Documentos de Académico
Documentos de Profesional
Documentos de Cultura
by
Mitra Mirhosseini
Doctor of Philosophy
Mitra Mirhosseini
August 2014
Sydney, NSW, Australia
PLEASE TYPE
THE UNIVERSITY OF NEW SOUTH WALES
Thesis/Dissertation Sheet
Grid-connected photovoltaic power plants (GCPPPs) are affected by grid requirements and affect
the system by different control strategies. The GCPPP should support the system based on the
grid code requirements including FRT capability, reactive power injection during voltage sags and
power quality. From the GCPPP side, the main reasons for inverter disconnection are: loss of
synchronization, excessive ac currents, excessive dc voltage, and voltage rise in non-faulty
phases under unbalanced voltage sags due to the reactive current injection.
The thesis deals with large-scale GCPPPs studies covering both static and dynamic aspects. The
focus on the static studies is to support the grid voltages by the capability of the inverter to inject
reactive currents. Four different methods based on the grid codes are designed and applied to a
10-MVA GCPPP. Considering dynamic studies, different control strategies are proposed to address
the mentioned problems for inverter disconnection. The solutions are offered for three main
controller strategies: controlling only the positive-sequence, controlling positive- and negative-
sequences and individual control of the phase currents.
Dealing with the positive-sequence control of the currents, inverter disconnection issues are
addressed for single-stage dc-ac and two-stage dc-dc-ac conversions by applying a current limiter
and using an advanced phase-locked-loop (PLL). The difference in these two conversions is the
protection of the dc voltage from overvoltage. While the dc voltage in the single-stage conversion
is self-protected, three different methods are proposed to control the dc voltage in the two-stage
conversion under voltage sag conditions.
Limitations on using PI controllers in the current loops are demonstrated when controlling both
positive- and negative-sequences of the grid currents. An alternative method based on resonant
controllers is proposed with the capability to operate without a PLL for grid synchronization.
The individual control of the phases is proposed, allowing the injection of a different reactive
current to each phase. This prevents from overvoltage in the non-faulty phases during unbalanced
voltage sags. For this control strategy a frequency-adaptive PLL is also designed that enables the
extraction of the individual phase angles. Finally, the results confirm the viability of the proposed
solutions using both simulation and experiments.
Declaration relating to disposition of project thesis/dissertation
I hereby grant to the University of New South Wales or its agents the right to archive and to make available my thesis or dissertation in whole or in
part in the University libraries in all forms of media, now or here after known, subject to the provisions of the Copyright Act 1968. I retain all
property rights, such as patent rights. I also retain the right to use in future works (such as articles or books) all or part of this thesis or dissertation.
I also authorise University Microfilms to use the 350 word abstract of my thesis in Dissertation Abstracts International (this is applicable to doctoral
theses only).
29/08/2014
Date
..
Signature
Witness
The University recognises that there may be exceptional circumstances requiring restrictions on copying or conditions on use. Requests for
restriction for a period of up to 2 years must be made in writing. Requests for a longer period of restriction may be considered in exceptional
circumstances and require the approval of the Dean of Graduate Research.
Signed:
Date: 29/08/2014
i
Abstract
Photovoltaic (PV) power plants have experienced remarkable growth over the past
decade and are on the way to become a mature and mainstream source of electricity.
Most of the PV power plants are connected to the electrical grid. Grid-connected PV
power plants (GCPPPs) are affected by the electrical system requirements and affect
the electrical system by different control strategies. GCPPPs need to meet specific
grid code requirements that include fault-ride-through (FRT) capability, reactive power
injection under voltage sag conditions, grid frequency support, power quality, etc. From
the GCPPP side, the main reasons for inverter disconnection are: loss of the grid voltage
synchronization, excessive ac currents, excessive dc voltage, and voltage rise in non-faulty
phases under unbalanced voltage sags due to the reactive current injection.
The thesis deals with large-scale GCPPPs studies covering both static and dynamic
aspects. The focus on the static studies is to support the grid voltages by the capability
of the PV inverter to inject reactive currents. Four different methods based on the grid
codes are designed and applied to a 10-MVA GCPPP. These methods are fixed reactive
power, fixed power factor (cos ), cos dependent of active power (cos (P )) and droop
control. Regarding dynamic studies, one of the recent requirements of the grid codes
for large-scale GCPPPs is not to be disconnected under some fault occurrences in the
electrical grid, i.e. the FRT capability requirement. All the disconnection issues men-
tioned above are considered in this thesis and different control strategies are proposed
to address the mentioned problems for inverter disconnection. The solutions include
three main controller strategies: controlling only the positive-sequences of the currents,
controlling positive- and negative-sequences of the currents and individual control of
grid phase voltages and currents.
panels, opening the circuit of the PV panels and controlling the PV panels operating
point to generate less power during the voltage sags. Moreover, under unbalanced voltage
sag conditions, the generated power as well as the dc-link voltage contain a second-
order harmonic oscillation, which is reflected into the active current reference signal. A
filtering technique based on moving average filters (MAFs) is applied to remove those
oscillations from the active current reference and thus to improve the quality of the
currents injected into the grid. Applying the MAF as a low-pass filter implies a delay in
the dc-link loop and may lead to instability. Therefore, a stability analysis is performed
to obtain the proper parameters for the proportional-integral (PI) controller considering
different environmental and voltage sag conditions.
Finally, individual control of each phase in three-phase grids is proposed for the cases
where the injection of the balanced reactive currents produces overvoltage in the non-
faulty phases during unbalanced voltage sags. In those cases, each phase has a different
requirement for reactive power injection. For this purpose a frequency-adaptive PLL is
also designed and proposed that enables the extraction of the individual phase angles.
The studies developed in this thesis are corroborated by simulation and experimental
results. The experimental results are obtained from a scaled-down GCPPP laboratory
prototype.
Acknowledgements
First of all, I would like to thank my God for letting me through all the difficulties. I
have experienced his guidance day by day. He is the one who let me finish my degree. I
will keep on trusting him forever.
A special thank to Dr. Baburaj Karanayil for his technical support, and also
Rosheila, Shamim, Amer, Ghias, Ali and all my friends in AERI for their support
during my studies.
I would like to express my deepest gratitude to my parents for their endless love and
affection. Their everlasting love was always the greatest support towards my success in
my PhD study. Their prayer for me was what sustained me thus far.
Last but not least, I would like to thank the Iran Ministry of Science, Research and
Technology for its financial support through scholarship and work opportunities.
iv
Contents
Originality Statement i
Abstract ii
Acknowledgements iv
List of Figures ix
List of Tables xv
1 Introduction 1
1.1 Research Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2.1 Static Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2.2 Dynamic Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2.2.1 Excessive AC Currents . . . . . . . . . . . . . . . . . . . 5
1.2.2.2 Excessive DC Voltage . . . . . . . . . . . . . . . . . . . . 6
1.2.2.3 Loss of Synchronization . . . . . . . . . . . . . . . . . . . 7
1.2.2.4 Overvoltage in Non-Faulty Phases . . . . . . . . . . . . . 7
1.2.3 Voltage and Current Control Strategies . . . . . . . . . . . . . . . 7
1.3 Thesis Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.4 Thesis Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.5 Tools Used . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.6 List of Publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.6.1 Journal Papers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.6.2 Conference Papers . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.6.3 Papers under Review . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.6.4 Association of Publications with the Chapters . . . . . . . . . . . . 12
1.6.5 Future Publications . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.7 Awards . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.8 Thesis Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
v
Contents vi
Bibliography 156
References 156
List of Figures
1.1 Power generation capacities added in the European Union (EU 28) in
2013 [1]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
ix
List of Figures x
4.1 Limiting curves of voltage at the grid connection point for PV power plant
in the event of a voltage sag in the grid. . . . . . . . . . . . . . . . . . . . 65
4.2 Principle of voltage support in the event of grid voltage sags. . . . . . . . 66
4.3 Performance of the SRF-PLL under LG with 60% voltage sag condition:
(a) input voltages, and (b) extracted angle. . . . . . . . . . . . . . . . . . 67
4.4 Diagram of the FSPLL. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
4.5 Performance of the FSPLL under a 60% LG voltage sag condition: (a)
input voltages and (b) extracted positive-sequence angle. . . . . . . . . . . 69
4.6 40% balanced voltage sag at the MV side of the transformer: (a) grid
voltages at the LV side, (b) active current reference, (c) reactive current
reference, and (d) grid currents at the LV side. . . . . . . . . . . . . . . . 71
4.7 Application of the current limiter to the reference currents. . . . . . . . . 71
4.8 Results after applying a current limiter with a 40% balanced voltage sag
at the MV side: (a) grid voltages at the LV side, (b) dc-link voltage, (c)
active current reference, and (d) grid currents at the LV side. . . . . . . . 72
4.9 Change in the PV array operating point under a voltage sag. . . . . . . . 72
4.10 PI controller with an anti-windup technique. . . . . . . . . . . . . . . . . 73
4.11 Results after applying an anti-windup technique to the voltage loop with
a 40% balanced voltage sag at the MV side: (a) grid voltages at the LV
side, (b) dc-link voltage, (c) active current reference, and (d) grid currents
at the LV side. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
List of Figures xi
4.12 Results for a 65% LG voltage sag in phase a at the MV side: (a) grid
voltages at the LV side, (b) voltage drop (c) reactive current reference,
(d) active current reference, and (e) grid currents at the LV side. . . . . . 75
4.13 Results for a 65% LG voltage sag in phase a at the MV side after applying
a MAF to the voltage measurement: (a) grid voltages at the LV side, (b)
voltage drop, (c) reactive current reference, (d) active current reference,
and (e) grid currents at the LV side. . . . . . . . . . . . . . . . . . . . . . 77
4.14 Results for a 30% LG voltage sag in phase a at the MV side after applying
the MAF for voltage measurement: (a) grid voltages at the LV side, (b)
voltage drop (c) reactive current reference, (d) active current reference,
and (e) grid currents at the LV side. . . . . . . . . . . . . . . . . . . . . . 78
4.15 Droop control diagram for the generation of the reactive current reference
based on the minimum rms value of the line-to-line voltages. . . . . . . . 78
4.16 Results for 65% LG voltage sag in phase a at the MV side with the
new voltage measurement: (a) grid voltages at LV side, (b) voltage drop
(c) reactive current reference, (d) active current reference, and (e) grid
currents at the LV side. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
4.17 Block diagram of a two-stage conversion GCPPP. . . . . . . . . . . . . . . 81
4.18 Control diagram of the dc-dc converter. . . . . . . . . . . . . . . . . . . . 81
4.19 Current path when short-circuiting the PV panels. . . . . . . . . . . . . . 82
4.20 Short-circuiting the PV panels when applying a 60% LG voltage sag at
the MV side of the transformer: (a) grid voltages at the LV side, (b) grid
currents, and (c) dc-link voltage. . . . . . . . . . . . . . . . . . . . . . . . 83
4.21 Short circuiting the PV panels: (a) overall generated power, (b) injected
active power, and (c) injected reactive power into the grid. . . . . . . . . . 83
4.22 Current paths in the dc-dc converter when turning on the switch: (a)
transition mode and (b) locked in state. . . . . . . . . . . . . . . . . . . . 84
4.23 Turning the dc-dc converter switch on when applying a 60% LG voltage
sag at the MV side: (a) grid voltages at the LV side, (b) grid currents,
and (c) dc-link voltage. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
4.24 P V characteristic and new power point under a voltage sag. . . . . . . 85
4.25 Adding a controller to the dc-dc converter to force the operating point to
move from the MPP to a lower power point. . . . . . . . . . . . . . . . . . 86
4.26 Triangle used to estimate the new operating point. . . . . . . . . . . . . . 87
4.27 Updated controller with feed-forward terms to enhance the dynamics of
the proposed controller. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
4.28 Control of the dc-dc converter to produce less power under a LG with
45% voltage sag at the MV side: (a) grid voltages, (b) grid currents, (c)
dc-link voltage, (d) input voltage of the dc-dc converter, (e) estimated
duty cycle, and (f) actual duty cycle. . . . . . . . . . . . . . . . . . . . . . 89
4.29 Control of the dc-dc converter to produce less power under voltage sag:
(a) grid voltages under a 3LG with 45% voltage sag at the MV side, (b)
related grid currents for G=300 W/m2 , and (c) related dc-link voltage;
(d) grid voltages under a LG with 65% voltage sag at the MV side, (e)
related grid currents for G=1000 W/m2 , (f) related dc-link voltage, (g)
related grid currents under G=300 W/m2 , and (h) related dc-link voltage. 91
4.30 Grid voltages exceeding due to symmetrical reactive current injection
(magnified peak voltages). . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
List of Figures xii
4.31 Application of a controller to limit the rise of the voltage in phases: (a)
reactive current reference and (b) grid voltages (magnified peak voltages). 92
4.32 Control diagram limiting the reactive current lower boundary. . . . . . . . 92
4.33 Example of a 48% 2LG voltage sag at the MV side of the transformer
under solar irradiation G=300 W/m2 : (a) voltages at the LV side, (b)
active current reference, and (c) output currents. . . . . . . . . . . . . . . 94
4.34 Inclusion of a MAF in dc-link loop. . . . . . . . . . . . . . . . . . . . . . . 94
4.35 Example of a 48% LG voltage drop with solar irradiation G=300 W/m2
when operating with a MAF in the voltage loop: (a) voltages at the LV
side, (b) dc-link voltage, (c) active current reference before the limiter,
(d) limited active current reference, and (e) grid currents. . . . . . . . . . 95
4.36 Bode plot of the MAF with a window width of 10 ms: (a) actual MAF
and (b) MAF approximations. . . . . . . . . . . . . . . . . . . . . . . . . . 96
4.37 closed-loop dc-link voltage control diagram including a MAF. . . . . . . . 96
4.38 PI control parameters: (a) for different voltage sag intensities, (b) for
different reactive current injections, (c) for different solar irradiation con-
ditions, (d) PI parameters without the MAF under rated grid voltages,
and (e) final stable region also considering optimal operation without the
MAF. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
4.39 Operating with a MAF in dc-link control loop with a 48% 2LG volt-
age drop under solar irradiation G=300W/m2 when choosing proper PI
parameters: (a) voltages at the LV side, (b) dc-link voltage, (c) active
reference current, and (d) grid currents. . . . . . . . . . . . . . . . . . . . 102
4.40 Final stable area of the dc-link voltage PI controller for the experimental
setup. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
4.41 Experimental results for the VSI under a LG fault with 100% voltage sag
in phase a at the grid side of the transformer: (a) without a MAF and
(b) including a MAF in the dc-link voltage control loop. . . . . . . . . . . 104
5.12 Results with current control loops based on PR controllers: (a) grid volt-
ages, (b) grid currents, and (c) injected active and reactive powers into
the grid. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
5.13 Performance of the VSI when applying the power limiter: (a) grid volt-
ages, (b) grid currents, (c) dc-link voltage, and (d) injected active and
reactive powers into the grid. . . . . . . . . . . . . . . . . . . . . . . . . . 120
5.14 Diagram of the proposed current limiter based on re-scaling the current
references. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
5.15 Application of the proposed current limiter: (a) grid voltages, (b) grid
currents, (c) re-scaling factor frs , and (d) injected active and reactive
powers into the grid. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
5.16 PI-based current control of the positive- and negative-sequences under
a 55% LG voltage sag at the MV side. From top to the bottom: grid
voltages at the LV side, output currents, injected active and reactive
powers into the grid. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
5.17 PI-based current control with the improved feed-forward term under a
55% LG voltage sag at the MV side. From top to the bottom: grid volt-
ages at the LV side, output currents, injected active power and injected
reactive powers into the grid. . . . . . . . . . . . . . . . . . . . . . . . . . 123
5.18 PR-based current control under a 55% LG voltage sag at the MV side.
From top to the bottom: grid voltages at the LV side, output currents,
injected active and reactive powers into the grid. . . . . . . . . . . . . . . 124
5.19 PR-based current control with power and current limiters under a 55%
LG voltage sag at the MV side. From top to the bottom: grid voltages
at the LV side, output currents, injected active and reactive powers into
the grid. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
5.20 Performance of the GCPPP including a FSPLL for phase angle detection.
Controlling the active power injected into the grid under a LG voltage
sag at the MV side: (a) voltages at the LV side of the transformer, (b) dq
components of the positive-sequence voltages, (c) dq components of the
negative-sequence voltages, (d) dq components of the positive-sequence
current references, (e) dq component of the negative-sequence current
references, (f) dc-link voltage, (g) grid currents at the LV side, (h) active
and reactive powers delivered to the filter and the grid, and (i) active and
reactive powers delivered to the grid. . . . . . . . . . . . . . . . . . . . . 127
5.21 Performance of the GCPPP without PLL for phase angle detection. Con-
trolling the active power injected into the grid under a LG voltage sag
at the MV side: (a) voltages at the LV side of the transformer, (b) dq
components of the positive-sequence voltages, (c) dq components of the
negative-sequence voltages, (d) dq components of the positive-sequence
current references, (e) dq component of the negative-sequence current ref-
erences, (f) dc-link voltage, (g) grid currents at the LV side, (h) active
and reactive powers delivered to the filter and the grid, and (i) active and
reactive powers delivered to the grid. . . . . . . . . . . . . . . . . . . . . . 128
List of Figures xiv
5.22 Performance of the GCPPP without PLL angle detector. Controlling the
active power delivered to the grid filter under a LG voltage sag at the MV
side: (a) voltage at the LV side of the transformer, (b) dc-link voltage,
(c) grid currents at the LV side, (d) active and reactive powers delivered
to the filter and the grid, and (e) active and reactive powers delivered to
the grid. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
5.23 The procedure of limiting the excessive ac currents by changing the op-
erating point of the PV array. . . . . . . . . . . . . . . . . . . . . . . . . . 130
5.24 Performance of the GCPPP after applying the current limiter when con-
trolling the active power injected into the grid under a LG voltage sag:
(a) voltage at the LV side of the transformer, (b) difference between the
maximum rms measured current and the maximum acceptable (Ier ), (c)
dc-link voltage, (d) grid currents at the LV side, and (e) active and reac-
tive powers delivered to the grid. . . . . . . . . . . . . . . . . . . . . . . 131
6.1 Proposed individual phase angle detector based on the FSPLL. . . . . . . 135
6.2 Linearized SRF-PLL when the input frequency is different from the nom-
inal value. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
6.3 Illustration of applying the slope limiter to the detected frequency (fo ) to
obtain an improved frequency (foimp ) under frequency transients. . . . . 140
6.4 Proposed PLL under unbalanced voltages. From top to the bottom: volt-
ages with a 2LG voltage sag (40% in phase a and 60% in phase b), de-
tected angles of phases a, b, and c, and the summation of the extracted
d- components of three independent phase voltages. . . . . . . . . . . . . . 141
6.5 Proposed frequency detector performance when the frequency jumps from
56 Hz to 47.5 Hz and vice versa. From top to the bottom: voltages with
the frequency changes, detected frequency, detected angles of phases a, b,
and c. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
6.6 Performance of the overall system with the slope limiter in the frequency
detector. From top to the bottom: voltages under four consequent dif-
ferent conditions (normal conditions, frequency change from 56 Hz to
47.5 Hz, phase jump of +/6 in phase b, and 40% voltage drop in phase
a of the input voltages), measured frequency (foimp ), detected angles of
phases a, b, and c. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
6.7 Control diagram to obtain the active and reactive current references. . . . 143
6.8 Current reference generation for phase a. . . . . . . . . . . . . . . . . . . . 144
6.9 Control diagram to re-scale the current references to avoid overcurrent. . 146
6.10 Current control loop with PR controllers. . . . . . . . . . . . . . . . . . . 146
6.11 Generation of a weak grid in the laboratory setup. . . . . . . . . . . . . . 147
6.12 Phasor diagram when the injected current is resistive-capacitive. . . . . . 147
6.13 Performance of the balanced control method for the GCPPP under a
100% LG voltage sag at the grid side of the transformer. From top to
bottom: grid voltages at the LV side of the transformer, output currents
at the LV side, and reactive current reference. . . . . . . . . . . . . . . . . 148
6.14 Performance of the proposed control method under a 100% LG voltage
sag at the grid side of the transformer. From top to bottom: grid voltages
at the LV side of the transformer, detected angles of phases a, b, and c,
generated reactive current references, and output currents at the LV side. 149
List of Tables
2.1 The different status of the inverter switches and corresponding output
voltages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
xv
List of Acronyms
Acronyms Definition
2LG Two-line-to-ground fault
3LG Three-line-to-ground fault
A/D Analog to digital converter
ac Alternative current
AERI Australian Energy Research Institute
CAPEX Capital expenditures
CB-PWM Carrier-based pulse-width modulation
CSI Current source inverter
D/A Digital to analog converter
dc Direct current
EHV Extra high voltage
EPLL Enhanced phase-locked-loop
ESS Energy storage system
FACTS Flexible ac transmission system
FRT Fault-ride-through
FSPLL Filtered-sequence phase-locked-loop
GC Grid code
GCPPP Grid-connected photovoltaic power plant
GW Gigawatt
HV High Voltage
I/O Input/output
IncCond Incremental Conductance
kVA Kilo volt-ampere
LCOE Levelized cost of electricity
xvi
Acronyms xvii
LG Single-line-to-ground fault
LV Low voltage
MAF Moving average filter
MPP Maximum power point
MPPT Maximum power point tracking
MV Medium voltage
MVA Mega volt-ampere
MVAr Mega volt-ampere reactive
MW Megawatt
NPV Net present value
OPEX Operational expenditures
P&O Perturb and observe
p.u. Per unit
PCC Point of common coupling
PI Proportional-integral
PI-R Proportional-integral-resonant
PLL Phase-locked-loop
PR Proportional-resonant
PV Photovoltaic
PWM Pulse-width modulation
rms Root mean square
SCP Short circuit power
SPWM Sinusoidal pulse-width modulation
SRF Synchronous reference frame
STC Standard Test condition
SV-PWM Space-vector pulse-width modulation
TSO Transmission system operator
VSI Voltage source inverter
List of Symbols
Symbol Definition
C Dc-link capacitance
d Duty cycle
deL Voltage drop with respect to the nominal rms value of the grid 1 voltage
dex Phase voltage drop with respect to its nominal rms value for x = {a, b, c}
eabc Grid voltages
edq dq component of the grid voltages in SRF
ei Imaginary component of the three-phase grid voltages
eL Three-phase rms voltage of the grid
eLmin Minimum rms value of line-to-line voltages of the grid
eLrms Rms value of line-to line voltages of the grid
Emax Maximum amplitude of the phase voltages of the grid
En Nominal three-phase rms value of the grid voltages
Enph Nominal single-phase rms value of the grid voltage
Enph Nominal single-phase amplitude of the grid voltage
er Real component of the three-phase voltages of the grid
Ermsph Rms value of the single-phase voltage of the grid
ex Grid voltage of phase x for x = {a, b, c}
Ex Grid voltage amplitude of phase x for x = {a, b, c}
f Frequency of the grid
fn Nominal frequency of the grid
frs Re-scaling factor
G Solar irradiation
iabc Grid currents
1
Note: All the grid measurements have been performed with respect to the low-voltage side of the
transformer.
xviii
Symbols xix
N Natural frequency
c Cut-off frequency
Damping factor
Reference value of that variable
+ Positive-sequence value of that variable
Negative-sequence value of that variable
/ / components of that variable in stationary frame
d/q d/q components of that variable in SRF
d/q Filtered d/q components of that variable in SRF
Dedicated to my parents SeyedAhmad and Mahin and my husband
SeyedHamed,
for their everlasting love, patience and support...
xxii
Chapter 1
Introduction
With at least 38.4 gigawatts (GWs) of newly-installed solar photovoltaic (PV) capacity
worldwide and a global cumulative installed capacity of 138.9 GWs, 2013 was a historic
year for solar PV technology [1]. The cumulative installed capacity refers to installations
that can make a real contribution to meeting energy demand.
Solar PV energy power plants are also one of the fastest growing types of renew-
able energy sources being integrated into the electrical grid. The increased efforts in
the semiconductor material technology resulted in the appearance of utility-scale PV
modules and consequently made the PV power plants an important alternative energy
source. In top positions in the european union (EU) in terms of new added capacity,
the PV ranked the second place after wind power with more than 10 GWs connected
to the grid in 2013. These two renewable electricity sources together beat gas and all
other sources of electricity [1] in 2013 as depicted in Fig. 1.1.
Solar PV energy is becoming one of the most popular renewable resources which has
experienced remarkable growth over the past decade and is on the way to becoming a
mature and mainstream source of electricity generation. Most of the PV power plants
are connected to the grid. However, because of the intermittent nature of these gener-
ation plants, they bring about new operational issues to the electricity grid, especially
considering the increasing penetration level all over the power system. One-third of
the land surface of the earth is covered by very dry desserts with high levels of solar
irradiation. If only 4% of these areas were used to install PV power plants, the resulting
annual energy generation would match the annual energy consumption [3]. However,
1
Chapter 1. Introduction 2
Figure 1.1: Power generation capacities added in the European Union (EU 28) in
2013 [1].
there are several technical requirements from the grid side that need to be fulfilled to
make grid-connected PV power plants (GCPPPs) reliable. This thesis reviews the most
important interconnection issues that have to be dealt with and proposes remedies to
improve the performance of the GCPPPs. The main motivation for this thesis is to con-
tribute to the technology of the GCPPPs in order to help to extend the implementation
of such clean electrical power generation in the world.
The studies of large-scale GCPPPs can be divided into static and dynamic operating
conditions. This classification is based on the requirements defined by the transmission
system operators (TSOs) in the national and international obligations known as grid
codes (GCs) [4, 5]. Besides the generation of active power to meet the demand from
the grid, GCPPPs should be able to support the grid voltages by the injecting reactive
power during voltage sag processes. In this thesis, no specific equipment such as storage
devices or reactive power compensators are considered to be used in GCPPPs to meet the
GC requirements. Therefore, the controllers developed to control the GCPPPs should
address such requirements.
Regarding the dynamic performance, which is the major focus of the studies devel-
oped in this thesis, there are several issues to be addressed in order to maintain the
reliability of both the PV power plant and the electrical grid. Among those issues, the
occurrence of voltage sags in the grid and therefore at the point of common coupling
(PCC) is one of the most challenging issues. The voltage source inverters (VSIs) used
as interfaces with the grid in PV power plants can be damaged during extreme grid
voltage disturbances. In the past, all GCPPPs connected to the medium voltage (MV)
grids were forced to be disconnected during voltage sags. However, as the amount of
GCPPPs has increased significantly in the recent years, nowadays the requirements for
Chapter 1. Introduction 3
these generating sources are more demanding and they have to stay connected during
some grid events besides providing support to the grid stability, which is known as
fault-ride-through (FRT) capability. The grid voltage support is provided by injecting
reactive power. On the other hand, some grid voltage abnormalities at the PCC may
have a direct impact on the protection of the GCPPP inverter that lead to VSI discon-
nection. The three main reasons for inverter disconnection are loss of synchronization,
excessive ac currents, and excessive dc-link voltage. Moreover, under unbalanced voltage
sags, the injection of reactive power should be limited so that the voltages of the non-
faulty phases do not exceed 110% of their nominal value. Therefore, the main target of
this thesis is to propose some control strategies that avoid inverter disconnection during
voltage sags and consequently provide the GCPPPs with FRT capability.
One of the most important consequences of voltage sags at the PCC is the excessive
currents produced in the GCPPPs. There are several active and passive methods to limit
these currents. The focus of this thesis is based on the active methods; therefore some
new techniques to limit the faulty currents are elaborated considering both active and
reactive currents. The reactive current is controlled based on the droop control [5] and
the active current is generally the one that is limited during a voltage sag. Another issue
is that when the current is limited during the fault, the excess of energy from the PV is
accumulated into the dc-link capacitor and forces the dc-link voltage to increase. If this
dc-link voltage increase exceeds the maximum acceptable by the inverter, a protection
is activated that leads to inverter disconnection. This issue is addressed in this thesis
for both single- and two-stage (with a dc-dc converter) inverter-based GCPPPs.
In order to apply different control strategies to address the FRT capability, the first
step is to extract the angle of the grid voltages properly. Under normal conditions and
balanced voltage sags the conventional phase-locked-loop (PLL) known as synchronous
reference frame PLL (SRF-PLL) can extract the angle of the voltages accurately. How-
ever, under unbalanced voltage sags, an advanced PLL is needed to extract the positive
sequence of the voltage. A method based on moving average filters (MAFs) is used for
this purpose in this thesis which was proposed in [6].
