Documentos de Académico
Documentos de Profesional
Documentos de Cultura
com
ScienceDirect
Solar Energy 97 (2013) 155167
www.elsevier.com/locate/solener
Received 4 December 2012; received in revised form 31 July 2013; accepted 12 August 2013
Available online 12 September 2013
Abstract
This paper presented a low cost and eective passive islanding detection method for single-phase photovoltaic grid-connected invert-
ers. An analog circuit for over/under voltage protection is developed to ensure fast detection and no delay to system. An under/over
frequency circuit is also developed, for accurate and fast frequency detection with minimal external components. A new algorithm is
developed in a low-cost PIC18F4550. An improved disconnection time in the proposed method compared with that in the previously
developed method is an attractive solution for single phase grid connected inverters. The low cost, eective and minimal external com-
ponent count are the advantages. A prototype is developed and tested to demonstrate the performance and feasibility of the proposed
method. The experiment results veried that the proposed islanding detection method able to detect islanding operation eectively under
various load types, inverter output powers and quality factors.
2013 Elsevier Ltd. All rights reserved.
1. Introduction a portion of utility system that contains both load and dis-
tributed resources remains energized while it is isolated
Development for renewable energy sources produce low from the remainder of the utility system. Such an undesir-
pollution compared to the fossil fuels and nuclear genera- able event could potentially occur due to the circuit
tion system (Yu et al., 2008). The new paradigm of distrib- tripping, accidental disconnection of the utility due to
uted generation (DG) thus gains technical importance and equipment failure, human error, temporary disconnection
creates business opportunities (Chowdhury et al., 2009). In (for maintenance services) or uncommon network recong-
principle distribution generation is a small scale generation uration (Chowdhury et al., 2009; Yu et al., 2010). Integrat-
unit that need to be installed to the load and also connected ing DG into utility is a major challenge to researchers. DG
to the grid for selling or buying energy purposes. One of could still be supplying local load demand while network is
the most important criteria that need to be considered is already isolated from the main system. Existing methods
the islanding issue (Yu et al., 2010). The islanding are still lacking, hence successful detection of islanding is
condition as specied in Recommended Practice for Utility an ongoing challenge.
Interface of Photovoltaic (PV) Systems (2000), occurs when Two factors must be highlighted to better understand
the islanding phenomenon. The rst one is the available
Corresponding author. Tel.: +60 3 22463246; fax: +60 3 22463257. standards that have been established for the grid connected
E-mail address: edhura@gmail.com (K.N.E.K. Ahmad). system. These standards addressed the issue of islanding as
0038-092X/$ - see front matter 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.solener.2013.08.011
156 K.N.E.K. Ahmad et al. / Solar Energy 97 (2013) 155167
Nomenclature
well as procedure for testing and qualifying DG system Active techniques developed include impedance measure-
(Chiang et al., 2010). The second feature is associated with ment (IM), sliding mode frequency shift (SMS) or active
the so called Non-detection zone (NDZ) which can be phase shift (APS), sandia frequency shift (SFS) or active
dened as the zone for which an islanding detection frequency drift with positive feedback, and reactive power
method would fail to operate in time. NDZ is an evaluating export error detection (RPEED) Chowdhury et al., 2009;
criterion of islanding detection methods. In principle, islan- Mohamad et al., 2011; Lopes and Huili, 2006. AFD
ding detection monitors changes in inverter output param- method varies the output current frequency through posi-
eter or other system parameters that indicate islanding. tive feedback. This method will inject a current with
There are two types of anti islanding methods which are slightly distorted in frequency into the PCC. Upon grid dis-
the local and remote methods. The local methods can be connection, the phase error between the PCC voltage and
divided into passive and active. Passive islanding detection the inverter current will be detected by the inverter, which
detects changes in electrical parameters to determine the then tries to compensate by increasing the frequency of the
occurrence of islanding (Chiang et al., 2010). The advanta- injected current until it exceeds the OF/UF limits. How-
ges of these passive techniques are their easy implementa- ever, the performance of this conventional AFD method
tion (no additional controller), no degradation of PV is inecient and researchers are facing diculty in choos-
inverter power quality, and their inexpensiveness. The pri- ing the suitable chopping fraction (cf) value to meet the
mary drawbacks are a relatively large NDZ and ineective- limit of harmonics. Hence, a novel AFD method (Jung
ness in multi-PV inverter systems (Syamsuddin et al., and Yu, 2007) with a periodic chopping fraction that devi-
2009). The most commonly used passive technique for ates from the frequency in an instant way from nominal
islanding detection are under/over voltage and under/over was proposed. Remote islanding detection techniques are
frequency (OV/UV & OF/UF), phase jump detection based on communication between utilities and PV inverter
(PJD), voltage harmonic monitoring, current harmonic units (Syamsuddin et al., 2009). This technique do not have
monitoring, rate of change of power output (ROCOP) the NDZ, do not degrade PV inverter power quality, are
and rate of change of frequency (ROCOF) Yu et al., eective in multi-PV-inverter systems, but are expensive
2010; Llaria et al., 2011; De Mango et al., 2006; Freitas to implement (especially in small systems), and have a com-
et al., 2005; Redfern et al., 1993. Active techniques inject plicated communication technique for multi-PV inverter
a small disturbance at the PV inverter output to detect systems. Common techniques that are communication-
islanding. The main advantage of these techniques is their based include power line carrier communication (PLCC),
relatively smaller NDZ than that of passive methods. Their signal produced by disconnect (SPD), and supervisory con-
main drawbacks are the possibly deteriorated output trol and data acquisition (SCADA). In application, each
power quality causing instability to the PV inverter and method has advantages and drawbacks. Among popular
normally require additional controllers which increased reference standards for islanding are IEEE 929-2000, IEC
the complexity of the method (Syamsuddin et al., 2009). 62116, IEE 1547, VDE 0126-1-1, and AS 4777.3-2005.