If a set of balanced reactive currents are injected to the grid during an unbalanced
voltage sag to support the grid voltages, the voltage in non-faulty phases may exceed the
110% of its nominal value which activates the over-voltage protection and disconnect the
GCPPP from the grid. As a result, two methods are developed in this thesis to address
this issue.
Based on the targets that the TSOs set in the interconnection of PVs with the grid,
different control methods can be implemented for the current control loops: (i) positive-
sequence control, (ii) both positive- and negative-sequence control, and (iii) individual
Chapter 1. Introduction 4
phase control. The first control method can achieve injecting balanced currents into
the grid even under unbalanced voltage sags. In the second control method, constant
active power can be injected into the grid, as well as constant power from the dc side,
helping to attenuate the low-frequency oscillations that appear in the dc-link voltage
during unbalanced voltage sags. In this case, the negative-sequence of the grid currents
needs to be controlled. The third control method is for the cases where the individual
control of each of the phases is needed. One example of this is to avoid overvoltage in
the non-faulty phases during unbalanced voltage sags. This solution needs a PLL able to
extract individual phase angles. In this thesis, a new method to extract the individual
phase angles is proposed.
The research documented in this thesis is divided into two key parts:
dynamic studies,
Regarding static studies for GCPPPs, one of the main focus is on the maximum power
point tracking (MPPT) technologies. The key information on MPPT technologies can
be found in [713]. Some methods to estimate the solar irradiation were proposed in [14]
using suitable re-parametrization of the current-voltage characteristic, and in [15] based
on a wavelet-neural network. Some research was reported on static voltage stability
using active power-voltage (P V ) and voltage-reactive power (V Q) curves [1620].
In the new German GCs [4, 5], four methods of reactive power control to support the
static grid have been defined: fixed reactive power, fixed power factor, power factor
based on the active power, and droop control. In Chapter 3, these methods are applied
to a large-scale GCPPP to evaluate their performance.
Because of the high penetration of GCPPPs, these plants should contribute to the
abnormalities that occur in the electrical grid known as FRT capability. The study
Chapter 1. Introduction 5
in [21] proposed the FRT enhancement by overloading the capability of the grid-tied
inverter to regulate the negative-sequence voltages and currents. In [22], some research
was done on transient stability of GCPPPs in a weak grid under challenging operating
conditions with a dynamic induction motor load and a battery energy storage system
(ESS).
Some abnormalities in the grid such as voltage sags may stop the functionality of the
GCPPPs due to the activation of the VSI protection. Many studies so far have focused
on addressing the different aspects of this issue. There are four main reasons for the
VSIs to disconnect from the grid:
excessive ac currents,
excessive dc voltage,
Regarding the limitation of excessive currents, two main methods have been proposed:
passive and active current limiters. Considering passive controllers to limit the fault
currents (requiring additional hardware), a method using a high temperature resistive
superconductive-based fault current limiter was introduced in [23, 24]. Another method
reported in [25] was based on a resonant-type current limiter which consists of a capacitor
and a reactor, where the inductance of the reactor was controlled by a spark gap. Flexible
ac transmission system (FACTS) devices have also been used extensively in this category
considering different approaches such those proposed in [2628].
Regarding the active methods, which involve developing control strategies to limit
the excessive currents, a classic strategy was proposed in [29] by limiting the gener-
ated power. This limitation was calculated based on a function of the positive-sequence
voltage and the maximum inverter current. This method resulted in a reduction in the
average power delivered by the inverter. However, the support of the grid voltages by the
reactive power was not considered. Another study was reported on regulating the cur-
rents according to a Lyapunov function to control the active and reactive powers under
unbalanced grid voltage conditions [30]. In [31], three control algorithms were applied to
extract continuous values of control parameters in order to achieve intermediate power
quality characteristics. In that work, the injection of the reactive power was not consid-
ered and although low currents could be programmed, the control did not guarantee its
Chapter 1. Introduction 6
minimum value. To deal with the mentioned problem, the authors in [32] introduced a
method that ensured minimum peak values in the grid-injected currents. However, the
currents produced were highly distorted compared with the methods mentioned in [31].
Another study [33] implemented the reactive current injection using a droop control.
However, the study only addressed symmetrical three-phase voltage sags in the grid. It
should be mentioned that in the current source inverter (CSI) configuration, the output
currents remain limited in all types of voltage sags due to the implementation of the
converter as a current source [34, 35].
In this thesis, different methods of protecting the GCPPPs from overcurrents are
proposed which take into account the injection of reactive currents under balanced and
unbalanced grid voltage sags. One method to limit the grid currents in the case of
controlling only the positive sequence of the currents is proposed in Chapter 4. Two
other methods of limiting the ac currents are developed in Chapter 5 when controlling
both the positive and negative sequences of the grid currents.
Losing the grid voltage synchronization is also one of the main reasons for inverter dis-
connection. PLL-based methods are extensively used for grid voltage synchronization,
and the most extended one is the SRF-PLL [38]. However, it has been demonstrated
that the conventional SRF-PLL configuration does not perform well under asymmetri-
cal faults and consequently leads to inverter being disconnected from the grid. Different
phase detection methods have been proposed so far. Some approaches were based on
zero-crossing detection such as those described in [3941]. Extracting the positive se-
quence of the grid voltages is another technique that has been extensively used [6, 4246].
Some other methods can be sorted based on cascaded delayed signal cancellation [47],
enhanced phase-locked loop (EPLL) [48] and a novel low-pass notch filter PLL using the
fast Fourier transform concept [49]. In this thesis, the method introduced in [6] is used
to extract the positive-sequence angle of the distorted grid voltages.
Last but not least is the protection of the inverter from grid overvoltage in the non-faulty
phases when injecting reactive power into the grid. Little research has been reported
in this area. In [50, 51], a flexible voltage support method was documented based
on the type and severity of the voltage sags. For this purpose, the amount of reactive
power injected via positive and negative sequences was controlled with an off-line control
parameter. Regarding wind power systems, a strategy to feed asymmetrical reactive
currents into the distorted grid was demonstrated in [52] which is capable of protecting
the non-faulty phases from overvoltage. However, the method was not generalized for all
types of voltage sags. In this thesis, two methods are proposed: one is based on injecting
balanced reactive currents and limiting them to avoid overvoltage, which is presented
in Chapter 4, and the other one is based on supporting the unbalanced voltages with
unbalanced currents, presented in Chapter 6.
The remedies to avoid inverter disconnection depend on the configuration of the voltage
and current loops in the grid-connected VSIs. Regarding the voltage loop, a study in [53]
discussed design issues of a dc-link voltage controller for VSIs. Parameter sensitivity, fre-
quency response, and load-disturbance rejection were investigated in [53] and compared
for two different control designs, double- and single-pole systems. In [54], a proportional-
integral-resonant (PI-R) controller-based, space vector modulated direct power control
Chapter 1. Introduction 8
topology was proposed to suppress the dc-bus voltage ripple under unbalanced condi-
tions, and also for controlling the instantaneous power of the VSI. A predictive control
scheme of the dc voltage for single-stage three-phase GCPPPs was presented in [55]
based on the analysis of the energy balance relationship in one control period. However,
the accuracy of parameters affected the performance of the proposed control method.
With regards to two-stage grid connected VSIs, the dc bus voltage is very vulnerable to
sudden changes in the output power. Therefore, a first derivative element in series to the
steady-state feedforward compensation is suggested in [56] to compensate for the delays
in the control response. However, it was only focused on sudden changes of the power at
the dc side. Another study in [57] for two-stage GCPPPs presented a control strategy to
eliminate dc voltage sensors while not downgrading the system performance with regard
to the MPPT accuracy and ac output quality. In this thesis, a conventional voltage
control loop is implemented with some modifications to enhance the performance of the
controller. These modifications include using anti-windup techniques, moving average
filters in the dc-link voltage loop to remove second-order harmonic in the active current
reference, and changing the reference of the dc-link voltage under voltage sag conditions
in the two-stage configuration.
The current control loops have been widely investigated from several aspects. From
the control strategy point of view, some of the common approaches were based on the
conventional synchronous PI controllers [5861], proportional-resonant (PR) controllers
[6264], and adaptive controllers [6569]. Some other studies were based on the com-
bination of those control strategies [7074]. Other concepts such as the use of a linear
quadratic regulator (LQR) [75], fuzzy logic controller [76], and deadbeat current con-
trollers [77, 78] were also developed for the current control loops.
To propose novel control strategies for large-scale GCPPPs to support the grid
voltages under various voltage sag events while maintaining the connectivity of
the inverter to the grid.
Chapter 1. Introduction 9
The methods have been developed further based on the requirement for power
factor conservation.
In Chapter 4, the GCPPP is developed to address the FRT requirements of the GCs
considering the positive sequence control of the grid currents. A current limiter
is proposed to protect the GCPPP from ac overcurrents. From the dc side over-
voltage protection, it is demonstrated that the single-stage VSI-based conversion
is self-protected while the two-stage conversion is not. Therefore, three different
methods are proposed to protect the dc voltage from overvoltage in the two-stage
configuration. The first two methods are based on not generating any power from
the PV arrays during the voltage sags, while the third one is based on changing
the operating point of the PV arrays to inject less power into the grid compared
with the pre-fault condition. As the dc-link voltage oscillates under unbalanced
voltage sags, a filtering technique to attenuate/remove those oscillations is used
based on a MAF. The main drawback of including the MAF in the dc-link control
loop is the delay produced by the filtering stage that may lead to instability. A
stability analysis is performed to obtain the proper controller parameters for the
dc-link control loop.
In Chapter 5, the positive and negative sequences of the grid currents are con-
trolled. It is demonstrated that the need for filtering the measured voltages and
currents for the current control loops compromises the stability of the GCPPP
and restricts significantly the values of the PI parameters. An alternative method
based on resonant controllers is proposed to enhance the dynamic performance of
Chapter 1. Introduction 10
the VSI. It is also demonstrated that the proposed controller can operate without
a PLL for grid voltage synchronization. Moreover, two new current limiters are
designed to protect the GCPPP from overcurrent. These methods are developed
based on the maximum acceptable rms value of the phase currents. The first
method is based on limiting the injected currents in two stages. First a power
limiter is implemented to limit the active power reference and then a re-scaling
factor is used to limit both the active and reactive current references. In the sec-
ond proposed method, the active power reference is limited in such way that the
output currents are limited to their nominal values during the voltage sags. This
is performed by increasing the reference of the dc-link voltage control to force the
PV array to generate less power.
The results presented in the thesis are based on simulation and experimental tests. For
the simulations, the following simulation platforms and models have been used:
DigSILENT PowerFactory software [79] is used for the static studies performed
in Chapter 3 in which a typical average model of 0.5-MVA GCPPP is embedded.
This model is modified further to address the static requirements of GCPPPs.
MATLAB/Simulink [80] software is used to model the grid, the PV arrays, and the
controllers. The PLECS Toolbox [81] is used to replicate the inverter with ideal
switches, dc link capacitor, grid filters, and the dc-dc converter. The simulation
results presented in Chapters 4-6 are based on using this software.
The experimental results where obtained from the following scaled-down laboratory
prototype:
The main results included in this thesis have already been presented and published in
different conferences and journals.
The following papers have been accepted for publication or have been published:
The following papers have been submitted to journals and are currently under review:
[12] M. Mirhosseini, J. Pou, and V. G. Agelidis, Power quality and stability consid-
erations in grid-connected photovoltaic power plants under various environmental
and voltage sag conditions submitted to IEEE Trans. Sustain. Energy
The journal and conference publications are associated with the chapters of this thesis
as follows:
The following paper based on the research developed in Chapter 6 of this thesis will be
submitted soon to the IEEE Transactions on Power Electronics:
Chapter 1. Introduction 13
Chapters Publications
Chapter 2. Grid-Connected Photovoltaic Power Plants
[7]
(GCPPPs) and Experimental Setup
Chapter 3. Performance of GCPPPs under Static
[8]
Conditions
Chapter 4. Performance of GCPPPs under Voltage Sags
[2], [3], [5], [6] and [12]
with Positive Sequence Control
Chapter 5. Performance of GCPPPs under Voltage Sags
[4], [9] and [10]
with Positive- and Negative-Sequence Control
Chapter 6. Performance of GCPPPs under Voltage Sags
[1] and [11]
with Individual Phase Control
1.7 Awards
Winner of the IEEE Australia Council 2013 Student Paper Contest for the category
Women in Engineering worth $300 with a certificate for the conference paper [3]
and the journal paper [2].
In Chapter 3, four different reactive power injection methods based on the German
GCs are developed to support the grid voltage under static conditions. These four
methods are, fixed reactive power, fixed power factor, power factor based on the
active power, and droop control. All these four methods are modified further to
maintain the power factor within acceptable limits based on the GC requirements.
Chapter 4 addresses inverter disconnection issues for both single-stage (dc-ac) and
two-stage (dc-dc and dc-ac) conversions considering only the positive-sequence con-
trol of the grid currents. A current limiter and an advanced PLL is implemented
Chapter 1. Introduction 14
with the capability to perform properly under unbalanced voltage sag conditions.
Regarding two-stage VSI-based GCPPPs, which are not self-protected against dc
overvoltage, three different remedies are proposed to maintain the connectivity of
the GCPPP under voltage sag conditions. As the dc-link voltage and consequently
the active current reference are distorted with second-order harmonics under un-
balanced voltage sag, a MAF technique is used in the dc-link loop to remove those
harmonics and enhance the power quality of the generated power. However, be-
cause of the delay introduced by the MAF, the dynamic of the dc-link voltage
is significantly affected. Therefore, a stability analysis is performed to determine
proper parameters for the dc-link controller to retain the connectivity of the in-
verter under various environmental and voltage sag conditions. Finally, a controller
is also proposed to limit the amount of balanced reactive current injection to avoid
overvoltage in the non-faulty phases.
Grid-Connected Photovoltaic
Power Plants and Experimental
Setup
2.1 Introduction
Nowadays, the need to decrease greenhouse gas emissions and the shortage of fossil fuel
resources make the role of renewable energy generation very important. Among all the
green energies resources, PV power plants attract a worldwide interest and they have
gained a huge popularity for the following reasons:
Sunlight is a free and abundant source of energy with no air contamination and
waste in the generation of electricity.
There has been a fast reduction in the cost of PV power plants in the last decade.
For example, the average price of PV power plants including PV inverter and
modules in 2014 has been reduced by 54.6% compared to that in 2010 [83].
The solar cell and solar module efficiency has been increased significantly. The
latest details on the efficiency of the PV cells and modules based on their materials
can be found in [84].
From the economical point of view, the cost of power production per kWh is deter-
mined according to the levelized cost of electricity (LCOE) which allows to calculate the
real cost of PV-generated electricity and to compare this with the cost of other sources
15
En
En I q min
+ Voltage
vdc Source N
Inverter
of electricity. Using LCOE makes it possible to compare a PV installation with any kind
of power plant [85]. For each system the LCOE calculation takes into account:
The location of the plant, which for a PV power plant is essential to consider the
difference in solar exposure.
where NPV stands for Net Present Value. As reported in [86] the LCOE for PV power
plants in 2013 has been declined almost 50% since 2009.
Solar PV technologies convert solar energy into electrical energy by directly absorbing
solar photonsparticles of the light. Generally, the PV power plants consist of a solar PV
array(s), a dc/dc converter (optional), a dc/ac converter, also known as an inverter, a
control system and an energy storage system (optional). In this thesis no energy storage
system is considered. A typical diagram of a GCPPP is depicted in Fig. 2.1. In the
following the main components of a GCPPP are briefly described.
The PV solar arrays consist of a series and/or parallel integration of solar modules,
depending on how much power is required. Solar modules are made up of several solar
cells wired together. Solar cells in turn are made up of semiconductor devices which
convert solar irradiation into electricity. PV modules are combined in series and parallel
connections to form an electrically and mechanically larger unit. The series-connected
Chapter 2. GCPPPs and Experimental Setup 17
I V curve
Pm , I m P V curve
Pmpp
I sc
I mpp
0
0 Vmpp Voc Vm
Figure 2.2: Characteristic curve of a PV module.
modules are called string and the number of PV modules in a string determines the
input dc voltage of the PV power plant.
The cost of PV power plants has decreased in the last decades and is now approaching
competitiveness. The last few years have witnessed more or less consistent declines in
the cost of modules and underlying components, pushing LCOEs lower and lower in a
market [86]. PV modules have already shown impressive price reductions, decreasing
by over 20% every time the cumulative sold volume of PV modules has doubled. The
average price of a PV module in Europe in July 2011 reached around 1.2 /W; this is
about 70% lower than 10 years ago [85].
One of the main characteristics of PV power plants is the high variability of their
output power. This variability stems from the fact that these power plants are static, and
thus, any instantaneous sun irradiation change leads to a corresponding variation in the
output power of the arrays. Thus, they are very sensitive to their ambient environment
[87]. PV module characteristics are usually given as current-voltage (I V ) and P V
characteristic curves. A typical characteristic of a PV module is depicted in Fig. 2.2.
The PV array characteristic is obtained by multiplying these curves by the required
number of series and parallel modules. The overall power of the PV array (Ppv ) under
standard test condition (STC) can be calculated as:
Ppv (ST C) = Ipv (ST C)Vpv (ST C) = (np Impp (ST C))(ns Vmpp (ST C)) (2.2)
where ns and np are the number of series and parallel modules, respectively, and STC
are the standard test condition defined as solar irradiation 1000 W/m2 and ambient
temperature 25 C. The variables Impp and Vmpp are the current and voltage at the
MPP, respectively, defined in the datasheet of the PV module as indicated in Fig. 2.2.
Chapter 2. GCPPPs and Experimental Setup 18
10 10
TM "! ,._.Rile
8 8
j
< 6 ..,< 6
!
~
. 4
2
IJ c
~
0 4
00 10 20 40 00 10 20 40
Volt6~!V. Volt.J~g~(VI
(a) (b)
Figure 2.3: I V characteristic changes due to: (a) variation in solar irradiation, and
(b) variation in temperature.
Generally, two factors can change the PV array output characteristics: cell temperature
and irradiation. Their impact on the I V curve is shown in Fig. 2.3.
The technologies of PV power plants are mainly categorized into the PV modules and
the power conversion characteristics. They are described in the following subsections.
2.3.1 PV Modules
PV cells are made of semiconductor materials. The main types of materials are crys-
talline and thin films, which vary from each other in terms of light absorption efficiency,
energy conversion efficiency, manufacturing technology, and cost of production. There
are many different types of solar cells. The major types are documented in the following
which are based on [8891] where more information is also available.
To produce monocrystalline silicon a crystal of silicon is grown from highly pure molten
silicon. The crystal obtained from such process is called an ingot and must be sliced,
using an extremely hard and fine saw, usually with a diamond tip. Very thin wafers are
obtained after sawing, then they are cut into a rectangular shape. They are the most
common, most efficient, most reliable and most expensive panels currently available.
Because of their high efficiency (15-20%), they are often used in applications where
there is environmental restrictions, giving the end user the maximum electrical output
for the installation area available.
Chapter 2. GCPPPs and Experimental Setup 19
Consisting of small grains of single-crystal silicon, polycrystalline PV cells are less energy
efficient than single-crystalline silicon PV cells. The grain boundaries in polycrystalline
silicon hinder the flow of electrons and reduce the power output of the cell. The silicon
is heated to a high temperature and cooled under controlled conditions as a mould. In
this method sawing is not required since the silicon is grown directly into sheets, and can
be one-third the thickness of regular crystalline cells since the polycrystalline material
is much stronger. It reflects the least and absorbs the most light. The energy conversion
efficiency for a commercial module made of polycrystalline silicon ranges between 12-
16%. They are the most commonly seen solar panels since they are less expensive that
the monocrystalline type.
Other thin film technologies using Cadmium telluride (CdTe) and Diselenide Cadmium
Telluride (CdCI2 ) are being developed and may have a higher efficiency than their silicon
counterparts. The conversion efficiency for a CdTe commercial module is about 7%,
almost similar to that of amorphous type. The instability of cell and module performance
is one of the major drawbacks of using CdTe for PV cells. Another disadvantage is that
cadmium is a toxic substance. CdCI2 has been one of the major research areas in the thin
film industry. CdCI2 is also one of the most light-absorbent semiconductors - 0.5 m
can absorb 90% of the solar spectrum. CdCI2 is an efficient but complex material. Its
efficiency in laboratory has reached to 17%. However, its complexity makes it difficult to
Chapter 2. GCPPPs and Experimental Setup 20
manufacture. Also, safety issues might be another concern in the manufacturing process
as it involves hydrogen selenide, an extremely toxic gas.
Compiled values of highest confirmed conversion efficiencies for research cells, from
1976 to the present, for a range of photovoltaic technologies can be found in Fig. 2.4
[92].
The conventional power conversion stage in PV power plants is the dc-ac power conver-
sion which is performed by the inverter(s). It may also include a dc-dc power conversion
stage which is performed by dc-dc converter(s). The dc-dc conversion is optional de-
pending on the selected configuration and the power rating of the GCPPP. These power
conversion stages are described in the following:
Solar panels offer dc output which is not compatible with the ac electricity system.
Therefore, PV power plants require inverters to convert dc to the standard sinusoidal
current form suitable for the grid. The inverter is usually designed for single- or three-
phase systems. As the focus of this thesis is on the high power PV systems only the
three-phase inverter is studied.
The inverters can be classified into two categories: two-level and multilevel topologies
[9396]. A two-level converter can only generate two voltage levels at the outputs. The
Chapter 2. GCPPPs and Experimental Setup 21
sa sb sc
vdc +
2C
_j
2 va 0
(a ) vb 0
vdc ( 0) (b) vc 0
(c)
s- a sb sc
vdc +
2C _j
2
The conventional two-level inverter consists of two power switches per phase-leg,
adding to six switches in a three-phase inverter. The structure of a three-phase two-
level inverter is shown in Fig 2.5. The switches on each leg operate in a complementary
mode, i.e., there is only one switch activated at any time (on state), while the other
one is in the off state. This can be expressed as follows:
where sx is the control function of the upper switch that can take two vales, 1 and
0 when the switch is in the on and off states, respectively.
Chapter 2. GCPPPs and Experimental Setup 22
Table 2.1: The different status of the inverter switches and corresponding output
voltages
The different states of the switches with the corresponding output phase and line-
to-line voltages are presented in Table 2.1.
Many PWM techniques have been developed for power inverters, some of them are based
on comparing the reference signals with a triangular carrier that has a significantly higher
frequency. These techniques are usually reffered as carrier-based PWM (CB-PWM). If
the reference signals are sinusoidal, the CB-PWM is called sinusoidal PWM (SPWM).
In this case, the duration (width) of the pulses generated at the outputs of the inverter
changes according to the sinusoidal reference signal. Another way of modulating a three-
phase inverter is based on space-vector PWM (SV-PWM). SV-PWM is based on a vector
representation in the plane of the output voltages/currents of the inverter. Both
techniques are described in the following.
I. SPWM SPWM is one of the most popular PWM techniques for harmonic reduction
of inverters. In a three-phase system, three reference signals are compared with a trian-
gular carrier. The results of this comparison determines the on or off states of the
power switches. Every time that a reference signal crosses the carrier a switching tran-
sition is produced in the corresponding phase-leg. Therefore, the switching frequency of
the power devices is essentially dependent on the carrier frequency. If the carrier fre-
quency increases, the switching frequency of the power devices increases proportionally,
and thus the switching power losses as well. On the other hand, the larger the carrier
frequency, the better quality of the output voltages. Therefore, there is a trade off when
Chapter 2. GCPPPs and Experimental Setup 23
va 0 vdc 2
vb 0 vdc 2
vc 0 vdc 2
vdc 2
vab
vdc 2
Amp.va 0 h
vdc 2
Amp.vabh
Figure 2.6: SPWM performance for a fundamental frequency of 50 HZ, carrier fre-
quency of 1650 Hz and modulation index of 0.8: (a) modulation signals and triangular
carrier signal, (b) normalized voltage of phase a va0 , (c) normalized voltage of phase
b vb0 , (d) normalized voltage of phase c vc0 , (e) normalized line-to-line voltage vab , (f)
harmonic spectra of a normalized phase voltage va0 , and (g) harmonic spectra of a
normalized line-to-line voltage vab .
choosing the carrier frequency between the quality of the output voltages generated and
the power losses of the inverter. The performance of the SPWM technique is illustrated
in Fig. 2.6 where the frequency has been normalized based on the fundamental output
frequency.
va* va* va 0
* *
vb v b
vb 0
vc 0
vc* vc*
v0
in Fig. 2.8. This zero-sequence waveform does not affect the fundamental components
of the line-to-line output voltages. The benefit of injecting this third harmonic is that
it reduces the amplitude of the reference signals while keeping the same fundamental
components. As a result, it extends the linear operation mode of the inverter beyond
m = 1 and up to 1.15.
2
Vk = (va +
a vb +
a 2 vc ), (2.4)
3
where
a = j2/3. The vector representation is depicted in Fig. 2.9. Out of the eight
vectors, two of them are null vectors (zero magnitude), while the other six are called
active vectors that define a hexagon and split it into six sectors. The magnitude of the
active vectors is 32 Vdc .
The reference vector Vref is generated on average by using at least three vectors
of the space vector diagram. The vectors selected are usually the two adjacent active
vectors and one (or the two) null vectors. Consider a reference vector located in one of
the six sectors as depicted in Fig. 2.10. Either of the null vectors (000) or (111) can be
Chapter 2. GCPPPs and Experimental Setup 25
va 0 vdc 2
vb 0 vdc 2
vc 0 vdc 2
vdc 2
vab
vdc 2
Amp.va 0 h
vdc 2
Amp.vabh
used as the third space vector (or both of them). The reference vector can be derived
as follows:
tk
tk+1
t07
Vref = V1 + V2 + V07 (2.5)
Ts Ts Ts
or
Vref = dk Vk + dk+1 Vk+1 + d07 V07 , (2.6)
where dk , dk+1 , and d07 are the duty cycles of the corresponding space vectors. The
duty cycles dk and dk+1 can be derived as:
sin k cos k
" #
dk 3 3 3
= (2.7)
dk+1 Vdc sin (k1) cos (k1)
3 3
Chapter 2. GCPPPs and Experimental Setup 26
V3 V2
010 110
vref M
V7
V4 V0
V1
6
011
V5 V6
Figure 2.9: Voltage vectors in SV-PWM and corresponding switching states.
Vk 1
ref
0
07 k
7
The maximum amplitude of the reference space vector as shown in Fig. 2.9 is calculated
as:
2 Vdc
VSV P W M M = VrefM = Vdc cos( ) = , (2.10)
3 6 3
Chapter 2. GCPPPs and Experimental Setup 27
and the ratio between the maximum fundamental output voltage amplitudes of SV-
PWM with regard to SPWM is:
V
dc
VSV P W M 3 2
= Vdc
= = 1.15 (2.11)
VSP W M 2
3
In this thesis, SPWM has been applied to the GCPPPs under study.
Dc-dc converter is an electronic circuit that converts a dc voltage level to another level.
The dc-dc conversion itself can be categorized in two different classifications: (i) step-up
or step-down dc-dc converters, and (ii) isolated or non-isolated dc-dc converters. The
following information is based on research in [98101] and more information can be
derived as well from these sources.
In this thesis, the boost converters are selected for the dc-dc conversion. Isolation is
provided by a 50 Hz transformer connected between the PV power plants and the grid.
The schematic of a boost converter is depicted in Fig. 2.11.
Chapter 2. GCPPPs and Experimental Setup 28
Li
v pv vdc
i pv
The solar PV power plants can be categorized into several classifications: off-grid and
grid-connected, small- and large-scale generation and different architectures. In this
section, each category is explained briefly and then the GCPPP is specifically elaborated
regarding different classifications in the next section.
PV power plants can be classified into off-grid and grid-connected systems. Off-grid
power plants are independent of the utility power grid. They generate electricity without
any connection to the power grid. Off-grid or namely stand-alone solar power structures
are usually installed on the rooftop of the buildings with the energy capacity of few
kWp (peak kW). Off-grid systems are most appropriate for remote locations where
the access to the electrical grid is impossible or too expensive. Off-grid PV power
plants operate independently from the local utility grid to provide electricity to a home,
building, boat, or remote agricultural pumps, etc. These systems typically require either
a battery bank (to store solar electricity for use during night-time or cloudy weather),
a backup source of energy (like a generator), or both. An off-grid PV power plant must
be large enough to produce enough electricity to cover 100% of the energy needs of the
building. In all off-grid scenarios, electrical usage must be monitored and kept below the
maximum output of the panels and batteries as there is no grid-source to supply excess
power. A typical off-grid PV power plant includes the PV modules, a bank of batteries,
charge controller (to protect the battery bank from overcharge), a controller, an inverter,
required electrical safety gear (i.e. fuses, breakers, disconnects), and monitoring system
to balance energy consumption with production. The advantages and disadvantages of
the off-grid configuration are listed below [102]:
Advantages:
Chapter 2. GCPPPs and Experimental Setup 29
Provides power independently of the utility grid - still having power when the grid
is down.
Can save significantly on initial cost if a long grid extension is needed to get to a
customer.
Disadvantages:
Higher ongoing time and cost- The cost of batteries over the time is significant.
The system must be large enough to supply full power needs during the lowest
sun part of the year (although a generator can be used to supplement during the
worst times).
The customer is the power company, and responsible for safe and reliable operation,
and maintenance.
GCPPPs are systems connected to a large independent electrical grid usually the public
electricity grid and feed power directly into the grid. In grid-connected systems the pub-
lic electricity grid functions as an energy storage. These systems are usually employed
in decentralised and centralized applications. Decentralised GCPPPs configuration in-
clude rooftop PV power plants, where the PV power plants are mounted on rooftops
of buildings and incorporated into the buildings integrated system. In the case of res-
idential or building mounted GCPPPs, the electricity demand of the building is served
by the GCPPP and the excess is fed into the grid. Their capacities are usually in the
lower range of kilowatts while the centralized GCPPPs configuration have capacities
ranging from the higher kWs to the MW range [103]. In centralized configuration, the
PV modules are serially string and several strings are connected in parallel to a single
dc-bus.
Chapter 2. GCPPPs and Experimental Setup 30
A typical GCPPP includes PV modules, inverter(s), electrical safety gear (i.e. fuses,
breakers, disconnects), and monitoring system to monitor energy production. The ad-
vantages and disadvantages of the grid-connected configuration is listed below [102]:
Advantages:
Lowest initial cost (because there is no need for batteries and charge controller).
Simplest to install.
Most efficient (because there are no losses associated with charging batteries).
The system can be small initially and can be extended at a later stage (with some
limitations).
Disadvantage(s):
The other category is a hybrid configuration which is GCPPP equipped with a battery
backup. The backup battery is charged by both the grid and the solar panels. In the
event of an outage, the backup battery will need to be switched on either manually or
with an automatic system to support the loads. This category has some advantages
when compared with the two previous categories and can provide backups under grid
outage. However, it involves with the batteries which are expensive and require services
and regular maintenance. As it is clear from the topic of this thesis, this category will
not be studied.
The grid-connected configuration has become the dominant part of the solar PV
installed in the world as shown in Fig. 2.12 [2]. Therefore, the focus of this thesis is on
the grid-connected configuration.
In this section the GCPPPs are described in detail considering different classifications
and then the basic control structures of a typical GCPPP are presented.
Chapter 2. GCPPPs and Experimental Setup 31
The GCPPPs can be sorted from different architecture aspects as described in the fol-
lowing.
Single-stage conversion means direct connection of the PV source to the dc side of the
VSI. The schematic of single-stage VSI conversion is depicted in Fig. 2.1. In this case
the dc-link voltage is equal to the total series voltages of the PV modules in the PV
array. Therefore, PV array sizing (series and parallel connections of the PV module)
are determined based on the rated power and maximum dc voltage of the inverter.
In the single-stage configuration the inverter performs the active and reactive currents
generation control and MPPT. The active current is controlled by regulating the dc-link
voltage. In this case, the voltage reference for the dc-link control loop is determined
by MPPT technique embedded in the control structure of the GCPPP which may be
changed under different environmental and voltage sag conditions.
Two-stage conversion means a dc-dc conversion or pre-regulator unit exists between the
PV source and the VSI. In some cases where the PV array output voltage does not
match the dc-link voltage, a dc-dc or several dc-dc converters are implemented between
the PV array and the inverter to transfer the generated power from the PV array to the
inverter. The different dc-dc conversion methods are described in Subsection 2.3.2.3.
Chapter 2. GCPPPs and Experimental Setup 32
vdc
Figure 2.13: A typical schematic of a two-stage VSI conversion for GCPPP applica-
tion.
A typical two-stage dc-dc-ac conversion for the GCPPP application is depicted in Fig.
2.13. In this configuration, the active and reactive currents generation are controlled by
the inverter while the dc-dc converter performs the MPPT to control the input voltage
of the dc-dc converter. In this case, the voltage reference of the dc-link loop remains
constant.
The sizes of the GCPPPs are determined based on the grid voltage level they are con-
nected to. The low PV power ratings with several kWs are those are connected to
low-voltage (LV) distribution network while the higher powers of tens of MWs such as
huge PV farms are connected to the high-voltage (HV) network. Generally, voltage
levels are sorted into LV, MV and HV and sometimes extra HV (EHV) grid. However,
these definitions are different in countries and there is no unique international standard
for that. For example in German GCs, the voltages lower than 1 kV are defined as LV,
between 1 kV and 60 kV are MV and higher than that are HV. In [104] the voltage
levels are categorized as: 100 V<LV< 1 kV, 1 kV<MV< 35 kV, 35 kV<MV< 230 kV
and 230 kV> EHV.
The minimum and the maximum power at which the PV power plant has to be
or can be connected to any of these voltage levels depends on the respective network
condition. Possible values for the GCPPPs to be connected to the MV grid are in the
range of 500 kW up to 100 MW in German GCs [5]. In some cases this can be only
determined by the calculation of the network operator.
The GCPPP can also be categorized in small-, medium-, large- and very large-scale
sizes. Small-scale are usually refers to those residential usages up to 10 kW, medium-
scale are for commercial, complexes and large buildings usage with the power rating of
10-1000 kW, large-scale (or utility-scale) are those connected to the conventional feeders
of the grid with the power rating of more than 1 MW and very large-scale are usually
ranging from 10 MW to several GWs and are applicable where there is a lot of space
available with high levels of solar irradiation like deserts [3]. The first two categories
Chapter 2. GCPPPs and Experimental Setup 33
PV Array
.;.
;: : :
: : :
.
;
. .
;: :: :: ; ...
' : : :
..
.
Centralized Grid
L 0 0 ~
Inverter
''''
PV Array
(a)
...
Inverter
... .. --- t\
... Grid
(b)
Figure 2.14: A typical schematic of (a) a centralized configuration, and (b) a decen-
tralized configuration.
are usually roof-mounted or integrated in the building while the last two categories are
ground-mounted. In this thesis the large-scale GCPPP is selected for the further studies.
The inverters of a GCPPP can be categorized into centralized and decentralized clas-
sifications. A typical schematic of both configurations is depicted in Fig. 2.14. When
determining the technology to be used based on these two classifications, many consid-
erations should be made. Installation capabilities, upfront and long-term costs, energy
production, maintenance, component availability, supplier stability and many more ul-
timately factors into the final decision [105].
Chapter 2. GCPPPs and Experimental Setup 34
In the centralized configuration, all the PV arrays are connected to one or limited number
of the inverter(s) located at one place. The benefits of the centralized configuration is
listed below based on the SMA report [106]:
Simple construction.
Further, another report [107] also indicated a number of other benefits such as better
maintenance, superior load management, improved security etc. that are associated with
the centralized configuration.
Maximum flexibility in plant design with regard to module types and their ar-
rangement.
In this thesis the central configuration is taken into the consideration. The perfor-
mance of a large-scale GCPPP with decentralized inverters will be the focus of research
of the author of this thesis in her future work.
The main control parts of a VSI-based GCPPP are: MPPT, dc-link voltage regulation,
ac currents control, and grid voltage synchronization. The MPPT generates the reference
for the dc-link voltage control loop. The current control loops are responsible for power
quality issues and current protection. Harmonic compensation and dynamics are the
important properties of the current controllers. The dc-link voltage controller is designed
for balancing the power flow in the system. Usually, the design of this controller aims
for system stability and has slow dynamics. The basic control strategies for each part is
briefly described in the following.
To extract maximum power (Pmpp ) from the PV modules, a MPPT technique needs to be
implemented. There are different MPPT techniques such as perturb and observe (P&O),
incremental conductance (IncCond), fractional open-circuit voltage, and fractional short-
circuit current. An introduction to these different technologies as well as a comparison
study on them was presented in [8, 108]. The more extended methods are P&O and
IncCond which are introduced in the following. The main advantages of these methods
are that they are generic, e.g., suitable for any PV array, they require no information
about the PV array, they work reasonably well under most conditions, and they are
simple to implement on a digital controller.
Chapter 2. GCPPPs and Experimental Setup 36
I. Perturb & Observe Method The P&O algorithms are widely used in control
of MPPT thanks to their simple structure and reduced number of necessary measured
parameters. As the name implies, the concept behind of this method is based on ob-
servation of PV array output power and its perturbation by changing the current or
the voltage of the PV array operation. The algorithm increments or decrements the
reference voltage based on the previous value of power until it reaches the MPP. The
flowchart of P&O algorithm is depicted in Fig. 2.15.
dP d(V I) dI
= =I +V = 0, (2.12)
dV dV dV
I dI
= . (2.13)
V dV
A set of inequalities can be derived from (2.13) that indicates whether the operating
voltage is above or below the MPP voltage. These relationships are summarized in
following:
dP I dI
<0 > (for V > Vmpp ) (2.14)
dV V dV
dP I dI
>0 < (for V < Vmpp ) (2.15)
dV V dV
The increase or decrease in Vref in both P&O and IncCond Methods can be applied
by fixed and variable step sizes [60, 110113]. Although the IncCond is slightly more
complicated to implement, it provides better performance than P&O under both static
and dynamic conditions. It has also been stated in the literature that the IncCond can
determine the position of the actual operating point relative to the MPP, and it can find
the distance to it. It can also stop perturbing when the MPP has been reached, thus
offering a superior performance compared with P&O [114] technique. Therefore, in this
thesis the IncCond technique has been implemented.
The dc-link voltage loop known as the outer loop in VSI-based GCPPP provides the
active current reference (idref ) for the current control loop (inner loop). Conventionally,
a PI controller is adopted to regulate the dc-link voltage. The schematic of a dc-link
voltage loop is depicted in Fig. 2.17.
In order to obtain the closed-loop block diagram of the Fig. 2.17, the small signal
model of the dc-link loop is obtained. The average model of the dc-link loop is depicted
in Fig. 2.18. The KCL law for the dc side can be written as:
dvdc vdc
iC = C = idc iin = ipv iin . (2.16)
dt Rp
As the dc-link voltage normally operates at the MPP, the maximum power is de-
livered to the ac side ignoring the inverter losses. The power delivered to the inverter
Chapter 2. GCPPPs and Experimental Setup 38
vdc
Figure 2.17: Schematic of dc-link voltage regulation.
Chapter 2. GCPPPs and Experimental Setup 39
idc iin
iC
i pv Rp vdc
C
is:
ppv = iin vdc , (2.17)
and the power injected from the inverter into the grid is:
pout = id ed + iq eq , (2.18)
where d and q subscripts are the transformed values of the three-phase components
into synchronous reference frame (SRF). The transformation used to convert from abc
coordinates to dq coordinates is known as the Park transformation:
id ia
h i
iq = Tdq0 ib , (2.19)
i0 ic
with
2 2
cos() cos( 3 ) cos( + 3 )
r
2
sin() sin( 2 2
[Tdq0 ] = 3 ) sin( + 3 )
, (2.20)
3
1 1 1
2 2 2
where is the angle of the grid voltage extracted by the synchronization technique
described later in this section.
Assuming 100% efficiency of the converter, (2.17) becomes equal to (2.18), obtaining:
id vd + iq vq
iin = . (2.21)
vdc
dvdc 1 vdc id vd + iq vq
= ipv . (2.22)
dt C CRp Cvdc
Chapter 2. GCPPPs and Experimental Setup 40
Vdc* s Id s Vdc s
Gp s
Linearizing the nonlinearity in (2.22) around the MPP, finally the small signal model
is derived as:
in which the subscript ss represents the steady-state value of that variable. Assuming
all the variables disturbances but the id component and converting (2.23) to s-domain:
1 Vdss
(s + M )Vdc (s) = Id (s) (2.24)
CRp CVdc
where
ns Vmpp
Rp = (2.26)
np Impp
Idss Vdss + Iqref Vqss
M= 2
(2.27)
CVdc
s 2
Vdc
EL 2 EL
Idss = + Idcss Iqref (2.28)
2R R 2R
Vdss = EL + RIdss LIqref , (2.29)
and I
and Vdc qref are the reference values of the dc-link voltage and reactive current
reference, respectively, while EL is the d component of the rms voltages of the grid and
L and R are the grid filter inductance and resistance, respectively.
Based on (2.25)-(2.30), the closed-loop block diagram of Fig. 2.17 in s-domain can
be obtained as demonstrated in Fig. 2.19
2
2N s + N
Gst = 2 , (2.31)
s2 + 2N s + N
did
Vd = Rid + L + Liq + ed (2.32)
dt
diq
Vq = Riq + L Lid + eq (2.33)
dt
where R and L are the grid filter impedance and inductance, respectively, ed and eq are
the transformed grid voltages to SRF and is the angular frequency of the grid.
Vd = vd + Liq + ed , (2.34)
Vq = vq Lid + eq . (2.35)
where
did
vd = Rid + L , (2.36)
dt
diq
vq = Riq + L . (2.37)
dt
Since the coupling terms (Liq and Liq ) and the grid voltages (ed and eq ) are
known, the current regulators only have to define the voltages drop on the grid filter
impedances (vd and vq ) to achieve the current references (idref and iqref ). The
schematic of the current regulator based on (2.36) and (2.37) is depicted in Fig. 2.20.
Compensating for the coupling terms, the closed-loop block diagram of the current
control is obtained as depicted in Fig. 2.21. The closed-loop transfer function of the
current control loop is also a second-order system. Therefore, the parameters of the
PI controller can be easily tuned when compared with a standard second-order transfer
function as stated in (2.31).
In the application of GCPPPs, one important issue is the phase synchronization with
the grid voltage by means of angle extraction techniques. The PLL method is one of
Chapter 2. GCPPPs and Experimental Setup 42
ed
idref kic vd vd
k pc
+ s
id
L
iq
L
iqref kic vq vq
+ k pc
s
eq
Figure 2.20: Control diagram of the current control loop.
.
l dref id
:~ +~ne.I I I I
1
Ls R
.
l qref iq
:~ l~ne.l I I I
1
Ls R
the key technologies extensively used in this regard and the conventional PLL is the
so-called SRF-PLL as shown in Fig. 2.22.
To obtain proper parameters of the PI controller for the SRF-PLL, the closed-loop
transfer function of the SRF-PLL should be obtained first. A balanced set of grid
voltages is expressed as:
cos t
eabc = E 2 , (2.38)
cos(t 3 )
2
cos(t + 3 )
in which E is the peak value of the phase voltage. The grid voltages eabc (2.38) are the
input of the SRF-PLL and the output of the SRF-PLL is the extracted angle . This
angle is used to transform the set of voltages in (2.38) into d-q components, as follows:
h i
edq0 = Tdq0 eabc , (2.39)
Chapter 2. GCPPPs and Experimental Setup 43
ed
eabc n
eq o
t 3
eq 1
E
2 s
where Tdq0 is defined in (2.20). The obtained grid voltages in dq coordinates are:
r
3
ed = E cos(t ) (2.40)
2
r
3
eq = E sin(t ). (2.41)
2
The SRF-PLL tries to impose the q component in (2.41) to be zero (eq =0). This
implies:
= t. (2.42)
The control loop in the SRF-PLL tries to keep close to t. Therefore, a simplification
can be applied to linearized (2.41) as follows:
r
3
eq = E(t ). (2.43)
2
q q
3 3
2 Ekp s + 2 Eki
Gcl = q q , (2.45)
3 3
s2 + 2 Ekp s + 2 Eki
Considering (2.46) and (2.47), the relationship between kp and kr can be calculated
based on defining = 1 as follows:
2
ki = 6Ekp2 , (2.48)
A fast transient response is needed for SRF-PLL. Defining the response time by:
4
tr = , (2.49)
N
4 4
tr = q kp = q . (2.50)
3 3
8 Ekp 8 Etr
The experimental results illustrated in this thesis have been obtained from the three-
phase GCPPP prototype shown in Fig. 2.24. The prototype can be categorized into
hardware and digital controller parts. The hardware parts of the prototype are:
3. Computer.
4. Oscilloscope.
Chapter 2. GCPPPs and Experimental Setup 45
5. Transformer used to step down the grid voltages from 415 V to 200 V.
6. Power Meter.
12. Dc-bus electrolytic capacitor of 1100 F built by two 2200 F capacitors connected
in series.
The digital controller part which is shown in Fig. 2.25 creates an interface between
the model of the GCPPP in MATLAB/Simulink in the computer and the hardware.
The controller is composed of several main parts which are:
Chapter 2. GCPPPs and Experimental Setup 46
Computer
2. DS1103 processor which provides the communication of the software with the
Digital input/output (I/O) interface and digital/analoge (D/A) interface.
3. Digital I/O interface used to generate high speed PWM signals through a digital
output port.
4. Digital/analoge (A/D) interface high speed board used to convert the measured
signals from analog to digital. The digital signals are further processed by the
DS1103 processor and send to the computer for data acquisition.
2.7 Summary
In this chapter, the basic structure of a PV power plant with the main emphasis on
large-scale GCPPPs has been elaborated. The different technologies, configurations,
and system structures have been described. This includes the PV modules and power
conversion technologies, solar PV configurations, and basic control structures of a VSI-
based GCPPP. Finally, the experimental setup prototype used for the experimental tests
in this thesis has been explained.
In the next chapter, a typical 10-MVA GCPPP is developed and equipped with
four different methods of supporting the grid with reactive power injection under static
conditions. These methods are modified further to maintain the power factor within the
acceptable boundary considering the GCs.
Chapter 3
3.1 Introduction
Modelling of large-scale GCPPPs is needed for different time frameworks to assess some
aspects of both component and system performance. This chapter presents extended
models of large-scale GCPPP with DigSILENT PowerFactory for static grid support
using different reactive power control strategies. GCs are taken into account to develop
the proposed models which are able to provide reactive power support under fixed reac-
tive power, fixed power factor (cos ), cos dependent of active power (cos (P )) and
droop control Q(E). Moreover, the models are developed further to address the require-
ments for power factor conservation. Selected simulation results of a 10-MW GCPPP
demonstrate the effectiveness of the developed models.
The integration issues of large-scale GCPPPs can be categorized into static and dynamic
ones. From the static point of view, several areas of research have been addressed
[16, 115117]. Considering voltage stability in static conditions, the impact of large-
scale PV penetration on the power system was investigated in [115] with MATLAB
software. In [116], a capability chart was defined in the P Q plane to find all possible
pairs of active and reactive power values (P, Q) that can be selected as the inputs of
the PV inverter control. From the grid support perspective, it was shown in [117] that
operating the PV inverter with a power factor different than the unity with a proper
47
Chapter 3. Performance of GCPPPs under Static Conditions 48
control, the requirements for voltage regulators and capacitor banks embedded in the
power system were reduced. In [16], the impact of large GCPPPs on the grid voltages
and the reactive power characteristics at the PCC were reported through a P V curve
method using DigSILENT software. The focus has been on the control mode of the
PV power plant, i.e., current-source or voltage-source mode. It should be noted that
in the above mentioned references, there is a lack of research on the various methods
of supporting the grid voltages under static conditions with reactive power injection
without using external reactive power compensation devices.
The DigSILENT PowerFactory 14.1 software has a model of GCPPP available. How-
ever, as it will be demonstrated later in this chapter, the current model requires some
modifications to address reactive power support under static conditions to meet the
GCs requirements. This chapter will fill the gap in the static reactive power support, by
developing a control system and applying it to the existing GCPPP model to operate
under four different methods of reactive power support as required by the GCs.
Electricity utilities must fulfill certain national and/or international obligations in sup-
plying electric power to customers. These obligations are known as GCs and are set by
TSOs. Several European countries are obliged to satisfy interconnection requirements
for large-scale GCPPPs based on power system operation and stability. Germany and
Spain are the leaders in utilizing PV power plants with respect to both the installation
and integration technologies. Therefore, they have the most comprehensive GCs in this
field. The studies in this chapter have been done according to the German GCs [4, 5] for
connecting PV power plants to a MV grid. The maximum size of a PV power plant to
be connected to a MV grid depends on the respective grid conditions and is determined
by the TSO. Generally, the GCs do not have a major impact on the design of the PV
inverters except that new control algorithms must be developed to satisfy them.
In MV grids, the grid support under static conditions implies the voltage stability
of the system under normal operation. This is done based on the system requirement or
on demand by the TSO to keep the voltage variation within acceptable limits. Besides
that, the GCPPPs should be capable of providing reactive power at any operating point
to keep the power factor within the boundary of 0.95 underexcited to 0.95 overexcited
at the PCC. The setpoint for reactive power control can be adjusted by the TSO via
each of the four methods of supporting reactive power in MV grids [5]: fixed Q, fixed
cos , power factor dependent of active power (cos (P )), and droop control Q(E). One
of these methods can be selected upon the decision of the TSO.
Chapter 3. Performance of GCPPPs under Static Conditions 49
Kxtemal <itid
};1V---...-....._-
LV - - - t - - -
l' V <l~neralor
Figure 3.1: PV template model in DigSILENT.
PV inverters are usually oversized to meet the requirements of the GCs so that in the
case of maximum power generation (Pmpp ), the Q requirement can always be achieved.
A reasonable oversizing is to set the apparent power of the inverter to be S=1.1Pmpp .
Therefore, the Q capability increases from zero to 45.8% of Pmpp [33].
Solar irradiation and temperature: the control system inputs are in W/m2 and
C, respectively.
PV model: calculates the current and voltage at the MPP for one module and
then for the whole array considering the number of parallel and series modules.
Power measurement: a device that measures the active and reactive power at the
LV bus.
Slow frequency measurement: a PLL that measures the frequency at the LV bus.
Active power reduction: a limiter of the generated power in the case of over-
frequency.
Phase measurement: a PLL device to derive the phase angle at the LV bus.
Controller: main part of the control in a rotational frame that adjusts the current
components idref and iqref to regulate the active and reactive power injected into
the grid. The reactive power support under dynamic conditions is done with a
droop controller embedded in the same controller.
In general, the input measurement devices, PV model, dc busbar, and capacitor model
are related to the dc side of the inverter and the remaining blocks refer to the ac side.
Chapter 3. Performance of GCPPPs under Static Conditions 51
The system considered in this study is a 10-MVA GCPPP as shown in Fig. 3.3, which
consists of 10 parallel 1-MVA PV generators connected to a MV grid through trans-
formers and an ac line. The systems specifications are provided in Table 3.1.
To connect this 10-MVA PV power plant to a stable external grid, the short circuit
power (SCP) of the external grid is selected to be 30 times the power of the PV plant,
Chapter 3. Performance of GCPPPs under Static Conditions 52
i.e., the short circuit ratio equals 30. In this case, the external grid is considered to be
strong and its SCP is calculated to be 10 MVA 30 = 300 MVA. According to [118], the
short circuit ratio value below 20 is considered as a weak grid connection. The details
of the selected PV module and inverter can be found in [119] and [120], respectively.
Based on the PV module parameters in Table 3.1, the maximum power for each PV
generator at STC is:
In this section, simulation results are initially reported from the tests of GCPPPs under
different environmental conditions. Then, the results under static conditions using all
four methods of reactive power support are reported. It should be mentioned that the
base value for the apparent, active, and reactive powers, are given per unit (p.u.) based
on a rated apparent power of 1 MVA.
The ability of the DigSILENT model to follow changes in solar irradiation and temper-
ature is shown here. The solar irradiation has a direct effect on the output power of the
PV power plant, while the temperature has an inverse effect on that [121]. An example
of active power tracking of the solar irradiation variation as well as the inverse effect of
a temperature rise on the output power is shown in Fig. 3.4. Since the active power is
within the range of 10 MW, this power is scaled-down by ten to fit in the same represen-
tation, which is indicated in the figure by base: 10 p.u.. It should be mentioned that
the increase in the temperature of the PV arrays produced by the increase in the solar
irradiation is not considered in the existing DigSILENT model of the PV power plant.
In order to evaluate the validity of the test system for supporting the grid with fixed
reactive power, the dispatch input of all 10 PV generators is set to (P, Q) mode with
the same variations in irradiance and temperature as in the previous test, and Pmpp is
set to 0.93 MW and Q to 0.2 MVAr. The results are presented in Fig. 3.5.
From the results shown in Fig. 3.5, it can be concluded that the reactive power
remains constant to 0.2 MVAr at the output of the PV generator throughout all the
Chapter 3. Performance of GCPPPs under Static Conditions 53
2.00
1.60
1.20
0.80
0.40
0 .00 ~--------~--------------------~----~----~--------~
0.0000 2.0000 4.0000 6.0000 8.0000 [sj 10.00
PV t\rrny: l'vlodule Temperamre in p.u . (base: 25.00 "C)
PV 1\rray: Solar Radiation in p.u. (base: 1000.00 Wlm' )
C'ub_ l\PQ Measurement_PCC: Active Power in p.u. (base: 10.00 ~fV:\.)
1.00
0.80
9.010s / /
O.S6l
0.60
0 .40
---------------------~~~~::;...~~
0.20
0 . 00 1----~-----'-----------'-----------...__ _________._____________..
0.0000 2.0000 4 .0000 6.0000 ll.OOOO [sl 10.00
PVI: Po\..-er Factor
Cub_ l\PQ Measurement_PCC: Active Power in p.u. (base: 10.00 .tvlVA)
PQ Measurement_LV l: Reactive Power in p.u. (base: 1.00 l\fVA l
Cub_ !\PQ Measurement_PCC: Reactive Power in p.u. (base: 10.00 MVA)
Figure 3.5: Performance of the DigSILENT PV model in the mode of injecting fixed
reactive power (Q=0.2 MVAr).
process, as expected. The deviation of Q from the setpoint value (0.2 MVAr10=
2 MVAr) at the PCC is due to the consumption of reactive power in the transformers
and ac lines. Therefore, the existing model of the DigSILENT PV power plant is capable
of providing fixed reactive power under static conditions.
However, as mentioned before, the reactive power has to be limited so that the cos
does not exceed the threshold defined in the GCs. But, as shown in Fig. 3.5, there is
Chapter 3. Performance of GCPPPs under Static Conditions 54
I q max
En I q min
Figure 3.6: Dynamic reactive power support integrated with the existing PV model
in DigSILENT.
no control on the reactive power and therefore on the power factor value in this mode.
To deal with this problem, a control system is designed to meet the required reactive
power support considering the GCs.
As depicted in Fig. 3.6, there is a voltage control slot (droop control) in the controller
block of the PV frame in DigSILENT that supports the grid in dynamic conditions in
the case of disturbances such as faults. This dynamic controller is triggered when the
rms voltage at the PV generator terminal (eL ) deviates more than 10% of the nominal
voltage (En ), i.e., |deL | > 0.1 p.u.. More information on dynamic grid support is given
in [5].
The output of the droop control block iqref is limited by the maximum and minimum
reactive current values Iqmax and Iqmin , respectively, as depicted in Fig. 3.6. Then,
the current references iqref and the calculated idref are used to regulate the active and
reactive power generated by the inverter, respectively.
These principles of the dynamic controller can be extended to a static control system
that can operate in parallel with the dynamic controller to support the grid with reactive
power under static conditions (|deL | < 0.1 p.u.). Such a static controller is based on
a PI controller and is shown in Fig. 3.7. The desired reactive power Qref (which is
set by the TSO) is compared with the actual reactive power (Q) measured at the ac
terminal of the PV generator and the mismatch value passes through the PI controller.
The output of the PI controller (Qnorm ) is then divided by eL to obtain the respective
reactive current (iqnorm ). A switch is introduced and embedded to select the suitable
reactive current reference (iqref or iqnorm ) depending on the level of voltage drop.
In order to implement the cos restrictions to the static control system, another
block is designed to calculate the acceptable maximum and minimum reactive power
values (Qmax and Qmin ) to be applied as the upper and lower limits of the PI controller,
respectively. Qmax and Qmin are defined according to the reactive power capability
curve of the inverter and also the power factor limitations as depicted in Fig. 3.8. The
Chapter 3. Performance of GCPPPs under Static Conditions 55
--~~IL..Q___
Limits. .~ : : ~ -- -- -- -:
Pin
Q max
I
I Qnorm
Qref I
PI Controller
+
Qmin
Qin Low-pass
Filter
Figure 3.7: Proposed control system to provide reactive power support in static
conditions.
P p.u.
cos 0.95 cos 0.95
(leading) (lagging)
1
shaded region is the acceptable area of reactive power operation and the reactive power
limits are determined based on the worst case scenario of cos =0.95.
The developed control system is evaluated in the test system to show its validity
to support the grid with reactive power based on the GCs. The performance of the
proposed controller is shown in Fig. 3.9. The reactive power remains fixed during the
first five seconds where cos is within limits as recommended by the GCs, and gains the
maximum acceptable value in the next five seconds to maintain the power factor within
the limits. This controller is added to the inverter controller in the constant reactive
power mode. Qref is set to 0.2-MVAr and the results are shown in Fig. 3.9. These
results show a clear improvement when compared with those in Fig. 3.5.
Chapter 3. Performance of GCPPPs under Static Conditions 56
1.20
------~-:.:::------- ---------------------;---
0.80 ......,_.__
....___ ___ 1.830 s .'/
0.971 9.2so s I
0.950
0.00
-0.~0
-0 . 80L--~----'---~--..L_--~----~--_L_--~----L
0.0000 2.0000 4.0000 6.0000 8.0000 [s] 10.00
Controller: M inimum Reactive Power Limit in p.u. (base: I .00 lvfVA)
Controller: I\faximum Reactive Power Limit in p.u. (base: 1.00 l\IIVA)
P Q Measurement- B l : Reactive Power in p .u . (ba.sc : 1.00 :tviVA)
PVl: Power Factor
Figure 3.9: Performance of the modified model in the fixed reactive power mode.
LOO
0.80
O.GO
0.40
o.zoj:::::==:::::==:=:=::::====:::::;
-;;;;;
_==- =4r--
-~=......
- -------- -
O.OOL-----------~----------~----------~----------~----------~
0.0000 2.0000 4 .0000 6.0000 s.oooo [s] 10.00
PV I : Po,vcr Factor
Cub_ J\PQ :Mcasurcmcnt_PCC: Active Power in p.lL (b ase: 10.00 :>.fVA)
PQ Mcasuremcnt_B l Reactive Power in p.u. (base: 1.00 MVA)
Cub_ 1\PQ :tvlt:asurernenl_P CC: Rem:tive Power in p.u. (base: 10 00 l'v1VA)
Figure 3.10: Performance of the existing PV model in fixed power factor mode.
In this subsection, the fixed power factor mode of reactive power control is selected. The
dispatch input is set to (P , cos ) mode with values of 0.93 MW and 0.97, respectively.
In this case, Q and P should follow the solar irradiation so that the cos remains
constant. However, since Q remains constant the cos varies, as depicted in Fig. 3.10.
Chapter 3. Performance of GCPPPs under Static Conditions 57
---------------------------~-------------
'-.,_
', 6.770 s
0 .970
----- - --------------------------
0.00
0.50 7.330 s
-0 .961 """,
- Loo~-o--
'' ......-,-....~--,
........- ........_..,.__.._., _______ , ___~-.~~
0.0000 2.0000 ~ .0000 6.0000 8.0000 [s] 10.00
Controller: Minimum Reactive P ower Limit in p.u. (base: 1.00 M VA)
Controller: Maximum Reactive Power Limit in p.u. (has~: : 1.00 M VA)
PQ Measurement_ E l : R eactive P ower in p.u. (base: 1.00 MVA)
PVI : Power Factor
F.xlemal Grid: Total Power Fador
Figure 3.11: Performance of the modified model in the mode of fixed power factor.
As discussed before, a static control method is applied based on Fig. 3.8, but with a
constant cos as a reference setpoint. Based on this, the related Qref is obtained. The
rest of the procedure is the same as in the previous one. The proposed control method
is tested under constant cos = 0.97. As shown in Fig. 3.11, a fixed power factor is
obtained achieving the same value as the reference one. The power factor delivered to
the PCC is almost the same as the reference value. The deviation from the setpoint
value at the PCC is due to the consumption of reactive power in the transformers and
ac lines. The negative sign of the power factor in the external grid is because the power
delivered to the external grid has a negative value.
In this method, the cos is controlled by the active power produced in the PV generator.
The cos P characteristic is selected from the German GCs [5] by defining two vertices
as 0.2 and 0.8. It is normalized based on the rated apparent power of the PV power
plant as shown in Fig. 3.12. There is no cos (P ) control technique implemented in
the template model of the DigSILENT PV power plant. As a result, and based on
the previous proposed solutions, a static control system is designed to address this
requirement. In this case, the Qref is derived based on the cos P relationship. The
active power in Fig. 3.13 refers to the power generated in the 1-MVA PV power plant,
which is also the same as P in Fig. 3.7, and here Pn is equal to the rated active power
of the PV inverter.
Chapter 3. Performance of GCPPPs under Static Conditions 58
cos
0.95
0.8
1
0.2 1 P P
I
n
0.95
1.20 -
-- --
,~,--- --- ----------------------
.,-
-0.40
- 0.80 '---~---'------'-------'--------''--------'-
0.0000 2 .0000 4.0000 6.0000 8.0000 rsl 10.01
PVl: Power Factor
PQ Measurement B l : Active Power in p.u. (base: 1.00 MVA)
ControUer: MininlUm Reactive Power Limit in p.u. (ba se : 1.00 MV A)
ControUer: Maxinmm Reactive Power Limit in p.u. (base: 1.00 l'v1VA)
PQ Measurement_B 1: Reactive Power in p.u. (base: 1.00 MVA)
As it can be observed, the reactive power remains within the acceptable limits and
the minimum value of power factor is 0.95.
The fourth method tested in this chapter for reactive power support under static condi-
tions is reactive power controlled by the grid voltages drop, i.e. Q(E), known as droop
control. This method attempts to alleviate the voltage variations of less than 10% re-
spect to the nominal value En . First, the existing model of the PV power plant in
DigSILENT is tested to check its capability to support the grid under static conditions
Chapter 3. Performance of GCPPPs under Static Conditions 59
Figure 3.14: Performance of the droop control in the existing PV model of DigSI-
LENT when the droop value is set to 2.
with the droop controller. For this purpose, the local voltage controller in the PV gen-
erator is adjusted to the droop mode and the droop value is set to 2 (200%). A voltage
drop less than 0.1 En is applied to the MV bus. During the voltage drop duration, this
controller should inject reactive power to mitigate the voltage drop in the PCC bus.
Fig. 3.14 shows the results after activating the droop control mode of the existing
model with a droop value equal to 200%. For better illustration of the results in the
voltage drop, solar irradiation and temperature values are set constant throughout the
simulation. It is obvious that there is no reactive power injection during the voltage
drop.
Therefore, a method which is the same as the droop control in Fig. 3.6 is proposed.
However, the static controller is triggered when the voltage drop is less than 10% of the
nominal voltage. The droop control equation is as follows:
where eL is the rms voltage at the output of the PV generator, In0 is the nominal phase
current of the inverter in dq coordinate, deL is the voltage deviation from the nominal
value En and droop is a constant value. More information on how to obtain In0 will be
given in Chapter 4. The capability of the developed model is shown in Fig. 3.15. For a
better comparison with the existing model, the droop value in the proposed method is
Chapter 3. Performance of GCPPPs under Static Conditions 60
1.60
7 .002 s
0.992
1.20
0.40
----..- ---------........-------------------
_ _ _ _ _ _ _ _ _ _ _ _ _ .J
t------
2.0000 4.0000 6.0000 ::!.0000 [s] 10.00
Voltage Mcasurcment-l'CC: Outout Volta2e. Absolute in o.u . (base: 20 .00 kV)
ConlroUt!r: l\finimUJn Rt!actiw Pow~r Limit in ]J.LI. (base: 1.00 :\1VA)
Controller: Maximum Reactr,.;e Power Limit in p.u . (oilli~: L OO MVAJ
PV l : Power Factor
Voltage Measurement-B J: Output Voltage, Absolute in p.u. (base: 4 15.00 V)
PQ Measurement Reactive Power in p.u. (b ast!: 1.00 IvfVA)
Figure 3.15: Performance of the modified static droop control when the droop value
is set to 2.
set to 2 (200%). Based on the results in Fig. 3.15, there is voltage improvement in the
PCC bus by the injection of reactive power. Nevertheless, the power factor drops a bit
due to the rise in reactive power, but it is still within the GCs limits.
3.7 Conclusion
In this chapter, the model of a GCPPP in DigSILENT PowerFactory software has been
modified and assessed to support the grid voltages under static conditions with different
reactive power control methods. The operation modes implemented and tested are fixed
reactive power, fixed power factor, power factor dependent on the active power, and
droop control. The model has been developed further to address the requirement of the
power factor conservation. The effectiveness of the modified models has been evaluated
and confirmed with a 10-MVA GCPPP. The proposed reactive power control methods
help to improve the performance of the GCPPP under static conditions.
In the following chapters, the thesis will focus on dynamic performance of the GCPPP
and how to make it FRT compatible based on the GCs requirements. In the next chapter,
these requirements will be addressed for the single- and two-stage GCPPP conversions
considering only the positive sequence control of the grid currents.
Chapter 4
4.1 Introduction
In this chapter the control of GCPPPs under voltage sag conditions is studied when only
the positive-sequence of the grid currents is controlled. The recent GCs [4, 5] require
the GCPPPs to stay connected under voltage sag conditions and support the grid by
reactive power injection. However, the inverters of GCPPPs have to be disconnected
from the grid in the cases of:
2. excessive ac currents
Furthermore, under unbalanced voltage sags, the generated active current reference
contains second-order harmonic distortion which may affect the power quality delivered
by the GCPPP.
GCPPPs with single- and two-stage conversion stages are studied in this chapter.
The two-stage conversion consists of a dc-dc converter (or several dc-dc converters)
61
Chapter 4. Performance of GCPPPs with Positive Sequence Control 62
in addition to the VSI needed to perform the grid connection. Both configurations
have been considered in this chapter to address the issues of inverter disconnection
under various grid faults. The proposed methods are tested on a single-stage conversion
GCPPP for the items 1, 2, 4. Nevertheless, they are also applicable to GCPPPs based
on a two-stage conversion. The only difference in single- and two-stage conversions is
the protection of the GCPPP from excessive dc-link voltage (item 3). While the single-
stage is self-protected from dc overvoltage, three different alternatives are proposed in
this chapter for the two-stage conversion.
The proposed VSI control methods also incorporate reactive power support in the
case of voltage sags based on the GCs requirements to ride-through the faults and
support the grid voltages. Another study on enhancing the power quality of the GCPPP
under unbalanced voltage sag conditions is also performed in this chapter. Simulation
and experimental results illustrate the capability of the proposed control methods to
ride-through different types of grid faults.
Fault studies are important in large-scale grid-connected renewable energy systems and
have been reported in the technical literature. However, most of these studies are focused
on grid-connected wind power plants [122, 123]. In the case of GCPPPs, some research
has been reported on FRT capability [22, 33, 37, 124126]. The studies in [33, 124] only
considered balanced voltage sags for addressing FRT capability. In [125], a strategy to
limit the generated current under voltage sag conditions was proposed based on updating
the operating point of the PV array. In that paper, a feedback linearization method was
used instead of a PI controller to track the dc voltage reference faster. Considering
passive methods, some remedies were presented in [126] for a single-stage conversion
using a super capacitor, and in [127] for a two-stage conversion using energy storage
to absorb the extra power generated by the PV array during the voltage sag to avoid
overcurrents.
Regarding dc voltage protection during voltage sags, a method was proposed in [37]
for a two-stage conversion which defined a threshold for the dc voltage and set the duty
cycle based on that to control the generated power from the PV array. A study was
reported in [128] for the control of the dc side of the inverter, which shows the impact
of various types of faults on the voltage and current of the PV array. Among several
studies for unbalanced voltage sags, a method was introduced in [32] to mitigate the
peak output currents of a 4.5-kVA PV power plant in non-faulty phases. Another study
in [29] presented a PR current controller for the current limiter to ensure a sinusoidal
Chapter 4. Performance of GCPPPs with Positive Sequence Control 63
output current waveform and avoid overcurrent. However, in the mentioned studies,
reactive power support was not considered. In the case of a two-stage conversion, no
technical paper so far has proposed a comprehensive strategy to protect the inverter
during voltage sags while providing reactive power support to the grid. All the designs
and modifications for the inverter in both the single- and two-stage conversions, have
to accommodate various types of faults and address FRT capability based on the GCs
[4, 5].
Another objective of this chapter is to improve the power quality of GCPPPs under
unbalanced voltage sags by including a filter in the dc-link voltage control loop. The
filtering technique applied is based on MAFs. However, including a filter in the dc-link
voltage loop introduces delays that may impact on the stability of that control loop and
Chapter 4. Performance of GCPPPs with Positive Sequence Control 64
eventually on the whole GCPPP. A stability analysis is performed when the MAF is
included in the dc-link control loop, which is used to determine proper parameters for
the dc-link controller.
When connecting PV plants into the MV grids, these plants need to contribute support-
ing the dynamic of the grid. Germany has the most comprehensive GCs in utilization
and application of PV power plants and therefore these GCs are taken into consideration
in this study. Based on the German GCs, [4, 5] the GCPPPs connected to MV grids
have to be provided with FRT capability and support the grid. This means that they
have to able to:
Support the grid voltages by the injection of reactive power during the voltage sag,
Consume the same or less reactive power after voltage sag clearance.
The requirements for FRT capability depend on the severity and duration of the
voltage sags. The limiting curves of the German GCs are depicted in Fig. 4.1. According
to this figure, the GCPPPs must not disconnected during a voltage drop of 100% of En if
the duration is less than 150 ms. The voltage sags with voltages above the Borderline 1
cannot lead to disconnection, and the voltage sags beneath the Borderline 1 and above
the Borderline 2 have to be ridden-through.
Regarding the amount of reactive power support based on the German GCs, the
reactive power injection must be activated when there is a voltage drop of more than
10% of rms value, as shown in Fig. 4.2. The voltage controller must assure the supply
of reactive current with a support of at least 2% of the rated current per percent of the
voltage drop:
where droop is a constant value and deL is the amount of voltage drop and In0 is the
nominal current of the inverter in dq coordinates. Considering the transformation
q from
2
abc to dq coordinates to be power conservative, i.e., using the coefficient 3 in the
Chapter 4. Performance of GCPPPs with Positive Sequence Control 65
e
En
Figure 4.1: Limiting curves of voltage at the grid connection point for PV power
plant in the event of a voltage sag in the grid.
transformation (2.20), the nominal apparent power of the inverter (Sn ) can be written
as:
Sn = 3En In = Ed Id + Eq Iq (4.2)
where En and In are the rms nominal line-to-line voltage and nominal phase current
of the inverter, respectively, Ed/q and Id/q are the d/q components of the voltages and
currents in dq coordinates, respectively. Since the SRF-PLL tries to impose the q com-
ponent of the voltage (Eq ) to be zero, and Ed in dq coordinate is equal to En in abc
coordinates, the nominal value of the the current in dq coordinate can be derived based
on (4.2) as:
In0 = Id = 3In . (4.3)
In the case of asymmetrical faults, the reactive currents must not exceed values that
cause the grid voltages of the non-faulty phases to rise above 110% of the nominal value
[4, 5].
In this section, the four main reasons for inverter disconnection will be analyzed, i.e.,
iq
In
Dead band limits:
Emax 1.1En ,
Emin 0.9 En
e
En
e e e0 , iq iq iq 0 ,
e intantaneous voltage,
e0 voltage prior to voltage sag,
iq reactive current,
iq 0 reactive current prior to voltage sag,
Figure 4.2: Principle of voltage support in the event of grid voltage sags.
excessive ac currents,
The SRF-PLL introduced in Section 2.5.4.4 performs well under balanced three-phase
systems, i.e., operating with nominal grid voltages or with balanced voltage sags. How-
ever, in the case of unbalanced voltage sags, the PLL is unable to follow the positive
sequence of the grid voltages accurately and a second-order harmonic is propagated
through the PLL that eventually appears in the extracted phase angle. Tuning the PI
controller parameters of the PLL is a trade-off between achieving a clean detected angle
in the steady state conditions and having fast dynamic response. Since voltage sags are
dynamic processes, a fast response is essential to address the FRT capability.
As can be seen in Fig. 4.3, the SRF-PLL performance is satisfactory if the voltages
are balanced but not under unbalanced conditions. The SRF-PLL is not capable of
removing the second-order harmonic produced in the dq-transformed grid voltages under
Chapter 4. Performance of GCPPPs with Positive Sequence Control 67
eabc
Time s
Figure 4.3: Performance of the SRF-PLL under LG with 60% voltage sag condition:
(a) input voltages, and (b) extracted angle.
ed ed ed
abc MAF dq
e abc n
eabc abc
eq eq eq
+ o
dq MAF abc dq PI
+
2 f f
such conditions. Therefore, a more elaborated method should be used able to extract
the angle of the positive sequence with no significant distortion. Several methods have
been proposed to deal with this problem [6, 4246, 133, 134]. Among those methods, the
filtered-sequence PLL (FSPLL) [6] is very effective and simple to apply. Fig. 4.4 shows
the diagram of the FSPLL method, which is based on using an arbitrary angle (f )
that evolves with the grid frequency, together with MAFs to remove the second-order
harmonics from the distorted dq components of the input voltages. Then, the extracted
positive-sequence is the input of a conventional SRF-PLL that obtains the angle of the
positive sequence. The SRF-PLL is tuned to perform fast since the fundamental negative
sequence and other harmonic components have been removed by the MAFs.
in which Tw is the window width of the MAF and x(t) is the moving average value of
x( ) calculated over Tw . The window-width of the MAFs (Tw ) needs to be calculated
properly to remove the second order oscillations. In the case of asymmetrical faults, the
Chapter 4. Performance of GCPPPs with Positive Sequence Control 68
ed = Ed + A sin(2t + ). (4.5)
A
ed = Ed + { cos(2t + ) + cos[2(t Tw ) + ]}. (4.6)
2Tw
Focusing on the ac components (Ed = 0) and assuming =0, the Laplace transform
applied to (4.6) gives:
Replacing s with j
In order to remove completely the sinusoidal component from the output of the MAF,
the magnitude of the transfer function in (4.8) is set to zero, i.e.:
1 p
|GM AF (j)| = 8 6 cos(4Tw ) 2 cos(2Tw ) = 0, (4.9)
4Tw
which gives:
k
Tw = k or Tw = for k = 1, 2, 3, ..., (4.10)
2
where T is the period of the grid voltages. From (4.10), the minimum window-width to
be selected is Tw = T /2, i.e., in the case of a 50-Hz system the minimum Tw is 10ms.
The larger Tw is, the larger the number of harmonics that are filtered, but the slower
the performance of the MAF. In Fig. 4.4, the d and q components of the input voltages
are filtered by MAFs with the same window-width. It should be mentioned that in the
case of frequency variation, a frequency detector is needed to extract the exact window
width for the MAF to perform properly. This has been addressed in Chapter 6.
Now the FSPLL is tested under the same voltage sag condition applied to the SRF-
PLL shown in Fig. 4.3 using the same parameters for the PI controller. The results are
depicted in Fig. 4.5. As it can be observed, the FSPLL detects the positive-sequence
angle of the grid voltages under unbalanced voltage sags without any low-frequency
distortion.
Chapter 4. Performance of GCPPPs with Positive Sequence Control 69
eabc V
_:l _
400~--~----~----~--~----~----~--~----~----~--~
400L.~~~~~~~~~~~~~~~~~~~~~~~~~~~
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
rad (a)
Time s
Figure 4.5: Performance of the FSPLL under a 60% LG voltage sag condition: (a)
input voltages and (b) extracted positive-sequence angle.
To protect the inverter from excessive ac currents, first some methods for obtaining
the reference currents are defined and then a protection strategy is applied. In a GCPPP,
the active current reference is generated by the dc-link voltage control as shown in Fig.
2.17. On the other hand, the reactive current reference is provided based on the droop
control presented in Fig. 3.6 and defined in (4.1). There are two ways of measuring the
grid voltages eL in Fig. 3.6, which will be described later in this section.
As mentioned in [5], the reactive power injection must be activated when the voltage
drop is more than 10% of the rms value. Therefore, the rms value of the three line-
to-line voltages needs to be considered. This value is compared with the nominal rms
voltage En and the difference goes into the droop control to determine the amount of
the reactive current required for the grid voltage support.
Consider the GCPPP in Fig. 2.1 with the case study specifications given in Table 4.1.
According to the characteristics of the PV modules and the number of units connected
in series and parallel, the maximum power injected under STC is 1 MW. This power
gives the rated current of 1400 A (1979.8 A peak value) at the low voltage side of the
Chapter 4. Performance of GCPPPs with Positive Sequence Control 70
transformer. According to the datasheet of the inverter [120], the maximum acceptable
output current at the low voltage side of the transformer is 1532 A (2166 A peak value).
However, for a 40% three-line-to-ground (3LG) voltage sag at the MV side, the peak
value of the output current exceeds the limits in the example shown in Fig. 4.6, which
would lead to inverter disconnection.
The performance of the GCPPP after applying the current limiter for the same case
study is illustrated in Fig. 4.8. If the generated current in the dc-side of the GCPPP is
more than the injected current into the electrical grid, the extra current goes through the
dc-link capacitor and the dc voltage starts to increase. Consequently, in a single-stage
PV power plant, as the dc-link voltage rises the operating point on the I V curve (in
Fig. 4.9) moves towards the open-circuit voltage point, which leads the PV current to
Chapter 4. Performance of GCPPPs with Positive Sequence Control 71
~400-
eabc
_::~ ~ ~: :~: ~~ ;;. :_- .. : : .:
- 4 0 0 ' - - - - - ' - - - - - - - ' -_
-_--_.__
-_--
;.
_-___.__
'- _-_
:. ~ .. .. _<..>:~ ,_
-__.__
. ._ ._
____,
_:<~
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
(a)
idref [kA]
:~L___
t ____L__
l _ _ _ _l_
. . .l....._
iqref [kA]
..........
l-----l-....____
l ______j__
l ___jj
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
(c)
-i~
iabc [kA]
4 . .
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
(d)
Time [s]
Figure 4.6: 40% balanced voltage sag at the MV side of the transformer: (a) grid
voltages at the LV side, (b) active current reference, (c) reactive current reference, and
(d) grid currents at the LV side.
Vdc* idref
PI
Current
idref
Limiter
vdc
En deL iqref
+
eac
Figure 4.7: Application of the current limiter to the reference currents.
Chapter 4. Performance of GCPPPs with Positive Sequence Control 72
400
~200 ~
. '. '.' .:
- 0 . ' . ' . ' . '
eabc
-200
400~--~----~--~----~--~----~--~~--~----~--~
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
(a)
vdc
0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
_: L
t ..._____...L..
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
(c)
iabc
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
(d)
Time [s]
Figure 4.8: Results after applying a current limiter with a 40% balanced voltage sag
at the MV side: (a) grid voltages at the LV side, (b) dc-link voltage, (c) active current
reference, and (d) grid currents at the LV side.
idc
Impp
inew
decrease. Therefore, the power generated by the PV power plant is reduced because the
operating point is taken away from the MPP and hence less active current is required at
the ac-side. Finally, the operating point reaches new values in which the input current
is lower and the dc-link voltage higher than the values prior to the voltage sag event as
shown in Fig. 4.9. By the implementation of the current limiter and injection of the
reactive current during the voltage sag, the active current is reduced as demonstrated
in Fig. 4.8(c). The reactive current reference, which is not shown here, has the same
value as the one in Fig. 4.6.
Chapter 4. Performance of GCPPPs with Positive Sequence Control 73
Vdc*
idref
-1 kpv
vdc I r----- I
1 : I
I : kiv I
I : I
I : I
I : I
I : :_no
I ._______ -; s
------
I
I
- - - .J
A problem that may arise because of the deviation of the MPP during the voltage sag
is that after fault removal the dc-link voltage and the ac currents may take a long time to
reach the pre-fault values, as shown in Fig. 4.8. The reason for this is the accumulation
of error in the integral part of the PI dc-link voltage controller. The existing dc-link
voltage error keeps the PI output increasing prior the current limiter, and therefore
excessive control action may be accumulated during the voltage sag process.
From another perspective, the measured rms value of the voltage contains even-
order harmonics under unbalanced grid voltage conditions which makes the voltage drop
distorted. An example is shown in Fig. 4.12 in the case of a 65% LG fault in phase
a. The variable deL is depicted in Fig. 4.12(b). Based on the amount of voltage drop,
the droop control is activated and deactivated continuously. This distorts the iqref and
consequently i0dref after passing through the current limiter. Distorted reference currents
produce the distorted ac currents shown in Fig. 4.12(e), which may lead to overcurrent
in some cases.
eabc
vdc
idref
iabc
Figure 4.11: Results after applying an anti-windup technique to the voltage loop with
a 40% balanced voltage sag at the MV side: (a) grid voltages at the LV side, (b) dc-link
voltage, (c) active current reference, and (d) grid currents at the LV side.
with
er = E cos(t + ), and (4.12)
ei = E sin(t + ), (4.13)
er and ei being the real and imaginary components of the voltages, respectively, and E
the peak value of the phase voltage. However, in the case of unbalanced voltage sags,
er and ei may change in both magnitude and phase-angle. Therefore, (4.12) and (4.13)
can be re-written under unbalanced voltages as:
er = E cos(t + + r ), (4.14)
ei = E sin(t + + i ), (4.15)
0 , 1,
Chapter 4. Performance of GCPPPs with Positive Sequence Control 75
eabc [V]
deL [V]
n
iqref [kA]
[kA]
idref
iabc [kA]
Figure 4.12: Results for a 65% LG voltage sag in phase a at the MV side: (a) grid
voltages at the LV side, (b) voltage drop (c) reactive current reference, (d) active current
reference, and (e) grid currents at the LV side.
in which r and i are due to a phase-angle jump in the vector phase. Therefore, the
rms value of the unbalanced voltage can be indicated by:
r
3p
eL = A1 + A2 , (4.16)
2
with
1 + cos(2t + 2 + 2r )
A1 = e2r = 2 E 2 [ ], and (4.17)
2
1 cos(2t + 2 + 2i )
A2 = e2i = 2 E 2 [ ]. (4.18)
2
In the worst case, and are different from the unity and r and i are non-zero
values. It is worth mentioning that the phase-angle jump occur in types C, D, E, and
F of faults [136]. After simplifying (4.16) one obtains:
r r
3 A4
eL = A3 1+ sin(2t + 1 ), (4.19)
2 A3
Chapter 4. Performance of GCPPPs with Positive Sequence Control 76
where
E 2 2
A3 = ( + 2 ), and (4.20)
2
r
2 1 2 2 cos(2r 2i )
A4 = E . (4.21)
2
The rms value of the voltage under unbalanced conditions will contain a dc value
plus even order harmonics. Among all these harmonics the most significant is the second
one and the others are negligible. However, by selecting a MAF with a window-width
equivalent to a double frequency of the grid voltages, i.e. Tw = 1/(2f ), all the even
harmonics will be removed. The improved results after applying a MAF to deL are
depicted in Fig. 4.13.
The drawback of measuring the rms voltage of the set of three-phase voltages together
is that it does not show the severity of the voltage drop. For example a 30% LG voltage
sag will not activate the droop control, as shown in Fig. 4.14. Therefore, another
method of measuring the voltage drop based on individual measurements of the phases
is introduced in the GCs [4] and explained in the next section.
Based on the GCs [4], the amount of deL used for the droop control in (4.1) is obtained
based on the lowest rms value of the line-to-line voltages of the three phases at the
terminal of the GCPPP, i.e., eLmin in Fig. 4.15. The rms voltage is obtained using the
following expression:
s Z t
1 T
eyrms = ey 2 dt with Tw = , (4.24)
Tw tTw 2
Chapter 4. Performance of GCPPPs with Positive Sequence Control 77
eabc
deac [V]
iqref [kA]
[kA]
idref
iabc [kA]
Figure 4.13: Results for a 65% LG voltage sag in phase a at the MV side after
applying a MAF to the voltage measurement: (a) grid voltages at the LV side, (b)
voltage drop, (c) reactive current reference, (d) active current reference, and (e) grid
currents at the LV side.
where ey is the instantaneous line-to-line voltage for y = {ab, bc, ca}, Tw is the window
width for the rms value calculation, and T is the grid voltage period, which is equal to
20 ms for a grid frequency of 50 Hz.
The control diagram for the reactive current generation is depicted in Fig. 4.15.
The results after applying the control diagram in Fig. 4.15 are shown in Fig. 4.16.
The results can be compared with the ones in Fig. 4.13 in which the rms value of the
three-phase voltages was considered for the voltage measurement. This method does
not require any filtering technique.
Protecting the GCPPPs from dc-link overvoltage requires different control actions de-
pending on if the GCPPP is based on single- or two-stage conversions. Therefore, this
issue is addressed separately for each case.
Chapter 4. Performance of GCPPPs with Positive Sequence Control 78
s:-'400
-~~ -~
......
eabc .. .. H. : 00 ~ U: UU .:. U .:... -- .. ..~.- .. .... , .....:. U HO' H H . : , U H O UO
-400 .
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
(a)
! '-------
~ -----'---!-------'
.. .. . ~_-- t _ ._ _ r ------'----:-------'~.
deac [V]
.. ------'----
! .______.__.. --'-----
.. r _
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
(b)
iqref [kA]
- i 1t - - -- - - - j - - - - - - - - j - - - - - - - - - j - - - - - - - - '- j - -- - - - - - j - - - - - - - j - - - - - - - - j - - - - - f
-~-
iabc [kA]
Figure 4.14: Results for a 30% LG voltage sag in phase a at the MV side after applying
the MAF for voltage measurement: (a) grid voltages at the LV side, (b) voltage drop
(c) reactive current reference, (d) active current reference, and (e) grid currents at the
LV side.
En deL iqref
eL min
eab
ebc
eca
Figure 4.15: Droop control diagram for the generation of the reactive current reference
based on the minimum rms value of the line-to-line voltages.
Chapter 4. Performance of GCPPPs with Positive Sequence Control 79
eabc [V]
deL [V]
iqref [kA]
[kA]
idref
iabc [kA]
Figure 4.16: Results for 65% LG voltage sag in phase a at the MV side with the new
voltage measurement: (a) grid voltages at LV side, (b) voltage drop (c) reactive current
reference, (d) active current reference, and (e) grid currents at the LV side.
If the active current reference is limited, i.e. idref < i0dref , the generated power from the
PVs is more than the injected power into the electrical grid. As a consequence, some
energy is accumulated into the dc-link capacitor, increasing the dc bus voltage. In a
single-stage GCPPP, as the dc-link voltage increases, the operating point on the I V
curve of the PV array moves towards the open-circuit voltage point, which leads the
PV current to decrease, as shown in Fig. 4.9. Therefore, the power generated by the
PV panels is reduced because the operating point is taken away from the MPP and less
active current is injected into the ac-side. This happens until the GCPPP reaches a new
steady state where the dc-link voltage stops increasing. Thus, single-stage GCPPPs are
self-protected because the generated power is reduced when the dc-link voltage increases
under ac faults.
It should be mentioned that the inverter has to withstand the worst case of the dc-link
voltage, which is produced when the voltage provided by the PV modules reaches the
open-circuit value (Voc ) under the maximum solar irradiation expected on the generation
Chapter 4. Performance of GCPPPs with Positive Sequence Control 80
site. Hence, the number of PV modules connected in series (ns ) has to be limited in the
design of the GCPPPs so that the dc-link voltage is never higher than the maximum
acceptable value of the inverter (Vdcmax ) as depicted in Fig. 4.9 based on the following:
Vdcmax
ns . (4.25)
Voc
A two-stage GCPPP includes a dc-dc converter between the PV array and the inverter.
In high-power GCPPPs, more than one dc-dc converter can be included, one per each PV
array. Despite having several dc-dc converters, these systems will be referred anyway
as two-stage GCPPPs. In two-stage GCPPPs, the MPPT is performed by the dc-dc
converter and the dc-link voltage is regulated by the inverter.
During a voltage sag, if no action is taken in the control of the dc-dc converter, the
power from the PV modules is not reduced and therefore the dc-link voltage keeps rising
and may exceed the maximum limit. Hence, the system is not self-protected during grid
fault conditions. A specific control action has to be taken to reduce the power generated
by the PV modules and provide the two-stage GCPPP with FRT capability.
The diagram of the case study for a two-stage GCPPP is shown in Fig. 4.17. It
consists of a 1-MVA inverter and ten parallel 100-kW dc-dc boost converters. Details
of the individual dc-dc converter as well as the PV array characteristics connected to
each dc-dc converter are summarized in Table 4.2. The rest of data for this system are
provided in Table 4.1.
In two-stage GCPPPs, the PV voltage (vpv ) is controlled by the duty cycle (d) of
the dc-dc converter (Fig. 4.18 ). The reference for the PV voltage is obtained from the
MPPT technique implemented.
Transformer Grid
+ Voltage
vdc Source N
~-.,..----+-----1 Converter
PV Array
No.n
Table 4.2: PV Arrays and Dc-Dc Converter Specifications in the Two-Stage GCPPP.
MPPT
where ipv and vpv are the PV current and voltage, respectively, and ed and eq are the
d and q grid voltage components extracted by the PLL. Since the PLL forces the eq
component to be zero, the estimated d current component is obtained as:
ipv vpv
idest = . (4.27)
ed
The estimated active current component idest is then added to idref to improve the
dynamic of the dc-link control loop.
In two-stage GCPPPs, three different ways to limit the dc-link voltage under fault
conditions are proposed:
Chapter 4. Performance of GCPPPs with Positive Sequence Control 82
Li
+
r- v pv vdc
i pv_.I
__
short circuiting the PV array by turning on the switch of the dc-dc converter
throughout the voltage sag duration,
leaving the PV array open by turning off the switch of the dc-dc converter, and
changing the control of the dc-dc converter to inject less power from the PV arrays
when compared with the pre-fault operating conditions.
Fig. 4.20 shows some results for a LG voltage sag with a 60% voltage drop at
MV side of the transformer occurred from t=0.1 s to t=0.3 s. The generated power of
the PV arrays and also the injected active and reactive power into the grid are shown
in Fig. 4.21. During the voltage sag, the dc-link voltage remains relatively constant,
idref becomes almost zero with some ripples, and only iqref is injected during the fault
period. Consequently, the current limiter does not have to be activated in this case.
Chapter 4. Performance of GCPPPs with Positive Sequence Control 83
400
. .' ........ . . . . .' . .... i . .' . . .. . .' .. .. .. ' . ...... '. . ' . . . . .... '. .
eabc V -
200 -~
~
: ; : :
0 : - - - -- - - - - -- - - ~ - - - - - - - ~ -
-200
-400
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
(a)
2000 ... - - -- ---- -- ...- ..-- ---- :-- ... - ...- .. -- ---- -- ... - ... - -- ...- ..-- ---- --- ... - .. - :- ..- - -- -- ... - ..-- ---- ---- -- -:-- ... - - -- ..
-
iabc A
: - - - -- - - - . -- -~- ---- -.--- --.- --- -;
~
0 . . . . . . . .. . . . .
-2000
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
(b)
900
11 1~ 11
vdc V
800
0.6
700
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
(c)
Time [s]
Figure 4.20: Short-circuiting the PV panels when applying a 60% LG voltage sag at
the MV side of the transformer: (a) grid voltages at the LV side, (b) grid currents, and
(c) dc-link voltage.
~ I lr I !r I ~
Ppv MW
1
0
-1
0
.......................................................... ,..
0.05 0.1
1
0.15
~
0.2
!i
0.25
I i
0.3
i
......................... .,...........................
~
0.35 0.4
(a)
2
Pout MW
1
0
-1
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
(b)
1
Qout MVAr
0
-1 l ! ~ i l
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
(c)
Time [s]
Figure 4.21: Short circuiting the PV panels: (a) overall generated power, (b) injected
active power, and (c) injected reactive power into the grid.
Under unbalanced voltage sags, the output power contains a second-order harmonic
[137], which will produce dc-link voltage ripples at the same frequency.
II. Opening the Circuit of the PV Panels. Another option to avoid transferring
power from the PV modules to the dc-link is to keep the dc-dc converter switch off
Chapter 4. Performance of GCPPPs with Positive Sequence Control 84
Li
+
r- v pv C vdc
i pv _.
__ I
I
(a)
Li
+
r- v pv C vdc
i pv_.II
__
(b)
Figure 4.22: Current paths in the dc-dc converter when turning on the switch: (a)
transition mode and (b) locked in state.
throughout the voltage sag (d=0), as shown in Fig. 4.22. Since the inverter is not trans-
ferring active power into the grid during the voltage sag, the PV voltage vpv increases
until the dc-dc converter inductor is completely discharged (ipv = 0). Then the diode
turns off and the PV modules stop providing energy into the dc-link (Fig. 4.22(b)). This
case is similar to the previous one where the diode was continuously on and no current
from the PV was provided to the dc-link. The main difference with the previous case is
the transition process, as depicted in Fig. 4.23.
III. Injecting Less Power from the PV Panels. In the two previous cases,
during the voltage sags, there is no power generated by the PV modules and therefore
only reactive current is injected into the grid. However, as mentioned in [4], the network
operator is allowed to feed the grid through the generating power plant during the
voltage sags. For this purpose, the GCPPP is controlled to inject less power into the
grid during the voltage sag compared with the pre-fault case, while avoiding overvoltage
in the dc-link.
In normal operation, the MPPT function is performed by the dc-dc converter while
the dc-link voltage is regulated by the inverter. However, under a voltage sag, some
modifications should be implemented in order to keep the GCPPP stay connected. The
proposed method tries to match the power generated by the PV modules with the
power injected into the grid while trying to keep the dc-link voltage constant. Unlike
the previous cases of keeping the switch on or off during the voltage sag, in this case
Chapter 4. Performance of GCPPPs with Positive Sequence Control 85
400
eabc V
200
iabc A
0
-200
-400
2000
-
0 0.05 0.1 0.15 0.2
(a)
0.25 0.3 0.35 0.4
900
0
-2000
-
0 0.05 0.1 0.15 0.2
(b)
0.25 0.3 0.35 0.4
I!!~
:~ 11
vdc V
800
700
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
(c)
Time [s]
Figure 4.23: Turning the dc-dc converter switch on when applying a 60% LG voltage
sag at the MV side: (a) grid voltages at the LV side, (b) grid currents, and (c) dc-link
voltage.
ppv
A
Pmpp
C B
pnew
power balance is achieved for a value different from zero. Therefore, both active and
reactive currents will be injected into the grid.
In the proposed method, the target of the dc-dc converter is no longer achieving
MPP operation but regulating the power generated by the PVs to match the maximum
active power that can be injected into the grid. The dc-dc converter is controlled to find
a proper value for the PV voltage (vpv ) that achieves such power balance. As a result,
the operating point should move from point A in Fig. 4.24 to a lower power point, for
example either the points B or C.
Here, moving the operating point in the direction from A to B is applied and ana-
lyzed. For this purpose a positive voltage value vpv should be added to the Vmpp value
Chapter 4. Performance of GCPPPs with Positive Sequence Control 86
Vdc* 2 1
v pv
C
2
vdc2
v pv
Figure 4.25: Adding a controller to the dc-dc converter to force the operating point
to move from the MPP to a lower power point.
To ensure a fast dynamic response and maintaining the stability of the GCPPP, a
feed-forward control strategy is proposed and applied to the dc-link control loop. For
this purpose, a linear estimation is made based on the P V curve shown in Fig. 4.24.
Let us suppose the triangle represented by the vertices (Pmpp , Vmpp ), (0, Vmpp ) and (0,
Voc ), as depicted in Fig. 4.26. The new point (pnew , vnew ) can be estimated by (pnewest ,
vnewest ) on the triangle hypotenuse. According to the Side-Splitter Theorem and using
interpolation, the estimation of vnewest is:
pnewest
vnewest = (Vmpp Voc ) + Voc , (4.29)
Pmpp
in which Pmpp and Vmpp represent the pre-fault values at the MPP. The pnewest can be
calculated from the power injected into the grid:
P mpp ,Vmpp
pnewest ,vnewest
pnew ,vnew
0 ,V
mpp 0 ,Voc
v pv est
v pv
ed idref
vnewest = (Vmpp Voc ) + Voc , (4.31)
Pmpp
and
vpvest = vnewest Vmpp . (4.32)
The value in (4.32) is added to the controller as a feed-forward term before the limiter
in Fig. 4.25, as shown in Fig. 4.27. In order to increase the dynamics of the proposed
controller further, another estimation can be derived using (4.31) which is the estimation
of the duty cycle as a feed-forward term, dest . Based on the relationship between the
input and the output voltage of the boost dc-dc converter under continuous conduction
operating conditions
vdc 1
= (4.33)
vpv 1d
the estimated duty cycle is
vnewest
dest = 1 . (4.34)
Vdc
Chapter 4. Performance of GCPPPs with Positive Sequence Control 88
p
v pv est
Vdc* 2 1
v pv
C
2
vdc2
v pv
d est
Figure 4.27: Updated controller with feed-forward terms to enhance the dynamics of
the proposed controller.
The updated version of the controller in Fig. 4.25 is illustrated in Fig. 4.27, which
contains the two feed-forward terms to enhance the dynamics of the proposed controller.
The PI controllers PI-1 and PI-2 compensate for the difference between the estimated
and the real values of d and vpv , respectively.
The only unknown variable in (4.31) is idref . The reason is that in the proposed
method, during the voltage sag, the dc-link control loop stops adjusting the active
current reference and instead regulates the input voltage of the dc-dc converter (vpv ).
The method proposed to estimate idref is the following. Considering Fig. 4.7, the
maximum value for the idref is i0dref . Depending on the voltage sag depth and solar
irradiation (G), the value of idref can be lower or equal to i0dref . Therefore, if the active
current reference idref can be estimated, i0dref can be obtained as well. If Ppv is the
power generated by the PV array:
Ppv
idref est = . (4.36)
ed
The maximum acceptable value for the idref can be obtained based on the pre-fault
value of Ppv i.e. Pmpp and ed , as follows:
Pmpp
idref est = . (4.37)
ed
Chapter 4. Performance of GCPPPs with Positive Sequence Control 89
400
eabc V
200
0
-200
-400
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
(a)
2000
iabc A
0
-2000
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
(b)
900
vdc V
800
700
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
(c)
500
v pv V
400
300
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
(d)
0.6
0.5
d est
0.4
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
(e)
0.6
d
0.4
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
(f)
Time [s]
Figure 4.28: Control of the dc-dc converter to produce less power under a LG with
45% voltage sag at the MV side: (a) grid voltages, (b) grid currents, (c) dc-link voltage,
(d) input voltage of the dc-dc converter, (e) estimated duty cycle, and (f) actual duty
cycle.
The estimated current idref est goes through the current limiter and based on the
required reactive current reference, i0dref can be obtained and substituted in (4.31). The
performance of the proposed controller under a 3LG with 45% voltage sag at the MV side
of the transformer is shown in Fig. 4.28. As the PI controller (PI-1) is tuned to be slow
in order to track the MPP during normal operation, the parameters of this controller
(PI-1) can be increased during the voltage sag in order to improve the performance of
the proposed method.
Selected results on the performance of the system under different voltage sags and
different solar irradiation conditions are shown in Fig. 4.29. As demonstrated, the out-
put currents always remain balanced during various types of faults and solar irradiations
Chapter 4. Performance of GCPPPs with Positive Sequence Control 90
and the dynamic performance of the proposed method to reach the new operating point
is considerably fast. It should be mentioned that the ripples in the dc-link voltage in
Fig. 4.29(f),(h) are due to the unbalanced voltage sag.
The preference of the third method, i.e. injecting less power from the PV panels,
compared to the first two methods is firstly due to its capability to inject active power
into the grid during the voltage sag to support the grid. Secondly, it has the capacity
to inject balanced currents into the grid even under unbalanced voltage conditions. The
reason is the fact that the estimated active current reference given to the current limiter
is obtained based on (4.37) which provides an almost constant value.
Under unbalanced fault conditions, and considering any type of reactive current injection
such as constant reactive current or using the droop control, there is a possibility of grid
voltage exceeding in the non-faulty phases if the injected reactive currents are balanced.
For a 415 V rms voltage the maximum acceptable voltage peak value (i.e. 110% of
nominal value) is 372.73 V (1.1Enph ). In Fig. 4.30, the grid voltages are magnified
to show the impact of reactive current injection in the grid voltages during a two-line-
to-ground (2LG) voltage sag. In this example, the reactive current injected during the
voltage sag is the rated current of the inverter. As it can be observed, the voltage in the
non-faulty phase is excessively increased due to the reactive current injection.
To overcome this problem, the maximum value of the injected reactive current has
to be limited. Until now, the lower boundary of the reactive current was the nominal
current of the PV generator. However, to avoid grid overvoltage, the lower boundary
should be reduced to limit the amount of injected reactive current. For this purpose, a
PI controller is employed to regulate the maximum value of the phase voltage. As shown
in Fig. 4.31, the amount of reactive current injection is limited, and the peak value of
the voltage is reduced. The whole process is shown in Fig. 4.32 in which Emax is the
maximum amplitude of the phase voltages. However, as it can be seen from Fig. 4.31,
the reaction of reactive current reduction is faster than its effect on the voltage value.
If the PI controller is selected to be slower it will take a longer time to achieve a voltage
reduction, while if it is selected to be faster the reactive current tends to zero earlier.
The reason is the slow nature of the voltage profile that can be affected by the reactive
current change, which is not really desirable under fault conditions.
Therefore, this is not actually a comprehensive method as the time period to ride-
through the fault is very short (in German GCs the maximum time to ride-through
for 100% voltage drop is only 150ms). Therefore, during unbalanced voltage sags, a
Chapter 4. Performance of GCPPPs with Positive Sequence Control 91
400
eabc V ~
. . .
200 . '' ...........................
.
'' '' ........ '' .. '' ...................
.
'' ..............................
.
0
-200
-400
0
~ 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
(a)
2000
iabc A
0
-2000
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
(b)
900
I I EF I I r 1
vdc V
800
700
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
(c)
400
eabc V
-
200 . '' '' .. '' ..... '' .. '' .. ;,..
: , : ,
. .. ; .. ..
' '[ . . .................. .
'' '' '' ''''
-200
-400
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
(d)
2000
iabc A
900
0
-2000
0
- 0.05 0.1 0.15 0.2
(e)
0.25 0.3 0.35 0.4
I lfVVAAT I
vdc V
800
700 I I
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
(f)
2000 ~
, ,
iabc A
i ~ ; ; ' i
~;;;~ '.::
0
.......................
-2000
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
(g)
900
I I !: I
vdc V
800
700
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
(h)
Time [s]
Figure 4.29: Control of the dc-dc converter to produce less power under voltage sag:
(a) grid voltages under a 3LG with 45% voltage sag at the MV side, (b) related grid
currents for G=300 W/m2 , and (c) related dc-link voltage; (d) grid voltages under a LG
with 65% voltage sag at the MV side, (e) related grid currents for G=1000 W/m2 , (f)
related dc-link voltage, (g) related grid currents under G=300 W/m2 , and (h) related
dc-link voltage.
Chapter 4. Performance of GCPPPs with Positive Sequence Control 92
380 lf~=~l . H
~ abc 360
Time [s]
Figure 4.30: Grid voltages exceeding due to symmetrical reactive current injection
(magnified peak voltages).
~ :i:lf I !Eft ! ! ! + =
iqref
-3000~-----~----~~----~----~----~----~------~--~
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
(a)
Figure 4.31: Application of a controller to limit the rise of the voltage in phases: (a)
reactive current reference and (b) grid voltages (magnified peak voltages).
idref
Current
idref
Limiter
I q max
deL iqref
I q min
1.1En In
emax
Figure 4.32: Control diagram limiting the reactive current lower boundary.
Chapter 4. Performance of GCPPPs with Positive Sequence Control 93
more effective remedy to be compatible with the FRT time frame based on supplying
unbalanced reactive currents is recommended. Such development is reported in Chapter
6.
Under unbalanced voltage sags, which are more likely to happen than symmetrical faults,
the generated active power and consequently the dc-link voltage and active current
references contain second-order harmonic oscillations. Two situations may happen to
the active current reference (idref ) in Fig. 4.7:
In the second case (ii) where the idref does not reach the limiting values and hence
i0dref =idref , two situations may happen. The first situation is when the voltage drop is
less than 10% which does not activate the droop control. Although the active current
reference contains some distortion in this situation, it is generally negligible. The second
situation is when the unbalanced voltage sag occurs during low solar irradiation, i.e. low
generated power. As a result, the active current reference is small and not limited or just
partially limited by the current limiter. Since the output power contains a second-order
harmonic [137], significant dc-link voltage ripples may appear, which produce oscillations
in the active current component as well. Fig. 4.33 shows an example where a 2LG
voltage sag with 48% voltage drop occurs under a solar irradiation of G=300 W/m2 .
The oscillating active current reference is introduced into the current loop generating
distorted grid currents. A filtering technique should be applied to remove such ripples
and to maintain the generated power with the high quality.
Fig. 4.34 shows the inclusion of a MAF in the dc-link voltage control loop. As it was
demonstrated in Section 4.4.1, the minimum window-width to be applied to remove the
second-order harmonics is Tw = T /2, i.e., for a grid frequency of f = 50 Hz, Tw should
be 10 ms. The filtering path is activated when a voltage drop more than 10% occurs at
Chapter 4. Performance of GCPPPs with Positive Sequence Control 94
eabc V
idref
iabc
Figure 4.33: Example of a 48% 2LG voltage sag at the MV side of the transformer
under solar irradiation G=300 W/m2 : (a) voltages at the LV side, (b) active current
reference, and (c) output currents.
Vdc* idref
kiv
k pv
s
vdc
the terminal of the PV generator (LV side of the transformer in Fig. 2.1). In this scheme,
the MAF filters the error between the dc-link voltage and its reference. This filtering
stage will affect the dynamic of the dc-link voltage significantly. If the PI controller
is relatively fast, the system may become unstable due to the delay introduced by the
MAF. An example of this is depicted in Fig. 4.35 when operating with PI parameters
kpv = 18 A/V and kiv = 2879 A/(Vs). Therefore, a stability analysis should be performed
to obtain proper values for the PI parameters.
In this section, a stability analysis is performed when the VSI has a MAF in the path
of the dc-link voltage error. To perform this analysis, the model of the dc-link including
the PV array is obtained, as illustrated in Fig. 2.18. Considering (2.25)-(2.30), the
relationship between Vdc (s) and Id (s) can be derived. By converting the MAF operator
(4.4) into the s-domain one obtains:
1
X(s) = [1 eTw s ]X(s). (4.38)
Tw s
Chapter 4. Performance of GCPPPs with Positive Sequence Control 95
eabc V
vdc V
idref A
4000
A
2000
0
idref
-2000
iabc A
Figure 4.35: Example of a 48% LG voltage drop with solar irradiation G=300 W/m2
when operating with a MAF in the voltage loop: (a) voltages at the LV side, (b) dc-
link voltage, (c) active current reference before the limiter, (d) limited active current
reference, and (e) grid currents.
The bode diagram of (4.38) is shown in Fig. 4.36(a). Equation (4.38) is a nonlinear
equation; it will be linearized around its operating point to perform a stability analysis.
The method used in this study is based on pade approximation, which is available
in MATLAB toolbox. This approximation is used to linearize a time delay transfer
function. P ade (T, n) returns the nth-order pade approximation of the continuous-time
delay exp(T s) in transfer function form. Three order approximations of the MAF are
considered to select the most convenient one for this study.
Magnitude [dB]
I I I
I I I I I I Ill I I I I I I I I
-100 ---,--,-,-rrrrn----,--r-r-rrr ---- -
I I I I I I Ill I I I I I I I I
I I I I I I Ill I I I I I I 11
-200 ---,--,-,-rrrrn----,--r-r-rrn
I I I I I I Ill I I I I I I I I
I I I I I I Ill I I I I I I 11
-300 ---,--,-,-rrrrn----,--r-r-rrn
0
I I I I I
Phase [deg]
-45 ---~--~-~-~~~~U----
1 I I I I I Ill
I I I I I I Ill I I
-90 ---,--,-,-rrrrn----,--r-r
I I I I I I Ill I I I I
-135 ---4--~-~-~~~~H----~--~-r4-~
I I I I I I Ill I I I I I I I
-180 ---~--~-~-~~~~U----~--~-~~-~~
0 1 2 3
10 10 10 10
Frequency [Hz]
(a)
0
I I I I I I Ill I I I I I
I I I I I I Ill I I I I I I I I I
Magnitude [dB]
I I I I I I Ill
I I I I I I Ill I I
-90 ---~--,-,-rTITr~---~--~-~
I I I I I I Ill I I I I I
-135 ---~--J_J_L!~!LL ___ J __ ~-~-Ll
I I I I I I Ill I I I I I I
I I I I I I 11! I I I I I I I
-180
0 1 2 3
10 10 10 10
Frequency [Hz]
(b)
Figure 4.36: Bode plot of the MAF with a window width of 10 ms: (a) actual MAF
and (b) MAF approximations.
Vdc* X ( s) X ( s) PI
I d ( s) Vdc ( s )
Gp
+ Controller
filter for this second-order components. The closed-loop diagram of the dc-link voltage
in s-domain is indicated in Fig. 4.37 in which Gp (s) is defined by (2.25)-(2.30) and the
MAF is substituted by its 4th order approximation.
The T an method in [138] is chosen to find out the stable region of the PI control
parameters in the closed-loop system. The general transfer function of the plant and
Chapter 4. Performance of GCPPPs with Positive Sequence Control 97
N (s)
G(s) = (4.39)
D(s)
ki
C(s) = kp + (4.40)
s
in which N (s) and D(s) are the nominator and denominator polynomials of the G(s)
transfer function, respectively. Substituting s = j and separating N (s) and D(s) into
their real and imaginary parts, the transfer function can be re-written as:
Ne ( 2 ) + jNo ( 2 )
G(j) = . (4.41)
De ( 2 ) + jDo ( 2 )
Determining the characteristic polynomial and imposing it to zero, the final equations
for the PI parameters are:
X()U () Y ()R()
kp () = (4.42)
Q()U () R()S()
Y ()Q() X()S()
ki () = (4.43)
Q()U () R()S()
where:
Q() = 2 No ( 2 ), R() = Ne ( 2 )
S() = Ne ( 2 ), S() = No ( 2 )
X() = 2 Do ( 2 ), Y () = De ( 2 ). (4.44)
This method is applied to the dc-link control loop shown in Fig. 4.37 to determine
the stable region of the PI control parameters. Different factors have different impacts
on the dc-link transfer function Gp (s) and eventually on the stability of the dc-link
loop. Therefore, these factors should be taken into the consideration to obtain the final
acceptable region for the PI controller parameters as explained in the following.
The impact of the voltage drop depth on the stability region of the dc-link voltage
control loop is analyzed first. In this analysis, the method discussed in Section 4.4.2.2
is taken into account to measure the voltage drop. For unbalanced voltage sags beyond
10% of the rated rms voltage value, the oscillation in the active current reference may
lead to a highly distorted output current or even PV power plant instability. On the
Chapter 4. Performance of GCPPPs with Positive Sequence Control 98
other hand, in the case of a 50% voltage drop and more, the grid codes indicate that the
reactive current injection should be equal to the rated current of the PV power plant.
Therefore, the i0dref is almost zero with negligible oscillations compared to the amount
of reactive current injection. The stable boundary for the cases of 50% and 10% voltage
sags is studied and depicted in Fig. 4.38(a). As it can be observed, the more severe the
voltage sag, the wider the stability region for the PI controller. Therefore, the case of
10% voltage drop is considered as a worst case scenario for the design of the controller.
Now, the impact of injecting reactive current during a voltage sag on the stability region
is assessed. Considering STC, i.e., G = 1000 W/m2 and temperature Temp = 25 C, the
stability region is obtained for two cases:
activation of the droop control during the voltage sag where Iqref 6= 0, and
Applying these two cases to (2.25)-(2.30), and considering a 10% voltage sag, the
stable regions are obtained as illustrated in Fig. 4.38(b).
The stable region without reactive current is smaller compared to the reactive current
injection case. It has been observed that the more injected reactive current, the wider
the stability region is. Therefore, the worst case scenario is considered in the following,
i.e., no reactive current injection.
The stability region is also assessed for different irradiation conditions. The most re-
stricted cases from the previous subsections are selected, i.e. a 10% voltage drop with
no reactive current injection. Two solar irradiation conditions are assessed:
4000
- - 10% voltage""!!
3000 - - SO% voltage sag
"'
~ 2000
kiv
00 10 20 30 40 so 60 70
k pv A/V
(a)
I qref 0
I qref 0
kiv
k pv A/V
k pv A/V
(b)
2000
- G-100W/m2
- G=lOOO W/m2
I I
~ I
--+----+----
I
I
I
kiv
I
---+-I ---
I
0
0 s 10 1S 20 2S 30 3S 40
0
0 s 10 15 A/V
k pv 20 2S 30 3S 40
pv A/V
k(c)
..,
..,~
~
v
iv
s 10 1S 20 2S 30 3S 40
k pv A/V
(d)
--------1----------r---------- --------
iv
: ;
10 15 20 25 30 35 40 45 50
k pv A/V
(e)
Figure 4.38: PI control parameters: (a) for different voltage sag intensities, (b) for
different reactive current injections, (c) for different solar irradiation conditions, (d) PI
parameters without the MAF under rated grid voltages, and (e) final stable region also
considering optimal operation without the MAF.
Chapter 4. Performance of GCPPPs with Positive Sequence Control 100
G
Impp (G) = ImppST C (4.46)
GST C
G
Voc (G) = VocST C + VT log (4.47)
GST C
G
Vmpp (G) = VmppST C + VT log , (4.48)
GST C
where VT is 25 mV, and the subscripts oc and sc stand for open and short circuit,
respectively. Note that the number of PV modules connected in series and parallel also
should be considered.
The stability regions for different solar irradiations are depicted in Fig. 4.38(c).
Observe that the case of low irradiation restricts the range of PI values more than
operating with high irradiation conditions.
Including a MAF in the dc-link control loop avoids low-frequency distortion in the grid
currents during unbalanced voltage sags. However, the MAF produces a delay and
compromises the dynamic of the GCPPP. Since most of the time the grid voltages are
relatively balanced and close to the rms rated value, the MAF is not required for normal
operating conditions. Therefore, the MAF should be removed from the dc-link control
loop when the rms values of the grid voltages are at least 90% of the rated value.
Consequently, the GCPPP performance with no MAF in the dc-link control loop should
also be studied. This will imply adding more restrictions to the selection of PI parameter
values.
The closed loop transfer function of the dc-link loop without the MAF is:
kpv s+kiv
sGp
Gcl = , (4.49)
k s+k
1 + pv s iv Gp
Two solar irradiation cases are considered under rated grid voltage conditions: (i)
G=1000 W/m2 and (ii) G=100 W/m2 . The transfer function Gp in these two cases is:
17.74
G = 1000 W/m2 Gp (s) = , (4.50)
s + 1.1114
17.87
G = 100 W/m2 Gp (s) = . (4.51)
s + 0.1115
Chapter 4. Performance of GCPPPs with Positive Sequence Control 101
Applying the transfer functions in (4.50) and (4.51) into (4.49) and compromising
between rise time (tr ) and settling time (ts ) with a damping factor =0.707, the resul-
tant plot showing kiv vs. kpv considering the smallest stable region in Fig. 4.38(c) is
illustrated in Fig. 4.38(d).
The two discontinuous lines in Fig. 4.38(d) are very close to each other. Therefore,
the changes in solar irradiation do not effect much on the performance of the dc-link
control loop. Hence, the curve for STC is considered for further studies.
In order to analyze the dc-link voltage loop independently of the current control
loops, the dynamic of the dc-link voltage loop should be at least five times slower than
that of the current controllers. Therefore, considering a settling time for the current
control loops of tsI =2 ms, the settling time of the dc-link control loop should be at least
tsV =10 ms. As a result:
4 4
tsV = = (1.1114+17.74kpv )
10ms kpv 45. (4.52)
N
2
However, the dynamic of the dc-link controller should limited to a maximum value.
A practical condition can be:
The limits of the region for kpv , defined by (4.52) and (4.53), are represented in Fig.
4.38(e) in red lines.
Considering all the restrictions, the best performance of the GCPPP is achieved
selecting values for kpv and kiv on the cyan curve (=0.707) between the point A and
B. Choosing kpv =12 A/V and kiv =1290 A/(Vs) from Fig. 4.38(e), Fig. 4.39 shows
simulation results under a 2LG voltage sag with 48% voltage drop when G=300 W/m2 .
As it can be observed, the MAF is able to remove the oscillations from the active current
reference and consequently the grid currents are sinusoidal. Furthermore, the stability
of the GCPPP is not compromised.
400
-
200 ........... r ........... ---- - '... .. -_ .. - .. - ... - . .......... -. ,....... ----
eabc V
0
-200
-400
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
(a)
850
111~ 11
vdc V
800
750
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
(b)
1500
1- + !:~
:.; ; 1 ~ i
idref A
1000 ! ! \\...----+-----!
500
0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
(c)
2000
iabc A
0
-2000
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
(d)
Time [s]
Figure 4.39: Operating with a MAF in dc-link control loop with a 48% 2LG voltage
drop under solar irradiation G=300W/m2 when choosing proper PI parameters: (a)
voltages at the LV side, (b) dc-link voltage, (c) active reference current, and (d) grid
currents.
The stability curve for this setup is depicted in Fig. 4.40. A LG with 100% voltage
sag is applied to phase a at the grid side of the transformer. The reactive current
reference is chosen to be zero in order to show the generated active current reference
without being limited. The results with no MAF in the dc-link voltage control loop are
shown in Fig. 4.41(a), while the improved results after applying the MAF with proper
parameters for the PI controller chosen from Fig. 4.40 (Table 4.3) are illustrated in
Fig. 4.41(b). It can be observed how the oscillations in the active current reference are
practically removed when the MAF is included in the dc-link voltage control loop.
Chapter 4. Performance of GCPPPs with Positive Sequence Control 103
k pv A/V
Figure 4.40: Final stable area of the dc-link voltage PI controller for the experimental
setup.
4.6 Conclusion
Performance requirements of GCPPPs under fault conditions for single- and two-stage
grid-connected inverters have been addressed in this chapter. Some modifications have
been proposed for controllers to make the GCPPP ride-through compatible to any type
of faults according to the GCs. These modifications include applying current limiters and
controlling the dc-link voltage by different methods. It is concluded that for the single-
stage configuration, the dc-link voltage is naturally limited and therefore the GCPPP is
self-protected, while in the two-stage configuration it is not. Three methods have been
proposed for the two-stage configuration to make the GCPPP able to withstand any
type of faults according to the GCs without being disconnected. The first two methods
are based on not generating any power from the PV arrays during the voltage sags, while
the third method changes the power point of the PV arrays to inject less power into the
grid compared with the pre-fault condition. The validity of all the proposed methods to
ride-through voltage sags has been demonstrated by multiple case studies performed by
simulations.
A method to improve the ac currents under unbalanced voltage sag conditions has
also been proposed by using a MAF to remove the second-order oscillations from the
Chapter 4. Performance of GCPPPs with Positive Sequence Control 104
eabc
iabc
idref
eabc
iabc
idref
Figure 4.41: Experimental results for the VSI under a LG fault with 100% voltage
sag in phase a at the grid side of the transformer: (a) without a MAF and (b) including
a MAF in the dc-link voltage control loop.
Chapter 4. Performance of GCPPPs with Positive Sequence Control 105
active current reference. Applying such a filter implies a delay and impacts on the
stability of the dc-link control loop. A stability analysis has been performed to obtain
proper PI parameters for that control loop.
In this chapter, the target has been on delivering constant balanced currents into
the grid even under unbalanced voltage sags. However, for delivering constant power
into the grid, the negative sequence of the currents should be controlled as well. In the
next chapter, this target will be addressed by controlling both positive and negative
sequences of the currents. However, It will be demonstrated that there exists some
limitations on using PI controllers in the current loops. As an alternative method, PR
controllers will be proposed with the capability to operate without a PLL for grid voltage
synchronization and also to address the FRT requirements.
Chapter 5
5.1 Introduction
This chapter discusses the control of large-scale GCPPPs operating under unbalanced
grid voltage sags. The positive- and negative-sequences of the grid currents need to be
controlled to regulate the power injected into the grid during unbalanced grid voltages.
It is shown in this chapter that the use of conventional PI-based controllers compromises
stability and dynamic performance of the inverter. The reason is the delays introduced
by the filters needed to extract the sequences of the transformed grid currents. Be-
cause of such delays, there is a strong restriction on choosing the parameters for the
current and voltage controllers which forces the GCPPP to perform slowly. This can
be improved by using resonant controllers instead, which avoid the need for filtering
the transformed grid currents. A PLL-less control structure for grid-connected VSIs is
also presented in this chapter which is able to perform very well without detecting the
angle of the grid voltages, even under strong unbalanced voltage sag conditions. Ad-
ditionally, over-current protection of the GCPPP during voltage sag processes is also
addressed. Simulation and experimental results are presented to evaluate and compare
the performance of the GCPPP when operating with the different controllers.
106
Chapter 5. Performance of GCPPPs with Pos-Neg Seq. Control 107
In order to extract the positive- and negative-sequences from the grid currents, the
measured dq-transformed variables have to be filtered to remove low-frequency ripples
produced under unbalanced grid voltages. The filtering process produces delays and may
make the VSI unstable. None of the references above has dealt with choosing proper
parameters for the PI controllers in the current control loops considering these delays.
no need for filtering the current components when using PR controllers, it provides the
GCPPPs with a faster dynamic avoiding the critical issues of stability.
The proposed controller is developed further achieving that the GCPPP can operate
without a PLL detector for grid voltage synchronization. This PLL-less strategy is tested
under various grid voltage conditions by simulation.
In this section, the VSI controllers used to control the positive and negative sequences
of the grid currents in a GCPPP are described.
The FSPLL introduced and evaluated in Section 4.4 is used in this study. It was shown
that this PLL-based detector performs well even under strong unbalanced voltage sags.
The instantaneous active and reactive powers under unbalanced voltage sags are
given by [137]
p = P0 + Pc2 cos(2t) + Ps2 sin(2t) (5.1)
where P0 and Q0 are average values of the instantaneous active and reactive powers
respectively, and Pc2 , Ps2 , Qc2 , and Qs2 are second-order harmonic terms of these in-
stantaneous powers. The power components in (5.1) and (5.2) expressed in terms of
transformed positive- and negative-sequences of voltages and currents (dq + and dq
frames) are [137]:
P0 = e+ + + +
d id + eq iq + ed id + eq iq (5.3)
Chapter 5. Performance of GCPPPs with Pos-Neg Seq. Control 109
Pc2 = e + + + +
d id + eq iq + ed id + eq iq (5.4)
+ +
Ps2 = e + +
q id ed iq eq id + ed iq (5.5)
+ + +
Q0 = e+
q id ed iq + eq id ed iq (5.6)
+ +
Qc2 = e + +
q id ed iq + eq id ed iq (5.7)
Qs2 = e + + + +
d id eq iq + ed id + eq iq (5.8)
where e+ +
d , eq , ed , and eq are the transformed components of the grid voltages (positive
and negative sequences), and i+ +
d , iq , id , and iq are the transformed components of the
grid currents (positive- and negative-sequences). It should be mentioned that (5.3)-(5.8)
are based on considering the dq transformation to be power conservative. This set of
equations can be converted into a 64 matrix which is not reversible. The terms Pc2 and
Ps2 are the ones that produce active power fluctuations under unbalanced voltage sags,
while the terms Qc2 and Qs2 can be disregarded because they represent reactive power
fluctuations. As the target in this chapter is to inject constant power into the grid, the
references for Pc2 and Ps2 are defined to be zero. Therefore, the current references can
be obtained as follows:
1
i+
d e+
d e+
q e
d e
q P0
+ + +
iq ed eq ed eq Pc2 = 0
= . (5.9)
+
id eq ed e+ e Ps2 = 0
q d
+
i
q
+
eq ed
eq ed Q0
If only the positive-sequence of the grid currents was controlled, using the angle obtained
through the PLL explained in Section 4.4, the transformed current components would be
constant with no ripples during unbalanced grid voltages. However, the power injected
into the grid would not be constant, producing large ripples to the dc-link voltage of
the GCPPP. To avoid this, constant power into the grid can be achieved by unbalancing
the currents, i.e. including a proper negative-sequence into the grid currents. When
the currents are measured and transformed into dq-components by rotating frames in
opposite directions (dq + and dq frames), both sequences interact with each other pro-
ducing ripples at twice the fundamental frequency, i.e. 100 Hz when the grid frequency
is 50 Hz. Therefore, a filtering technique is needed to remove these ripples. MAFs can
be used for this purpose, as shown in Fig. 5.1. The same procedure can be applied to
extract the positive- and negative-sequence components from the grid voltages.
Chapter 5. Performance of GCPPPs with Pos-Neg Seq. Control 110
iabc id id
iq iq
id id
iq iq
It has been proven in Section 4.4 that under unbalanced voltage conditions a MAF
with Tw = T /2, T being the grid period, can remove all the second-order harmonics from
the dq transformed components. The filtering process introduces a delay, regardless of
the filtering technique implemented. In the case of using MAFs, the delay is equal to
the window width (Tw ).
Two control loops are needed for the conventional control of the positive- and negative-
sequences. Each loop contains two PI controllers as well as decoupling terms (L) and
feed-forward terms from the dq-transformed grid voltages, as shown in Fig. 5.2 for
the positive sequence. The same control diagram is used in parallel for the negative-
sequence control. The only difference is that the sign of L is negative due to the inverse
direction of the negative-sequence frame rotation. The angle used for the negative-
sequence loop is the positive-sequence angle extracted with the FSPLL with a negative
sign, i.e. =+ . Finally, the summation of the positive- and negative-sequence
. In Fig. 5.2, i+ and
voltage components provides the voltage reference to the inverter vabc d
i+
q are the positive-sequence components of the currents obtained through the filtering
diagram shown in Fig. 5.1, e+ +
d and eq are the filtered positive-sequence components of
the grid voltages, and finally vd+ and vq+ are the positive-sequence voltage references
for the VSI. It should be mentioned that the grid voltage components used in (6.30) to
obtain the current references are the filtered ones through the MAFs, i.e. e+ +
d , eq , ed ,
and e
q . Therefore, these measurements are also delayed when there is a transient such
as a voltage sag.
Chapter 5. Performance of GCPPPs with Pos-Neg Seq. Control 111
ed
id* id vd vd* *
vdq
dq *
* vabc *
vabc
id v q
L
abc
iq
L
iq* iq vq
eq
ed
id* id vd vd* *
vdq
dq *
vabc
*
id v q
L
abc
iq
L
iq* iq vq
eq
Figure 5.2: Current control loops for the positive- and negative-sequences of the grid
currents.
Fig. 2.1 shows the case study of a three-phase inverter-based GCPPP. The system
specifications are defined in Table. 4.1. A LG with 55% voltage sag in phase a is applied
at the grid side of the transformer defined as MV side, while the ac voltages and currents
shown in the figures are taken from the LV side of the transformer.
In some related studies the dc-link voltage has been imposed to be constant by means of
a dc voltage source such as the one presented in [142] and shown in Fig. 5.3. However,
Chapter 5. Performance of GCPPPs with Pos-Neg Seq. Control 112
Vdc
Figure 5.3: Diagram of the GCPPP using a constant dc voltage source in the dc side.
id* 1 id
+ Ls R
id
iq* 1 iq
+ Ls R
iq
Figure 5.4: Average model of current control loops after compensating for the cou-
pling terms.
the dc-link voltage has to be regulated in a real GCPPP such as that shown in Fig. 2.1.
This issue is addressed in the following.
Before evaluating the VSI performance during the voltage sag, the stability of the current
control loops when applying the MAFs is analyzed. Consider the MAF operator (4.4)
into the s-domain as presented in (4.38). Once the coupling terms in Fig. 5.2 are
compensated, two independent control loops are obtained [150], as shown in Fig. 5.4
for the positive-sequence control. The same control loops are obtained for the negative-
sequence currents.
In order to evaluate the stability region for the PI control parameters, the MAFs are
linearized through the pade approximation method, which is available in MATLAB
toolbox. After linearization, the T an method elaborated in (4.39)-(4.44) is used to
Chapter 5. Performance of GCPPPs with Pos-Neg Seq. Control 113
12.-----~----~----~----~----~
10~;~';'~ 'H
kic V As
SrHooHoHo+ /'- oooooooHooo:ooHoHooj-oHoHooo ~ ooHoHoo H
6
Tw T / 4
4
2
Tw T / 2
k p c V I A
Figure 5.5: Stable region for the parameters of the PI current controllers considering
the MAFs for the extraction of the positive and negative sequences.
calculate the stable region for the PI controller parameters. Considering a 4th order
approximation for the transfer function of the MAFs given in (4.38), the stable region
for the PI controller parameters is the area enclosed by the lower curve (pink color) in
Fig. 5.5. Since the window-width of the MAFs in the current control loops is Tw =T /2,
they produce a delay of T /2. When using another filtering method that produce a
delay of a quarter of the fundamental period (T /4), such as that one used in [142], the
stable region is the area enclosed by the upper curve (blue color) in Fig. 5.5, which is
considerably larger. Therefore, as expected, smaller delays lead to wider stable area for
the PI controller parameters in the current control loops. In any case, the values of the
PI parameters have to be selected small, which makes the current control loops slow.
5.5.1.2 Simulations
Considering the lower curve in Fig. 5.5 (a filtering delay of T /2), the values selected for
the PI parameters are kpc =0.015 V/A and kic =0.15 V/As. In this study, the transformed
grid voltages are introduced as feed-forward terms in the positive- and negative-sequence
control loops (5.2). Since there are second-order harmonics in the transformed grid
voltages, they also need to be filtered by MAFs to extract the d and q components of
each sequence. Again, this filtering process produces some delays that deteriorate the
overall dynamic performance of the VSI.
Vdc* P0*
vdc
Figure 5.6: Dc-link voltage regulation to obtain the active power reference.
constant power into the grid, this requires the injection of a negative sequence in the
grid currents when the grid voltages are unbalanced. Despite the power injected into
the grid being constant, power fluctuation occur in the grid filter when the currents are
unbalanced. Because of this fluctuation, the dc-link voltage has a ripple at twice the
grid frequency. Such a ripple is relatively small, however it will produce oscillations in
the active power reference (P0 ), which will eventually distort the grid currents. In order
to avoid such distortion, a MAF is included in the dc-link voltage loop (Fig. 5.6) with a
window-width of Tw = T /2 that is able to eliminate the oscillations in the active power
reference, as explained in Section 4.5. The reactive current reference Q0 is determined
from the droop control described in Section 4.3.
The performance of the VSI is assessed using the same PI parameters of the Section
5.5.1.2 for the current control loops, providing a settling time of 36 ms. For the PI
controller of the dc-link, the parameters are kpv =3320 A/V and kiv =109 560 A/Vs with
a settling time of 76 ms. As shown in Fig. 5.7, the output currents oscillate, which may
lead to inverter disconnection. The reason for this is the slow dynamic behavior of the
current control loops that interacts with the voltage loop due to the inclusion of the
MAFs.
In order to avoid interaction between the voltage and current loops, the current
control loops should be at least five times faster than the dc-link voltage loop. This
would also allow to tune the control parameters separately. Therefore, an alternative to
improve the performance of the GCPPP is to re-tune the PI parameters of the external
loop to perform slower, although it will slow down the dynamic performance of the VSI.
The new parameters chosen for the PI controller of the voltage loop are kpv =1037.5 A/V
and kiv =22 410 A/Vs, which produce a settling time of 290 ms. The grid currents for
this case are depicted in Fig. 5.8(a) and the generated active and reactive powers are
shown in Fig. 5.8(b). As expected, the VSI is stable but its dynamic performance is
relatively slow.
Another reason for the slow dynamic performance is that the measured grid voltages
are also filtered to extract the positive- and negative-sequences to be used as feed-
forward terms in the current loops, as shown in Fig. 5.2. An improvement to this would
be to include the grid voltages eabc after the summation of the positive and negative
Chapter 5. Performance of GCPPPs with Pos-Neg Seq. Control 115
400
eabc V
. . . .
200
:
iabc kA
0
-200
-400
0
5
-= 0.05 0.1 0.15 0.2 0.25
(a)
0.3 0.35 0.4 0.45 0.5
-5
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
(b)
Time [s]
Figure 5.7: Example using a PI-based control scheme with regulation of the dc-link
voltage under a LG voltage sag produced at the MV side: (a) grid voltages and (b)
0 0.05
grid currents. 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
(a)
~5~ :! : ! : '.,!'.,.,.
1 ; : : j . j : .
-5~--~--~--~--~--~--~--~--~--~~
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
(a)
106
:
p
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
(b)
Time (s)
40 -...
Figure 5.8: Reduced PI parameter values in the dc-link voltage loop: (a) grid currents
and (b) output powers.
eabc V
-~:g i . .. . : . . . . I . : . . I .. : : i .. : . 1_
voltage references of the inverter, as shown in Fig. 5.9. In this case, the feed-forward
... l. . ... _1 ..... . .... .
voltage terms do not need to be separated into sequences, thus avoiding the need for
filtering. As a result, any change in the grid voltages is immediately transferred to the
references provided to the inverter without filtering delays. This direct feed-forward
strategy is applied to the GCPPP under study and the grid currents and injected active
and reactive powers into the grid are shown in Fig. 5.10. It can be observed that,
although the same parameters of the PI controllers are used, the dynamic performance
is improved.
Chapter 5. Performance of GCPPPs with Pos-Neg Seq. Control 116
iq* iq vq
id* id vd vd* *
vdq
dq *
vabc
id vq*
L
abc
iq
L
iq* iq vq
Figure 5.9: Current control loops with grid voltage feed-forward terms without filter-
ing.
5
iabc kA
0 .
.. . ..... : ' : : .' .
-5
. ' . ' ' ; : '
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
(a)
106
p W ,q VAr
2
1
p
0
q
-1
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
(b)
Time (s)
Figure 5.10: Results after applying the actual grid voltages as feed-forward terms:
(a) grid voltages and (b) output currents.
Chapter 5. Performance of GCPPPs with Pos-Neg Seq. Control 117
idq *
dq
i*
eabc
+ v
+ *
i i v + v*
vabc + abc
+ abc
idq * i* v
dq
+
i
As discussed in the previous section, the use of PI controllers in the current loops to
control both the positive- and negative-sequences has important limitations. As an
alternative, PR controllers can be used. PR controllers perform on a stationary frame
( and ) and are able to track both sequences simultaneously. They have infinity
gain at the tuned frequency and therefore can achieve zero steady-state error at that
frequency. A control diagram with PR controllers in the current control loops is depicted
in Fig. 5.11.
Based on [151], the equivalent transfer function of the PI control in the stationary
frame can be obtained by applying a frequency shift of s s j for the positive-
sequence and s s + j for the negative-sequence. Applying these changes to (5.10)
and (6.16) , and adding both equations, the derived matrix expressed in the stationary
Chapter 5. Performance of GCPPPs with Pos-Neg Seq. Control 118
frame becomes:
" # " #" #
kic s
v kpc + s2 + 2
0 i
=2 kic s
. (5.12)
v 0 kpc + s2 + 2
i
As a result, the proper control parameters kpc and kic in (5.12) are twice the values of
the SRF determined when there is no signal filtering. Considering the control diagrams
in Fig. 5.4 without MAFs and with PI controllers as:
kic
GP I (s) = 2(kpc + ), (5.13)
s
kpc kic
2( L s+ L )
GCL (s) = 2k +R
. (5.14)
s2 + ( pcL )s + 2kLic
Comparing (5.14) with the standard second-order transfer function (2.31), one ob-
tains:
2kpc + R
2N = and (5.15)
L
2 2kic
N = . (5.16)
L
Two factors are considered to determine the control parameters: (i) achieving a good
compromise between rise time (tr ) and settling time (ts ), which is achieved by imposing
a damping factor = 1/ 2, and (ii) fast dynamics of the controller. Substituting
= 1/ 2 in (5.15), the relationship between kpc and kic becomes:
R 2
1
kic = kpc + . (5.17)
L 2
4 4
ts = = 2kpc +R
= 2ms kpc = 0.2 V /A, (5.18)
N
2L
Once kpc is defined, kic can be calculated from (5.17), obtaining kic = 1608 V/(As).
These values are used to define the parameters of the PR controller as kp = 2kpc = 0.4 V/A
and kr = 2kic = 3216 V/(As). The performance of the GCPPP under a LG voltage sag
is shown in Fig. 5.12. Comparing the results with those obtained previously using PI
controllers, the dynamic behavior with PR controllers is much faster.
Chapter 5. Performance of GCPPPs with Pos-Neg Seq. Control 119
eabc
iabc
106
,q
p
Figure 5.12: Results with current control loops based on PR controllers: (a) grid
voltages, (b) grid currents, and (c) injected active and reactive powers into the grid.
In order to satisfy the GCs [4] and protect the inverter while maintain its connectivity to
the grid under voltage sags, some modifications need to be implemented in the controller.
Based on these GCs, the priority is to support the grid voltages with the injection of
reactive power during a voltage sag, therefore the generated active power into the grid
should be limited if the resulting injected apparent power exceeds the rated power of the
0
inverter (SN =1.1 MVA). Thus, P0 in (6.30) is replaced by a new value P0 after passing
through the power limiter. The results after applying the power limiter is depicted in
Fig. 5.13.
Additionally, the grid currents should also be limited to avoid the inverter to be
disconnected. The proposed method is based on measuring the rms current references
of the three phases (imax ) and compare the maximum one with the rated current of the
inverter (In ). First, the rms value of the current references is obtained based on:
s Z t
1
ixrms = (ix )2 dt, (5.19)
Tw tTw
where x represents the three phases (x = {a, b, c}), and Tw is the window-width used
for the rms calculation, typically T /2 or T , T being the grid voltage period (T = 1/f ).
o- ; ... .~
. v
iabc k
: : : . : : .
~ . . . . . .
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
Chapter 5. Performance of GCPPPs with Time
Pos-Neg
[s] Seq. Control 120
400 ~
eabc V
200 . . . ... .. . .. .. .. . . . . . . . . .. . . . .. ... .. ... f
0 . ... . . . . ; ... ' ' ' ..
-200
-400 . . . .
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
(a)
i
iabc kA
~~ 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
(b)
1000
v dc V
l j j
900
800 L l : :
i if
700
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
(c)
106
p W ,q VAr
- ---ill
f~-t.~~.~v~vv~
. . . . . ~.\\Vo;~
. . . . . . . . ...~~.~~ii~
. . .~"""""""["""~-~\~~;v~
. . . . . .~ --~-=-< ~f.~ f~.
2
0 ~--
r 1
. ' . . .
.
. . q
-1~--~--~--~----~--~--~----~--~--~====
'. p
!.._:a
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
(d)
Time [s]
Figure 5.13: Performance of the VSI when applying the power limiter: (a) grid
voltages, (b) grid currents, (c) dc-link voltage, and (d) injected active and reactive
powers into the grid.
Then, the current references are re-scaled based on the factor frs defined as follows:
iIn
if imax > In ,
max
frs = (5.20)
imax
1 if In .
The diagram of this current limiter is depicted in Fig. 5.14. The new current
0
references (i ) are obtained from this diagram. The results after applying this method
are presented in Fig. 5.15. The grid currents and injected active and reactive powers
are re-scaled so that the inverter maintains its connectivity during the voltage sag while
supporting the grid with reactive power. The delay seen in frs , and consequently in the
output currents, is due to the time needed for the calculation of the rms values of the
grid currents.
The MAFs in the SRF control and the resonant controllers are frequency dependent.
Therefore, if the grid frequency changes, the GCPPP will not perform as expected.
Nevertheless, a frequency detector such as that proposed in Chapter 6 can be applied
Chapter 5. Performance of GCPPPs with Pos-Neg Seq. Control 121
*'
i
ia* ia*rms f sr
*
i
*
iabc ib* ib*rms *
imax Re-scaling
1----~ Max 1-----+1 Equation
L..----1 (5.20)
ic* ic*rms
Figure 5.14: Diagram of the proposed current limiter based on re-scaling the current
references.
0
400 -1
eabc V
200 0.1
0
-200
-400
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
(a)
1
f rs
0.8
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
(b)
iabc kA
2 : :
: . . :
0 .. . . . . .
-2
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
(c)
106
p W ,q VAr
2
1
p
0 q
-1
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
(d)
Time [s]
Figure 5.15: Application of the proposed current limiter: (a) grid voltages, (b) grid
currents, (c) re-scaling factor frs , and (d) injected active and reactive powers into the
grid.
Chapter 5. Performance of GCPPPs with Pos-Neg Seq. Control 122
to update the window width of the MAFs as well as the frequency of the resonant
controllers. This will make the controllers of the GCPPP frequency adaptive.
The performance of the grid-connected VSI with both PI and PR controllers in the
current control loops has been experimentally verified with the 2.8-kVA GCPPP setup
introduced in Section 2.6. The main parameters of this laboratory setup are given in
Table 5.1.
In the experimental tests, a 55% LG voltage sag is applied at the MV side. Fig. 5.16
shows the results using PI controllers when the positive- and negative-sequence of the
grid voltages are filtered and introduced as feed-forward terms (control diagram shown
in Fig. 5.2). The results in Fig. 5.16 can be improved by replacing the filtered voltages
with the actual voltages at the end of the current loops (Fig. 5.9), as demonstrated in
Fig. 5.17. The control diagram in Fig. 5.11 based on PR controllers is also implemented
and used to control the experimental setup. The performance of the VSI with PR
controllers is depicted in Fig. 5.18. Comparing the results in Fig. 5.18 with those in
Figs. 5.16 and 5.17, the best performance of the PR controllers versus PI controllers in
the current loop is demonstrated. Finally, Fig. 5.19 shows some results when applying
the power and current limiters in order to maintain the connectivity of the VSI under
grid voltage sags.
Figure 5.16: PI-based current control of the positive- and negative-sequences under
a 55% LG voltage sag at the MV side. From top to the bottom: grid voltages at the
LV side, output currents, injected active and reactive powers into the grid.
Figure 5.17: PI-based current control with the improved feed-forward term under a
55% LG voltage sag at the MV side. From top to the bottom: grid voltages at the LV
side, output currents, injected active power and injected reactive powers into the grid.
Chapter 5. Performance of GCPPPs with Pos-Neg Seq. Control 124
Figure 5.18: PR-based current control under a 55% LG voltage sag at the MV side.
From top to the bottom: grid voltages at the LV side, output currents, injected active
and reactive powers into the grid.
Figure 5.19: PR-based current control with power and current limiters under a 55%
LG voltage sag at the MV side. From top to the bottom: grid voltages at the LV side,
output currents, injected active and reactive powers into the grid.
Chapter 5. Performance of GCPPPs with Pos-Neg Seq. Control 125
the FSPLL in Subsection 5.3.1. This strategy avoids the requirement of phase angle
detection of the grid voltages and also the need for proper tuning of the PLL controller.
In the proposed method, the positive sequence angle (+ ) extracted by the FSPLL
is replaced by an arbitrary angle, such as:
+ = 2f t + 0 , (5.21)
where 0 is an initial angle. The angle + in (5.21) can be obtained by integrating the
angular frequency of the grid voltages ( = 2f ). For the negative sequence control,
the angle used is obtained by changing the sign to + in (5.21), as follows:
= + . (5.22)
With this method, the dq components of the positive and negative sequences of the
grid voltages extracted based on Fig. 5.1 are no longer synchronized with the grid.
Consequently, the dq component applied to (5.3)-(5.8) and to (5.9) generate different
current references compared to those obtained when the FSPLL was used. Nevertheless,
the power references remain the same and thus the GCPPP delivers the targeted active
and reactive powers to the grid.
In order to evaluate the proposed PLL-less method, two targets are pursued under
unbalanced voltage sags:
In the first case, although the instantaneous power injected to the grid is constant,
the VSI has to deliver non-constant power because the energy stored in the grid filter
Chapter 5. Performance of GCPPPs with Pos-Neg Seq. Control 126
oscillates under unbalanced voltage sags. In the second case, the VSI delivers constant
instantaneous power to the filter and grid, therefore the instantaneous power injected to
the grid is not constant. The practical interest of the second target is to attenuate/cancel
the low-frequency dc-link voltage ripples that typically appear during unbalanced grid
voltages.
The first target can be achieved by using (5.9). For the second target the oscillating
and P in (5.9) are defined as follows [75]:
active power components Pc2 s2
+ + +
Pc2 = 2R(i+
d id + iq iq ) 2L(id iq iq id ) (5.23)
+ + +
Ps2 = 2R(i+
d id iq iq ) 2L(id iq iq id ), (5.24)
in which R and L are the resistance and inductance of the grid filter located between
the VSI and the LV side of the transformer.
PR controllers are selected for the current control loops due to their better perfor-
mance demonstrated in Section 5.6.
Simulation results to demonstrate the good performance of the proposed PLL-less method
are presented in this subsection. A LG voltage sag with 70% voltage drop in phase a
is applied to the MV side of the transformer. The results when including the FSPLL
for angle detection are shown in Fig. 5.20, while the same results using the proposed
PLL-less method are illustrated in Fig. 5.21. One can observe that although the dq
components of the filtered voltages and currents are different, the resulting ac currents
and the active and reactive powers delivered to the filter and the grid are the same in
both cases. For convenience, the active and reactive powers delivered from the inverter
to the filter and the grid are named pi and qi , while the ones delivered to the grid are
named pg and qg , respectively. The currents in the faulty phases (blue and green phase
currents in Figs. 5.20 and 5.21) increase compared to the non-faulty phase (red phase
current) in order to achieve the requirement of delivering constant active power into the
grid. As expected, pg has no oscillations in both cases.
The proposed PLL-less method is evaluated for the second target, i.e., injecting
constant active power to the filter and grid. In this case, there is no need to include
the MAF depicted in Fig. 5.6 in the dc-link voltage loop as the oscillating terms are
removed when applying the terms given in (5.23) and (5.24) into the matrix (5.9). The
same voltage sag of the previous test is applied and the results are depicted in Fig. 5.22.
Chapter 5. Performance of GCPPPs with Pos-Neg Seq. Control 127
i:-
0 :':'::
eabc V
. . .... i . '... . .
-200 ..... ! .. . . . .. ' . . ' . . ! . .. . . +. . . . . . . ! ....
400~--~--~--~--~----~--~--~--~--~--~
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
(a)
edq V ed
eq
0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
(b)
ed
edq V
eq
0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
(c)
4
idq kA
2
: : i
0 . : -d
-2
\ j J - iq
4 i i
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
(d)
idq kA
id
!
. i[\/ L : !\
0 if iq
-1
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
(e)
:ll ~ f lll
v dc V
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
~
iabc kA
~~ 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
(g)
106
pi W ,qi VAr
l : ~ ! ll~
pi
qi
0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
(h)
106
pg W ,qg VAr
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
(i)
Time [s]
Figure 5.20: Performance of the GCPPP including a FSPLL for phase angle detection.
Controlling the active power injected into the grid under a LG voltage sag at the
MV side: (a) voltages at the LV side of the transformer, (b) dq components of the
positive-sequence voltages, (c) dq components of the negative-sequence voltages, (d)
dq components of the positive-sequence current references, (e) dq component of the
negative-sequence current references, (f) dc-link voltage, (g) grid currents at the LV
side, (h) active and reactive powers delivered to the filter and the grid, and (i) active
and reactive powers delivered to the grid.
Chapter 5. Performance of GCPPPs with Pos-Neg Seq. Control 128
eabc V
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
(a)
edq V ed
eq
0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
(b)
100,---,---,---,---,---,---,---,-------r=~
ed
edq V
' '
i i
................... ,....................; ................... ,.....................,.~-~-~-~
eq
0 1---~-,
i Jl
iq
i
-4~--~--~--~--~--~--~--~--~--~~
I
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
(d)
~ ~ i
:::j:_ ::: ::::: :: ::::::!_: ::::: :::::
j i i
mmom mm _jl ........................
i
!:_ . . . . . . . . . . . . . . . . . . . . . . .
i
-
iq
. . ' . . . . . .
-1~-~--~--~--~--~--~--~--~--~~
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
(e)
::111~ 111
v dc V
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
(f)
4
iabc kA
~~ 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
(g)
106
pi W ,qi VAr
] l aRk !! ~
pi
qi
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
(h)
106
pg W ,qg VAr
- -~-'- - v~_
- - -'- w~_
_ _ ._vw~.____
<wvv_~_.___ _ _ _._~
_ _ _ _.
ppgg
qqg
g
_;1._______.___I __.____I______.__
I
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
(i)
Time [s]
Figure 5.21: Performance of the GCPPP without PLL for phase angle detection.
Controlling the active power injected into the grid under a LG voltage sag at the
MV side: (a) voltages at the LV side of the transformer, (b) dq components of the
positive-sequence voltages, (c) dq components of the negative-sequence voltages, (d)
dq components of the positive-sequence current references, (e) dq component of the
negative-sequence current references, (f) dc-link voltage, (g) grid currents at the LV
side, (h) active and reactive powers delivered to the filter and the grid, and (i) active
and reactive powers delivered to the grid.
Chapter 5. Performance of GCPPPs with Pos-Neg Seq. Control 129
eabc V
v dc V
iabc kA
106
pg W ,qg VAr pi W ,qi VAr
i
i
106
g
g
Figure 5.22: Performance of the GCPPP without PLL angle detector. Controlling
the active power delivered to the grid filter under a LG voltage sag at the MV side:
(a) voltage at the LV side of the transformer, (b) dc-link voltage, (c) grid currents at
the LV side, (d) active and reactive powers delivered to the filter and the grid, and (e)
active and reactive powers delivered to the grid.
As demonstrated, the dc-link voltage and eventually the active power delivered to the
filter and the grid are almost constant under this unbalanced voltage sag.
As mentioned before, the priority under a voltage sag is the injection of reactive currents
into the grid. Therefore, the active currents may have to be limited to prevent from over-
currents. The current limiter proposed in this subsection is based on limiting the active
power reference only; while the method explained in Subsection 5.6.1 limits the active
Chapter 5. Performance of GCPPPs with Pos-Neg Seq. Control 130
In Voc
* * ier vdc Vdc* new
i imax ier 0 ? S Q
Q0* Vdc*
Q *
0 0 ? R Q!
Figure 5.23: The procedure of limiting the excessive ac currents by changing the
operating point of the PV array.
power reference first and then it limits both the active and reactive current references
by means of a re-scaling factor.
For designing this new current limiter the only information required is the ac currents
and the activation signal of reactive power injection. The maximum value of the rms
current references of the three phases (imax ) is measured and compared with In . The
difference (ier = imax In ) goes through a PI controller and the output is added to the
dc-link voltage reference as a feed-forward term which forces the voltage of the PV array
to move away of the MPP toward the open-circuit voltage Voc , as depicted in Fig. 4.9.
This in turn forces the grid currents to decrease until the operating point reaches a new
steady-state point in which imax becomes equal to In .
Due to the accumulation of error in the integral part of the PI controller, the PI
should be reset after fault removal. For this, two conditions need to be considered:
(i) ier 0 and (ii) deactivation of reactive power injection. Some logic needs to be
implemented with the following functionalities:
if ier 0 and the reactive power injection is activated no action (maintain the
previous situation),
This can be applied by using a S-R flip-flop. The diagram of the proposed current
limiter is illustrated in Fig. 5.23. The new dc-link voltage reference is defined as Vdcnew
< V
where Vdc dcnew < Voc . By resetting the PI controller, vdc becomes zero and dc-
. It should be mentioned
link voltage reference returns back to the prefault value Vdc
was obtained from the MPPT technique embedded in
that the previous reference Vdc
the inverter control. During the voltage sag, the MPPT is disabled temporary and Vdc
is set to the prefault value. The MPPT technique is activated again after fault removal.
The performance of the proposed current limiter is depicted in Fig. 5.24 for a
maximum apparent power of 1.1 MVA when controlling constant active power during
Chapter 5. Performance of GCPPPs with Pos-Neg Seq. Control 131
eabc V
ier kA
vdc V
iabc kA
pg W ,qg VAr
106
p-grid
q-grid
Figure 5.24: Performance of the GCPPP after applying the current limiter when
controlling the active power injected into the grid under a LG voltage sag: (a) voltage
at the LV side of the transformer, (b) difference between the maximum rms measured
current and the maximum acceptable (Ier ), (c) dc-link voltage, (d) grid currents at the
LV side, and (e) active and reactive powers delivered to the grid.
the same voltage sag of previous examples. As it can be observed, the active power
is reduced while the average value of the reactive power remains unchanged compared
with Fig. 5.21.
5.9 Conclusions
This chapter has analyzed different alternatives for improving the dynamic performance
of GCPPPs under unbalanced voltage sags. The positive and negative sequences of the
grid currents have been controlled to regulate the power injected into the grid. This
study has shown the restrictions in the selection of the parameters when using a control
diagram based on PI controllers in the current control loops. The need for filtering to
Chapter 5. Performance of GCPPPs with Pos-Neg Seq. Control 132
separate the dq current components of the sequences limits the selection of parameter
values for stable operation of the GCPPP. As a consequence, the dynamic performance
becomes very slow. On the other hand, the use of a control diagram based on PR
controllers improves the dynamic performance of the grid-connected VSI because the
sequences do not need to be separated for the current control loops.
A PLL-less method has also been proposed in this chapter that achieves the same
results than using a phase angle detector for the current controllers. With the proposed
method, no technique to extract the grid voltages angle and thus no tuning of the PLL
control parameters has been required. Two new current limiters have also been proposed
in this chapter. The first one has been based on applying a two-stage power and current
limiter. First, the active power has been limited based on the GC requirement of reactive
power injection and the apparent power of the VSI. Then a re-scaling factor has been
applied to limit both the active and reactive current references to avoid over-currents.
The second method has been based on changing the operating point of the PV array by
increasing the dc-link voltage reference. The benefit of this method is that it only limits
the active power and the required reactive power is injected into the grid.
The method proposed in Chapter 4 for protecting the non-faulty phases from over-
voltage has the drawback of limiting the reactive current injection in all the phases
equally. This may avoid the GCPPP to address the FRT requirement properly. As
a result, in the next chapter, a novel method based on controlling the phase currents
independently will be proposed. This allows the injection of different reactive currents
to each phase. For this purpose, the angle of each phase voltage needs to the detected
individually which will be carried out by designing a frequency-adaptive PLL capable
of extracting individual phases.
Chapter 6
6.1 Introduction
Independent control of each current of a three-phase VSI under unbalanced voltage sags
is proposed in this chapter. Based on the grid codes (GC), the GCPPPs should support
the grid voltages by injecting reactive currents during the voltage sags. Such current
injection must not lead the grid voltages of the non-faulty phases to exceed 110% of their
nominal value. However, grid overvoltage can occur in the non-faulty phases, especially
if the currents injected into the grid are balanced. In this chapter, this problem is
addressed by injecting unbalanced currents into the grid during the voltage sags with
independent control for each phase. For this purpose, a phase-lock loop (PLL) technique
capable of extracting the individual angle of each phase independently within the context
of a three-phase system is proposed. The PLL utilizes moving average filters and is very
robust under unbalanced and highly distorted grid voltages. A frequency estimator
is also presented and incorporated in the PLL angle detector to make it frequency
adaptive. Experimental results taken from a 2.8-kVA inverter are presented confirming
the effectiveness of the proposed control method.
133
Chapter 6. Performance of GCPPPs under Voltage Sags with Ind. Phase Control 134
The injection of balanced reactive currents to support the grid during unbalanced volt-
age sags may lead to overvoltage in the non-faulty phases as demonstrated in Chapter
4. To prevent from this, the new GCs require the injection of unbalanced reactive cur-
rents during unbalanced voltage sags. In [50] and [51], a flexible voltage support method
was introduced based on the type and severity of the voltage sags. For this purpose,
the amount of reactive power injected via positive- and negative-sequence is controlled
with an off-line control parameter. An extended generalization of previous studies was
reported in [152] in which the reactive power reference and the control parameter were
updated in order to restore the dropped voltage amplitudes. Another study in [153]
proposed a method to set the positive- and negative-sequence reactive power references
based on an equivalent impedance grid model to avoid over- and under-voltages in the
phases. In that paper, the new current references were updated based on the previous
reactive power references. A decoupled double synchronous reference frame current con-
troller was introduced in [154] with the capability of controlling the active and reactive
power of the positive- and negative- sequences independently. Regarding the individual
control of currents and voltages of the three phases, some research was reported in [52]
to support the phases with unbalanced reactive power. Nonetheless, the method used in
that paper was not universal for all types of voltage sags. Therefore, a way of avoiding
overvoltage in non-faulty phases is the use of individual controllers per phase without
losing the overall performance of the three-phase system. In such controllers and also in
the case of unbalanced networks, individual detection of the phase voltages is required.
The objective of this chapter is to propose a control method based on the individual
control of the phase currents under unbalanced voltage sags. The amount of reactive
current in each phase is determined based on the amount of voltage drop in that phase,
which implies no reactive current injection for the non-faulty phases. The implementa-
tion of this method requires the knowledge of the grid voltage angle of each phase. For
this purpose a new phase extraction method is proposed as well. A frequency detector
is also proposed in this chapter to update the window width of the MAFs and allow the
phase angle detectors to operate under frequency changes, hence making the overall pro-
posed PLL also frequency adaptive. Moreover, the grid currents including both active
and reactive currents are limited in order to protect the GCPPP from ac over-currents,
addressing the FRT requirement. Since the grid currents are defined independently for
each phase, two methods are proposed to avoid the controllers from trying to inject a
zero sequence into the grid.
Chapter 6. Performance of GCPPPs under Voltage Sags with Ind. Phase Control 135
ea eb ec
a b c
Figure 6.1: Proposed individual phase angle detector based on the FSPLL.
This section proposes a PLL able to extract the phase angle of each of the grid voltages
in the context of three-phase systems. The extracted information allows the independent
control of both the active and reactive power for each phase.
In this chapter, the implementation of the FSPLL described in Chapter 4 Section 4.4,
based on [6] is used to extract the individual angles of each phase in a three-phase
grid. This information is necessary to perform individual control for each phase. Three
FSPLLs are used in this case, one per phase, as shown in Fig. 6.1. A single-phase
voltage is introduced to each FSPLL while the other inputs are imposed to be zero, as
follows: ea0 = (ea , 0, 0), eb0 = (eb , 0, 0) and ec0 = (ec , 0, 0) in which ea , eb and ec are the
phase voltages. In order to extract the angle of a specified phase, it should be remarked
that the voltage of that phase should be the first component of the three inputs of the
FSPLL, as shown in Fig. 6.1.
A general set (ex , 0, 0) for x = {a, b, c} is introduced to the FSPLLs in Fig. 4.4, as shown
in Fig. 6.1. The set is firstly converted into d-q components, as follows:
edx ex
h i
eq = Tdq0 0 , (6.1)
x
e0x 0
ex = Ex cos x , (6.2)
with
x = 2f t + x , (6.3)
where Ex is the amplitude of the phase voltage and x the initial angle of that specific
phase. Considering (2.20), (6.2) and (6.3) and substituting into (6.1), one obtains:
r
2
edx = Ex cos f cos x (6.4)
3
r
2
eqx = Ex cos f sin x . (6.5)
3
= f x . (6.6)
Substituting x from (6.6) into (6.4) and (6.5) and using a trigonometric identity,
edx and eqx become
Ex
edx = [cos() + cos(2f )] (6.7)
6
Ex
eqx = [sin(2f ) sin()]. (6.8)
6
If the window-width of the MAFs in Fig. 4.4 is Tw = 1/(2f ), the terms with 2f in
(6.7) and (6.8) are filtered and removed. Therefore, the outputs of the MAFs become
Ex
edx = cos() (6.9)
6
Ex
eqx = sin(). (6.10)
6
Applying the inverse matrix of (2.20) to (6.9) and (6.10) with the same arbitrary
angle f , the positive sequence is obtained in a stationary abc frame, as follows:
cos x
Ex
e+
= cos(x 2 . (6.11)
abcx 3 3 )
2
cos(x + 3 )
Chapter 6. Performance of GCPPPs under Voltage Sags with Ind. Phase Control 137
As it can be observed in (6.11), the voltage sequence obtained does not depend on
, and the angle of the first component is x . A conventional PLL stage (SRF-PLL in
Fig. 4.4) obtains the angle of this sequence, which coincides with the angle of the first
component, as it is demonstrated in the following.
The output angle of the SRF-PLL is x+ and this angle is used in the d-q transfor-
mation (2.20) to transform the set of voltages (6.11). It is obtained:
Ex
e+
dx = cos(x+ x ) (6.12)
6
Ex
e+ +
qx = sin(x x ). (6.13)
6
and
Ex Erms
e+
dx = = ph , (6.15)
6 3
where Ermsph is the rms value of the phase voltage (6.2).
The proposed PLL is capable of extracting the angle of each phase, and based on the
e+
dx values the magnitude of the input voltage can be derived from (6.15). In addition,
the summation of the extracted e+
dx for x = {a, b, c} through individual PLLs gives the
average value of the rms line-to-line voltages:
Erms = e+ + +
da + edb + edc . (6.16)
The proposed PLL is capable of extracting the angle of each phase under nominal grid
frequency (fn ). However, if the frequency of the grid changes, the accuracy of the angle
detection is affected. Two parameters are highly dependant on the grid frequency, i.e.
the angle f in the d-q transformation (2.20) and the window width of the MAFs (Tw ).
The difference between the frequency used in the d-q transformation and the MAFs
(fn ) and the grid frequency (f ) produces d-q components with a frequency equal to this
difference (fn f ). In the case of the MAFs, if its window width (Tw ) does no match
half the period of the grid voltages, even order harmonics of the grid frequency appear
at the output of the MAF. All these result in inaccuracy in the detected angle.
Chapter 6. Performance of GCPPPs under Voltage Sags with Ind. Phase Control 138
In order to make the proposed PLL frequency adaptive, a frequency detector should
be implemented to update the window width of the MAFs and the frequency for the angle
f . There are many methods to detect the grid frequency in the technical literature [6, 44,
155, 156]. However, most of them produce significant transients under grid frequency
changes. When the measured frequency is used in the angle detector introduced in
Section 6.3.1, these transients produce measured angles with excessive error. This is
why a new method for detecting the grid frequency with limited transients is needed.
The frequency detector introduced here is based on the same FSPLL structure shown
in Fig. 4.4 but in this case used to extract the angular frequency instead of the angle.
The input is the three-phase voltages and the MAFs and f in the proposed frequency
detector are designed based on the nominal frequency fn . It is assumed that the angular
frequency of the input voltages is , the angular frequency of the MAFs and f is n ,
and the output angular frequency is o . A balanced set of grid voltages such as those
in (2.38) are introduced. The voltages obtained at the output of the filtering stage are:
N cos t M sin t
e+
=K 2 2 , (6.17)
abc N cos(t 3 ) M sin(t 3 )
2 2
N cos(t + 3 ) M sin(t + 3 )
where:
E
K= , (6.18)
(n )Tw
N = sin[(n )Tw ], and (6.19)
Fig. 6.2 shows the illustration of (6.22). The proportional-integral (PI) controller
forces e+
q to be zero. Since the term o is multiplied by the time (t), o has to be
Chapter 6. Performance of GCPPPs under Voltage Sags with Ind. Phase Control 139
Low-Pass Filter
M
1
1
fo
N
1 s 2
c
t 3
e q o
KN
2
ot o 1
s
Figure 6.2: Linearized SRF-PLL when the input frequency is different from the nom-
inal value.
the same as , otherwise the error produced in the loop grows or decreases continuously.
Therefore, the PI controller generates an angular frequency o that is not approximated
but exactly (o = ). Furthermore, the initial angle o is such that the disturbance
M/N is canceled, i.e. o = M/N .
If the grid voltages were unbalanced, it can be demonstrated that the detected angu-
lar frequency (o ) would contain, in addition to the dc component providing the value
of , other components with angular frequencies of 2(n ) and 2n . The compo-
nent with angular frequency 2(n ) has a very small amplitude and therefore can be
neglected. The component at 2n can be attenuated by a low-pass filter. This is the
reason why Fig. 6.2 includes a low-pass filter.
As it will be shown in Section 6.3.3, if the input frequency changes suddenly, large
transients may be produced in the detected frequency. In order to attenuate such tran-
sients, a limiter to restrict the slope of the measured frequency is proposed. The method
is based on comparing two consecutive samples of the measured frequency and when the
difference is more than the slope limiter value (lim), the output frequency is restricted
based on that slope. However, when the consequent samples are relatively close, such
as when the input frequency has a tendency to be stable, the frequency value is set to
the later frequency sample. The overall implementation for this frequency improvement
strategy is given as follows:
fo (k) if abs(D1 (k)) < lim
foimp (k) = fo (k 1) + sign(D1 (k))lim if abs(D1 (k)) > lim & abs(D2 (k)) < lim
oimp (k 1) + sign(D2 (k))lim if abs(D1 (k)) > lim & abs(D2 (k)) > lim
f
where:
fo
f oimp
Time
Figure 6.3: Illustration of applying the slope limiter to the detected frequency (fo )
to obtain an improved frequency (foimp ) under frequency transients.
Fig. 6.3 shows by simulation the performance of the frequency limiter under a strong
transient in frequency changes. The detected frequency preserves fast dynamics while
avoiding strong transients.
The proposed PLL is tested experimentally using a platform based on the dSpace 1103.
In these experiments, the rated rms line-to-line voltage is 415V and the frequency
f =50 Hz. The cut-off frequency for the first-order low-pass filter in Fig. 6.2 is set
to c = 400 rad/s. The MAF window width is generally defined as 10 ms.
Fig. 6.4 shows results when a 2LG with 40% and 60% voltage sag in phase a and
b is produced in the grid voltages. As it can be observed, the angles are detected very
accurately even under such a disturbance.
The proposed detector is also tested under frequency changes. The performance
of the frequency detector as well as the PLLs is illustrated in Fig. 6.5 when the grid
frequency drops from 56 Hz to 47.5 Hz. The slope limiter is not applied to the frequency
detector in this representation.
The frequency detector is fast and it reaches the new value in less than a cycle.
However, the error in the frequency detection during the transients produces significant
errors in the detected angles. The band-width of the low-pass filter in Fig. 6.2 can
be reduced to avoid such large transients, however it would reduce the dynamic of the
frequency detector and consequently the proposed PLL.
The slope limiter introduced in Subsection 6.3.2 , with a maximum slope defined as
50 Hz/s, is applied to the measured frequency. New experimental results are obtained
Chapter 6. Performance of GCPPPs under Voltage Sags with Ind. Phase Control 141
eabc
a
Erms
Figure 6.4: Proposed PLL under unbalanced voltages. From top to the bottom:
voltages with a 2LG voltage sag (40% in phase a and 60% in phase b), detected angles
of phases a, b, and c, and the summation of the extracted d- components of three
independent phase voltages.
eabc
fo
Figure 6.5: Proposed frequency detector performance when the frequency jumps from
56 Hz to 47.5 Hz and vice versa. From top to the bottom: voltages with the frequency
changes, detected frequency, detected angles of phases a, b, and c.
Chapter 6. Performance of GCPPPs under Voltage Sags with Ind. Phase Control 142
eabc
f oimp
Figure 6.6: Performance of the overall system with the slope limiter in the frequency
detector. From top to the bottom: voltages under four consequent different conditions
(normal conditions, frequency change from 56 Hz to 47.5 Hz, phase jump of +/6 in
phase b, and 40% voltage drop in phase a of the input voltages), measured frequency
(foimp ), detected angles of phases a, b, and c.
and shown in Fig. 6.6 under different test conditions, i.e. a frequency change, a phase
jump, and unbalanced grid voltages. It can be observed that the angles detected are not
very distorted now because the large transients in the frequency detector are significantly
attenuated.
In this section, the grid currents of a GCPPP are controlled independently. The single-
phase PLL detector introduced in the previous section is used for the individual con-
trollers. The performance of the GCPPP operating with the proposed control method
is evaluated.
In this subsection, the way the current references are obtained to feed the current control
loops is presented. The amplitude of the active current (ia ) is defined to regulate the
dc-link voltage, while the individual reactive current amplitudes (irx ) are found from
Chapter 6. Performance of GCPPPs under Voltage Sags with Ind. Phase Control 143
vdc I n
*
Vdc id* ia
2 ia a
3
ea I n
En ph
dea ir a ia b
eb
deb ir b I n
ec
ia c
dec ir c
Figure 6.7: Control diagram to obtain the active and reactive current references.
where droop is a constant value, |dex | is the amount of phase voltage drop from its
nominal rms value (Enph ), and In is the amplitude of the nominal current of the
inverter. The dc-link voltage loop is controlled by a proportional-integral (PI) controller
equipped with an anti-windup technology that helps to reach the pre-fault values very
fast after fault removal. This can be seen in the control diagram of Fig. 6.7.
Under a voltage sag condition, the controller increases the active currents to maintain the
power injected into the grid. At the same time, reactive current needs to be injected into
the faulty phases to support the grid voltages. Consequently, the total phase currents
may increase above the maximum acceptable values, which would eventually trigger the
over-current protection. To avoid this situation, priority is given to the reactive current
injection to support the grid voltages. Therefore, the amplitudes of the active currents
are limited based on the reactive current required for each phase (Fig. 6.7).
The actual current reference for each phase is obtained by multiplying the amplitudes
of the active and reactive currents by cos and sin of the phase angle obtained from the
PLLs in Section 6.3, respectively. The final current reference for each phase is achieved
by adding the active and reactive current components. Fig. 6.8 illustrates the procedure
Chapter 6. Performance of GCPPPs under Voltage Sags with Ind. Phase Control 144
ia a
a ia*
ir a
to obtain the current reference ia for phase a. The current references for the other phases
are obtained using the same procedure.
Since the currents of the three phases are regulated independently, the sum of the three
currents may not be zero. This would mean the circulation of a zero-sequence current
component through the ground that will not happen if the ground circuit is open or, in
the case the ground circuit offers a low impedance, the circulation of this current may not
be a desired situation. Therefore, this zero-sequence should be removed from the current
references. This can be achieved by applying the Clarke transformation (abc/) to the
current references. In this case, the third component in the Clarke transformation, i.e.,
the or zero-sequence component, is disregarded. As a result, the current vector will
lay on the plane coinciding with its projection before the zero-sequence was removed.
Therefore, the components of the reference currents will be preserved.
where:
i0 = ia + ib + ic and (6.28)
ka = kb = kc = 1/3. (6.29)
Chapter 6. Performance of GCPPPs under Voltage Sags with Ind. Phase Control 145
During balanced operation, the common component i0 will be zero or very low.
However, during unbalanced voltage sags, the common component may take significant
0 0 0
values. Consequently, after applying (6.25)-(6.29), the new references ia , ib , and ic may
differ significantly respect to the original values. Therefore, the reactive components
of the non-faulty phases may increase, causing a voltage rise above the limits. An
alternative solution to avoid this problem is explained next.
The proposed solution to avoid injecting reactive currents in the non-faulty phases
is based on changing the current references depending on the amount of voltage drop,
keeping the reference(s) of the phase(s) with no voltage drop unchanged. For example,
if phase a is non-faulty under an unbalanced voltage sag, ka will be set to zero and
the zero-sequence is eliminated by changing the current references of the other phases,
i.e., kb + kc = 1. In this study, the zero-sequence rejection is divided equally between
the faulty phases, i.e., kb = kc = 1/2 in this example. However, it can be distributed
according to the severity of the voltage sag in these two phases. This method also
requires the implementation of over-current protection, as it is discussed in the next
subsection.
Once the zero-sequence component is removed from the current references, the ampli-
tudes of the currents change, which may produce overcurrents. In order to limit the
phase currents to the maximum value (In ), the method introduced in Section 5.6.1 is
applied. The final current references are set as:
0
iabc = frs iabc . (6.30)
where frs is defined in (5.20). The proposed method to re-scale the currents is illustrated
in Fig. 6.9.
It can be observed that the process of generating the phase current references includes
two limiters. The first one, shown in Fig. 6.7, is to limit the active currents to give
enough room to the required reactive current injection. The second one is based on
re-scaling all the current references after the zero-sequence elimination. This process is
proposed here for the first time and has never been addressed in any other studies.
The current control is composed of two parallel loops that regulate the currents in a
stationary frame. Since the control variables are sinusoidal, PR controllers are chosen
Chapter 6. Performance of GCPPPs under Voltage Sags with Ind. Phase Control 146
*'
iabc *
iabc
ia*' ia*rms f sr
ib*' ib*rms *
imax
ic*' ic*rms
Figure 6.9: Control diagram to re-scale the current references to avoid overcurrent.
i* eabc
* abc i* i v
iabc v vabc
*
vabc
abc
i* v
In order to illustrate the impact of injecting reactive power on the grid voltage magni-
tudes, a weak grid is required. In this study, a weak grid is emulated by adding inductors
(Lg ) in series with the grid generator. Since Lg is assumed to be a parasitic inductor
intrinsic to the grid, the grid voltages are measured after these inductors (indicated by
Eg in Fig. 6.11) and used for PLL synchronization. In the steady state, Eg can be
obtained as follows:
1
Eg = (kV + E), (6.31)
1+k
where k = Lg /L.
Based on (6.31), the lower the value of k, the lower amplitude the switching frequency
components in Eg . However, k should not be too small otherwise the injection of reactive
currents will have no significant effect on the voltage Eg . In this study, the selected value
of Lg to be used in the laboratory experimental setup is 1 mH.
Chapter 6. Performance of GCPPPs under Voltage Sags with Ind. Phase Control 147
V L Lg E
I Eg
V E
Figure 6.11: Generation of a weak grid in the laboratory setup.
V
Eg V
E
I E
As the flux of power is from the inverter to the grid, V is always greater than E. The
magnitude of Eg depends on the characteristic of the current injected into the grid as
depicted in Fig. 6.12. As Lg < L, the magnitude of the voltage drop in Lg is less than L,
i.e., |E| < |V |. If the current injected is purely resistive (=0), the magnitude of the
voltages can be sorted as: |Eg | < |E| < |V |. If the current injected is purely capacitive
(=90 ) or resistive-capacitive (0< <90 ) the voltage magnitudes can be sorted as
|E| < |Eg | < |V |. As a result, the magnitude of the voltage Eg can be increased above
E when there is reactive current injection. This is the principle used to regulate the grid
voltages through the injection of reactive current.
The proposed control method is tested in the scaled-down laboratory GCPPP. The
scheme of the GCPPP is presented in Fig. 2.1, where the grid filter impedance is assumed
to be zero (R=0). The main specifications of the experimental setup are summarized in
Table. 5.1.
bCI!lsldi\'
\'1 8.547 .,
\"l. S./31 'I
eabc
iabc
iq ref
r - ---------- \
I
\.
CH 1.2,3: 10 V/div CH 4,5,6: 20 Ndiv CH 7: 20 Ndiv
Figure 6.13: Performance of the balanced control method for the GCPPP under a
100% LG voltage sag at the grid side of the transformer. From top to bottom: grid
voltages at the LV side of the transformer, output currents at the LV side, and reactive
current reference.
a imposed at the grid side of the transformer. The voltage magnitudes are scaled-
down by 20 times to be able to show them on the oscilloscope. As demonstrated, the
injection of balanced reactive currents under unbalanced voltage sags leads to voltage
rise in the non-faulty phase. In this case, the amplitudes of the grid voltages rise from
8.547 20 = 170.94V to 8.731 20 = 174.63V , which means an increase of 1.84% of
the grid voltages during the voltage sag. Although this value does not reach the limit
defined by the GCs, this test shows the effect of rising the voltage of the non-faulty
phases above the nominal value.
f'N!I \f \f !l~r ~r [\
~Arrurm " ,r~~~~ tf [ (\ f\I IN If\ I~ "
-
. ~
\I
mm!f/~
11{\!fYA
mfffli\ ~ V I~ . . ~
-; lv I I I I 1/ I I I I 1/ V
II I lv V Vv I I I 1/
1/
b
1/ v
~I 1
-
I 1/I 1/ I II V1/1/ IV1/IVVI v; I I I lj I I 1/
c
~ ~ %~ ~ ~ ~ ~ ir abc
v
~ ~ ~ ~ wJ J J
iabc
:r..
~~r 1/\ w. "" A " " A lh " " " " MIA,AA fl. M.A ,., M~" .~"r~~{l. "'~
"' ,rr.. 11 h r~ t "' r~ rA ~ M I
I
I
'
I i: I ;f: I \ ! I\ I /\ V I \ I I 11 III .I I \
,
vVVv vvVVv V
VVWV fJ IV~v V fJ Jv ,v rJJ rv
V
fV '~ VV 'V
V vVVvvvV V 'Vvv 1V ~ vv vv
V V V IV V V
6.4.8 Conclusions
In this chapter, a new control method based on the individual control of the three phases
of a GCPPP has been proposed. This independent control of the reactive currents
injected into the grid prevents the non-faulty phases from overvoltage. A new PLL
detector has been proposed with the capability to extract the angle of the three phases
independently. The angles detected of each phase are used for the independent control of
the grid currents. Furthermore, a novel frequency detector is proposed to make the angle
detector frequency adaptive. The reactive currents have been determined separately
based on the amount of voltage drop in each phase. The active current references
of each phase need to be limited based on the required amount of reactive currents.
Furthermore, in a three-phase system, it is necessary to cancel the zero-sequence from
the current references generated. This chapter has proposed two solutions to remove
the zero-sequence component. Also, a method to re-scale the instantaneous current
references to avoid producing overvoltage in the non-faulty phases while preventing the
GCPPP from over-currents is designed. The proposed control method has been tested
experimentally on a scaled-down laboratory prototype operating with a weak grid.
Next chapter will summarize the main achievements and conclusions of the thesis
and also the potential research that can be taken into consideration in future works.
Chapter 7
In this thesis, different methods of supporting the grid voltages in GCPPPs have been
proposed. Under dynamic conditions, the proposed methods have been developed to
maintain the connectivity of the GCPPPs under voltage sag conditions and support the
grid with reactive power injection as required by the GCs. This is known as fault-ride-
through capability requirement and it should prevent the GCPPPs from disconnection.
The main reasons for disconnection from the grid are:
excessive ac currents,
The proposed methods have been sorted based on three control strategies:
positive-sequence control,
All these methods have been tested by simulations and most of them have been
verified through experiments showing excellent results. The main conclusions of this
thesis are summarized in the next section.
151
Chapter 7. Conclusions and Future Work 152
7.1 Conclusions
One of the requirement of the new GCs for GCPPPs is to support the grid voltages with
reactive power injection under static conditions. In Chapter 3, a 10-MVA GCPPP based
on a 0.5-MVA model in DigSILENT software has been modified to address this require-
ment. Four this purpose four methods of reactive power support have been designed and
applied to the control of GCPPP. These methods are based on (i) fixed reactive power
injection, (ii) fixed power factor (cos ), (iii) cos dependent of active power (cos (P )),
and (iv) droop control Q(E). All these four methods have been developed further to
maintain the power factor cos value within acceptable limits based on the German
GCs.
In Chapter 4, maintaining GCPPPs connected to the grid has been addressed for
both single-stage (dc-ac) and two-stage (dc-dc and dc-ac) conversions by using controllers
that regulate the positive-sequence of the currents only. For the simulations, an improved
version of the GCPPP model in Chapter 3 has been developed using MATLAB/Simulink
and PLECS software. It has been demonstrated that the conventional SRF-PLL does
not perform well under unbalanced voltage sag conditions. As result, a more elaborated
PLL has been used based on the study in [6] which uses MAF technique to extract
the angle of the positive-sequence voltage. A current limiter has been designed to limit
the active current reference based on the maximum acceptable inverter current and the
required reactive current reference. Under unbalanced voltage sag conditions when the
active current reference is not limited by the current limiter, the reference contains a
second-order harmonic that leads the output current to be distorted. As a consequence,
a MAF has been included in the dc-link control loop to remove that harmonic. However,
applying such a filtering technique implies delays and may lead to instability in the dc-
link control loop. Hence, a stability analysis has been developed to obtain the best PI
parameters for the dc-link loop.
The main difference between single- and two-stage configurations regarding FRT
capability is the protection for excessive dc-link voltage. It has been demonstrated that
the single-stage conversion is self-protected while the two-stage conversion is not. As a
result, three different methods have been proposed to protect two-stage configurations
from dc-link overvoltage. These methods are:
controlling the PV array operating point to generate less power under voltage sag
conditions.
Chapter 7. Conclusions and Future Work 153
Finally, a method based on PI controllers has been proposed to regulate the balanced
reactive currents injected into the grid under voltage sags in order to avoid overvoltage
in the non-faulty phases. It should be remarked that although this method prevents
the non-faulty phases from overvoltage, the dynamic response of the controller maybe
relatively slow.
Another problem related to controlling only the positive-sequence of the grid currents
is the oscillating nature of the flux of power injected into the grid during unbalanced
voltage sags, which might be a problem for the stability of the grid. Furthermore, as
a result, an oscillation at a frequency twice the grid frequency appears on the dc-link
voltage. This oscillation may reach significant amplitudes under strong voltage sag
conditions. All these problems can be avoided by the control of not only the positive
sequence of the grid currents but also the negative one.
In Chapter 5, the GCPPP FRT capability has been addressed by controlling both,
the positive and negative sequences of the grid currents. By controlling both sequences,
the flux of power injected into the grid can be practically constant during unbalanced
voltage sags, even if the voltage imbalance is very strong. The main problem when using
two parallel loops for the control of the positive and negative sequences based on standard
PI controllers is that some filters need to be used to avoid distorting the grid currents,
which introduce delays and may compromise the GCPPP stability. A stability analysis
has been performed to obtain the proper parameters for the PI controllers for both
the current and voltage loops. However, the solutions obtained for the PI parameters
have been based on making the dynamic of the overall GCPPP slow. An alternative
control method based on using resonant controllers has been proposed, which is capable
of controlling both sequences without any need for filtering. This solutions can achieve
the two benefits, i.e., injecting constant power into the grid or to the filter and grid
besides providing the GCPPP with fast dynamics. Two new current limiters have also
been designed and tested for this control configuration. A further development of this
controller has also been performed in this chapter that allows the GCPPP to operate
without a PLL for grid synchronization (PLL-less control).
a novel frequency-adaptive PLL has also been designed with the capability of extracting
the individual angles of the phase voltages. A module that determines the frequency
of the grid has been developed that makes the proposed single-phase PLLs frequency
adaptive.
From the research work reported in this thesis, some future work can be devised in the
following areas:
Several power converters and different intersystem control methods and com-
pare their performances under static and dynamic conditions.
Combined performances by using different inverter sizes. For example for a
50-MVA GCPPP two options that can be compared are: (i) 50 PV inverters
of 1 MVA each, or (ii) 500 PV inverters of 100 kVA each.
Several dc-dc converters with independent MPPT control and how the tran-
sients affect on the dc voltage ripples and the overall performance of the
GCPPP.
Designing and evaluating GCPPPs with LCL filters: The grid filter used
for GCPPPs can be based on the LCL configuration. These filters can be designed
and compared with the conventional L filter used in this thesis.
Chapter 7. Conclusions and Future Work 155
[1] European Photovoltaic Industry Association (EPIA), Global market outlook for
photovoltaics 2014-2018. [Online] accessed Aug. 2014, http://www.epia.org.
[3] K. Kurokawa, K. Komoto, P. Vleuten, and D. Faiman, Energy from the Desert:
Practical Proposals for Very Large Scale Photovoltaic Systems. Earthscan, 1st ed.,
2007.
[5] E. Troester, New german grid codes for connecting PV systems to the medium
voltage power grid, in Proc. 2nd International Workshop on Concentrating Pho-
tovoltaic Power Plants: Optical Design, Production, Grid Connection, Mar. 2009.
156
Bibliography 157
[9] W. Xiao, H. Zeineldin, and P. Zhang, Statistic and parallel testing procedure
for evaluating maximum power point tracking algorithms of photovoltaic power
systems, IEEE J. Photovolt., vol. 3, pp. 10621069, Jul. 2013.
[12] F.-S. Pai, R.-M. Chao, S. H. Ko, and T.-S. Lee, Performance evaluation of
parabolic prediction to maximum power point tracking for PV array, IEEE Trans.
Sustain. Energy, vol. 2, pp. 6068, Jan. 2011.
[15] H. Zhou, W. Sun, D. Liu, J. Zhao, and N. Yang, The research of daily total
solar-radiation and prediction method of photovoltaic generation based on wavelet-
neural network, in Proc. Asia-Pacific Power and Energy Engineering Conference
(APPEEC), pp. 15, Mar. 2011.
[17] Y. Xue, M. Manjrekar, C. Lin, M. Tamayo, and J. Jiang, Voltage stability and
sensitivity analysis of grid-connected photovoltaic systems, in Proc. IEEE Power
and Energy Society General Meeting, pp. 17, Jul. 2011.
[18] S. Kabir, M. Nadarajah, and R. Bansal, Impact of large scale photovoltaic system
on static voltage stability in sub-transmission network, in Proc. IEEE ECCE Asia
Downunder (ECCE Asia), pp. 468473, Jun. 2013.
Bibliography 158
[21] M. El Moursi, W. Xiao, and J. Kirtley, Fault ride through capability for grid
interfacing large scale PV power plants, IET Generation, Transmission and Dis-
tribution, vol. 7, pp. 10271036, Sept 2013.
[25] X. Zhang and L. Ming, Using the fault current limiter with spark gap to reduce
short-circuit currents, IEEE Trans. Power Del., vol. 23, pp. 506507, Jan. 2008.
[30] S. Dasgupta, S. Mohan, S. Sahoo, and S. Panda, A lyapunov function based cur-
rent controller to control active and reactive power flow in a three phase grid con-
nected PV inverter under generalized grid voltage conditions, in Proc. IEEE 8th
International Conference on Power Electronics and ECCE Asia (ICPE ECCE),
pp. 11101117, May 2011.
[36] C. Yaai, L. Jingdong, Z. Jinghua, and L. Jin, Research on the control strategy of
pv grid-connected inverter upon grid fault, in Proc. International Conference on
Electrical Machines and Systems (ICEMS), pp. 21632167, Oct. 2013.
[37] G. Islam, A. Al-Durra, S. M. Muyeen, and J. Tamura, Low voltage ride through
capability enhancement of grid connected large scale photovoltaic system, in
Proc. 37th Annual Conference on IEEE Industrial Electronics Society (IECON),
pp. 884889, Nov. 2011.
[38] V. Kaura and V. Blasko, Operation of a phase locked loop system under distorted
utility conditions, IEEE Trans. Ind. Appl., vol. 33, pp. 5863, Jan. 1997.
Bibliography 160
[39] L. Arruda, S. Silva, and B. J. C. Filho, PLL structures for utility connected
systems, in Proc. IEEE Industry Applications Conference, vol. 4, pp. 26552660
vol.4, Sep. 2001.
[40] Q. Nasir and S. Al-araji, Enhanced performance zero crossing DPLL with lin-
earized phase detector, in Proc. International Symposium on Telecommunications
(IST), pp. 7376, Aug. 2008.
[41] H. Geng, D. Xu, and B. Wu, A novel hardware-based all-digital phase-locked loop
applied to grid-connected power converters, IEEE Trans. Ind. Electron., vol. 58,
pp. 17371745, May 2011.
[47] Y. F. Wang and Y. W. Li, Grid synchronization PLL based on cascaded delayed
signal cancellation, IEEE Trans. Power Electron., vol. 26, pp. 19871997, Jul.
2011.
[49] K.-J. Lee, J.-P. Lee, D. Shin, D.-W. Yoo, and H.-J. Kim, A novel grid synchro-
nization PLL method based on adaptive low-pass notch filter for grid-connected
PCS, IEEE Trans. Ind. Electron., vol. 61, pp. 292301, Jan. 2014.
[54] L. Xiao, S. Huang, and K. Lu, Dc-bus voltage control of grid-connected voltage
source converter by using space vector modulated direct power control under un-
balanced network conditions, IET Power Electronics, vol. 6, pp. 925934, May
2013.
[55] F. He, Z. Zhao, T. Lu, and L. Yuan, Predictive dc voltage control for three-
phase grid-connected pv inverters based on energy balance modeling, in Proc.
IEEE International Symposium on Power Electronics for Distributed Generation
Systems (PEDG), pp. 516519, Jun. 2010.
[56] L. Xialin, G. Li, W. Chengshan, L. Zhichao, and L. Shu, Coordinated control re-
search of grid-connected two-stage power converter system, in Proc. Asia-Pacific
Power and Energy Engineering Conference (APPEEC), pp. 15, Mar. 2012.
[57] J. Yuan, F. Gao, and H. Gao, Dc voltage sensorless control strategy of grid-
tied two-stage three-phase photovoltaic system, in Proc. IEEE 8th International
Conference on Power Electronics and ECCE Asia (ICPE ECCE), pp. 19081912,
May 2011.
Bibliography 162
[58] J. Dannehl, F. Fuchs, and P. Thogersen, PI state space current control of grid-
connected PWM converters with LCL filters, IEEE Trans. Power Electron.,
vol. 25, pp. 23202330, Sep. 2010.
[60] S. Liu, T. Bi, A. Xue, and Q. Yang, An optimal method for designing the con-
trollers used in grid-connected pv systems, in IEEE International Conference on
Power System Technology (POWERCON), pp. 16, Oct. 2012.
[61] M. Beza and M. Bongiorno, Improved discrete current controller for grid-
connected voltage source converters in distorted grids, in Proc. IEEE Energy
Conversion Congress and Exposition (ECCE), pp. 7784, Sep. 2012.
[64] G. Shen, X. Zhu, M. Chen, and D. Xu, A new current feedback PR control
strategy for grid-connected VSI with an LCL filter, in Proc. IEEE Applied Power
Electronics Conference and Exposition (APEC), pp. 15641569, Feb. 2009.
[66] J. Hu, Y. He, L. Xu, and D. Zhi, Predictive current control of grid-connected
voltage source converters during, IET Power Electronics, vol. 3, pp. 690701,
Sep. 2010.
[68] J. Massing and H. Pinheiro, Adaptive current control of grid-connected VSC with
LCL-filters using parallel feedforward compensation, in Proc. Annual Conference
on IEEE Industrial Electronics Society (IECON), pp. 31853191, Nov. 2010.
Bibliography 163
[71] D. sha, D. Wu, and X. Liao, Analysis of a hybrid controlled three-phase grid-
connected inverter with harmonics compensation in synchronous reference frame,
IET Power Electronics, vol. 4, pp. 743751, Aug. 2011.
[72] N. Thao and K. Uchida, Control the active and reactive powers of three-phase
grid-connected photovoltaic inverters using feedback linearization and fuzzy logic,
in Proc. 3rd Australian Control Conference (AUCC), pp. 133140, Nov. 2013.
[73] Y. Quan, H. Nian, J. Hu, and J. Li, Improved control of the grid-connected con-
verter under the harmonically distorted grid voltage conditions, in Proc. Inter-
national Conference on Electrical Machines and Systems (ICEMS), pp. 204209,
Oct. 2010.
[74] Q.-N. Trinh and H.-H. Lee, Improvement of current performance for grid con-
nected converter under distorted grid condition, in IET Conference on Renewable
Power Generation (RPG 2011), pp. 16, Sep. 2011.
[78] L. Wang, N. Ertugrul, and M. Kolhe, Evaluation of dead beat current controllers
for grid connected converters, in Proc. IEEE Innovative Smart Grid Technologies
- Asia (ISGT Asia), pp. 17, May 2012.
Bibliography 164
[83] Bloomberg New Energy Finance, H1 2014 levelized cost of electricity - PV.
[Online] accessed Aug. 2014, https://www.iea.org/media/workshops/2014/
solarelectricity/BNEF2LCOEofPV.pdf.
[86] World Energy Council, World energy perspective- cost of energy technologies.
[Online] accessed Aug. 2014, http://www.worldenergy.org/publications/
2013/world-energy-perspective-cost-of-energy-technologies/.
[90] R. Varieras, J. Wang, and D. King, System performance considerations for low-
concentration linear-focus silicon-based photovoltaic modules, IEEE J. Photo-
volt., vol. 3, pp. 14091414, Oct. 2013.
[92] National Center for Photovoltaics (NREL), Research cell efficiency records.
[Online] accessed Aug. 2014, http://www.nrel.gov/ncpv/images/efficiency_
chart.jpg.
[93] Y. Zhang, G. Adam, T. Lim, S. Finney, and B. Williams, Voltage source converter
in high voltage applications: Multilevel versus two-level converters, in Proc. 9th
IET International Conference on AC and DC Power Transmission (ACDC), pp. 1
5, Oct. 2010.
[95] J. Rodriguez, J.-S. Lai, and F. Z. Peng, Multilevel inverters: a survey of topolo-
gies, controls, and applications, IEEE Trans. Ind. Electron., vol. 49, pp. 724738,
Aug. 2002.
[97] D. Holmes and T. Lipo, Pulse Width Modulation for Power Converters:Principles
and Practice. Wiley-IEEE Press, 1st ed., 2003.
[99] W. Li, X. Lv, Y. Deng, J. Liu, and X. He, A review of non-isolated high step-
up DC/DC converters in renewable energy applications, in Proc. Twenty-Fourth
Annual IEEE Applied Power Electronics Conference and Exposition (APEC),
pp. 364369, Feb. 2009.
[101] H. Choi, Dc/dc converters for grid integration of large-scale solar photovoltaic
systems, Masters thesis, Electrical Engineering and Telecommunications, Uni-
versity of New South Walse, 2012.
[102] Deciding what type of system to install. [Online] accessed Aug. 2014, http:
//www.builditsolar.com/Projects/PV/EnphasePV/TypeOfSystem.htm.
Bibliography 166
[103] German Solar Energy Society (DGS), Planning And Installing Photovoltaic
Systems- A Guide For Installers, Architects And Engineers. Routledge Publi-
cations Ltd, New York, 3 ed., 2013.
[111] J.-H. Lee, H. Bae, and B.-H. Cho, Advanced incremental conductance MPPT
algorithm with a variable step size, in Proc. 12th International Power Electronics
and Motion Control Conference EPE-PEMC, pp. 603607, Aug. 2006.
[113] Z. Yan, L. Fei, Y. Jinjun, and D. Shanxu, Study on realizing MPPT by improved
incremental conductance method with variable step-size, in Proc. 3rd IEEE Con-
ference on Industrial Electronics and Applications ICIEA, pp. 547550, Jun. 2008.
Bibliography 167
[119] SUNTECH POWER, 280 Watt RELIATHON solar module. [Online] accessed
May 2013, http://am.suntech-power.com/images/stories/pdf/datasheets/
2011/STP275_VRM_UL.pdf.
[120] SMA Solar Technology, Sunny CENTRAL 1000 kVA. [Online] ac-
cessed May 2013, http://www.sma.de/en/products/central-inverters/
sunny-central-800mv-11-1000mv-11-1250mv-11.html.
[124] Y. Bae, T.-K. Vu, and R.-Y. Kim, Implemental control strategy for grid stabi-
lization of grid-connected PV system based on german grid code in symmetrical
low-to-medium voltage network, IEEE Trans. Energy Convers., vol. 28, pp. 619
631, Sep. 2013.
[126] H. Tian, F. Gao, and C. Ma, Novel low voltage ride through strategy of single-
stage grid-tied photovoltaic inverter with supercapacitor coupled, in Proc. 7th
International Power Electronics and Motion Control Conference (IPEMC), vol. 2,
pp. 11881192, Jun. 2012.
[127] X. Wang, M. Yue, and E. Muljadi, Modeling and control system design for an
integrated solar generation and energy storage system with a ride-through ca-
pability, in Proc. IEEE Energy Conversion Congress and Exposition (ECCE),
pp. 37273734, Sep. 2012.
[131] X. Bao, P. Tan, F. Zhuo, and X. Yue, Low voltage ride through control strategy
for high-power grid-connected photovoltaic inverter, in Proc. Twenty-Eighth An-
nual IEEE Applied Power Electronics Conference and Exposition (APEC), pp. 97
100, Mar. 2013.
[132] I. Andrade .A, R. Blasco-Gimenez, and R. S. Pena, Constant power control of grid
connected inverters during unbalanced faults, in Proc. 15th European Conference
on Power Electronics and Applications (EPE), pp. 17, Sep. 2013.
[134] P. R.-Sanchez, X. del Toro Garcia, A. Torres, and V. Feliu, Fundamental positive-
and negative-sequence estimator for grid synchronization under highly disturbed
operating conditions, IEEE Trans. Power Electron., vol. 28, pp. 37333746, Aug.
2013.
[144] Z. Li, Y. Li, P. Wang, H. Zhu, C. Liu, and W. Xu, Control of three-phase boost-
type PWM rectifier in stationary frame under unbalanced input voltage, IEEE
Trans. Power Electron., vol. 25, pp. 25212530, Oct. 2010.
Bibliography 170
[147] Q. N. Trinh and H. H. Lee, An enhanced grid current compensator for grid-
connected distributed generation under nonlinear loads and grid voltage distor-
tions, IEEE Trans. Ind. Electron., Early access, doi:10.1109/TIE.2014.2320218.
[149] I. Gabe and H. Pinheiro, Impact of unbalance voltage dips on the behavior of volt-
age source inverters, in Proc. Brazilian Power Electronics Conference (COBEP),
pp. 956963, Sep.-Oct. 2009.
[156] L. Wang, Q. Jiang, L. Hong, C. Zhang, and Y. Wei, A novel phase-locked loop
based on frequency detector and initial phase angle detector, IEEE Trans. Power
Electron., vol. 28, pp. 45384549, Oct. 2013.