K.N.E.K. Ahmad et al. / Solar Energy 97 (2013) 155167 157
Table 1
Existing anti-islanding standards.
Quality factor, Qf Required islanding Normal frequency range, Normal voltage range,
detection time, t f (nominal frequency f0) V (% of nominal voltage V0)
IEC 62116 1 t<2s f 0 1:5 Hz 6 f and 85% 6 V 6 115%
f 6 f 0 1:5 Hz
IEEE 1547 1 t<2s 59:3 Hz 6 f 6 60:5 Hz 88% 6 V 6 110%
IEEE 929-2000 2.5 t<2s 59:3 Hz 6 f 6 60:5 Hz 88% 6 V 6 110%
Japanese standard 0 (+rotating Passive: t < 0.5 s active: Setting value Setting value
machinery) 0:5 s < t < 1 s
Korean standard 1 t < 0.5 s 59:3 Hz 6 f 6 60:5 Hz 88% 6 V 6 110%
VDE 0126-1-1 2 t < 0.2 s 47:5 Hz 6 f 6 50:2 Hz 80% 6 V 6 115%
AS4777.3-2005 1 t<2s Setting value Setting value
Table 1 shows all the standards having their own quality from the variable ac supply, and (2) reading of the
factor (Q) value, the required islanding disconnection time, over/under voltage or over/under frequency values at
and range of operational frequency and voltage (Yu et al., which disconnection occurred. A criteria of acceptance
2010). According to IEEE929-2000 standard the quality for over/under voltage is that the value at step (2) should
factor Q is equals to; equal an under-voltage set point of 5 V; for frequency,
the value should equal an under-frequency set point of
Q tanarccosinepf 1 0.1 Hz.
The primary objective of this paper is to propose a new
The selected Q = 2.5 is equals to a power factor of 0.37.
low cost and eective passive method for islanding detec-
As the power factor increases, Q decreases. Thus according
tion of single-phase grid-connected PV inverter and to
to IEEE929-2000, the test requirement that sets Q = 2.5
compare between the proposed algorithm and existing
equals lines with uncorrected power factors from 0.37 to
algorithm.
unity and seems to cover all reasonable distribution line
congurations (Recommended Practice for Utility Inter-
face of Photovoltaic (PV) Systems, 2000). Meanwhile, Jap- 2. Research methodology
anese standard are proposing Q = 0 based on adding
rotating machinery while islanding test is being conducted 2.1. PV inverter system connected to grid
(Yu et al., 2010). Other than quality factor, disconnection
time is also important. The German VDE0126-1-1 stan- Fig. 1 is the block diagram of a typical single-phase grid-
dard has the strictest disconnection time limit which is less connected PV inverter; here, it has PV arrays, DCDC
than 0.2 s. Normal frequency and voltage range also are boost converter connected to two capacitors in series
crucial to determining islanding detection capability. (and functioning as maximum power point (MPP) tracker),
Australian standard AS4777.3-2005 requires nominal four switching devices connected in full bridge congura-
frequency and voltage range to be set by manufacturer. tion (S1S4) and presented as ideal switches, and a lter
Steps to obtaining tripping values are: (1) determination inductor (Lf) ltering the current injected into the grid.
of under/over voltage and under/over frequency values The power produced by the system is transferred directly
through gradual increase or decrease of voltage and to the grid. Since grid voltage is uncontrollable, the sim-
frequency until the device tested (the inverter) disconnects plest way of controlling system operation is by controlling
15V
R3
R1
LED
+
C1 R2 To Port
Grid RD6
- R1
Op
R1 Amp1
2W06 LM311
15V
R3
R1 LED
+
To Port
RD7
- R1
Op
Amp2
LM311
the current owing into the grid. The injected current must
be sinusoidal and have low harmonic distortion (Rahim
et al., 2011; Rahim and Selvaraj, 2010).
Table 2
Parameters of the single-phase grid-connected PV inverter prototype.
Parameter Value
D1 RHRP30120 VRR = 1200 V, I = 30 A
S1S4 IGBT IRG4PH50UD VCE = 1200, Ic = 24 A
Filter inductor (Lf) 5 mH
Switching frequency 20 kHz
Table 3 Table 4
Load parameters of the simulation and experiment at Q = 2.5, and Load parameters of the simulation and experiment at Q = 1, 2, 2.5, and
P = 1 kW. P = 1 kW.
Types R (O) L C Quality factor R (O) L (mH) C (lF)
Pure resistor load 57.6 1 57.6 183.35 55.26
Resistorinductor load 57.6 73.33 mH 2 57.6 91.67 110.52
Resistorcapacitor load 57.6 138.13 lF 2.5 57.6 73.33 138.13
Resistorinductor-capacitor load 57.6 73.33 mH 138.13 lF
Voltage sensor
PM6000
Current sensor
LAN Area
LeCroy
Oscilloscope
R C L Inverter PV Array
Simulator
Programmable
AC source
GPIB and serial USB
(a)
Computer
PM6000
Oscilloscope
under and over
voltage circuit
Inverter
RLC load
(b)
Fig. 7. (a) Islanding test and (b) the laboratory test equipment with the proposed passive islanding detection method.
K.N.E.K. Ahmad et al. / Solar Energy 97 (2013) 155167 161
standard reconnection time. For frequency calculation, a Fig. 7 shows the PV grid-connected inverter hardware
simple zero-crossing circuit is utilized. The square wave prototype controller interface with the islanding detection,
produced by the zero-crossing detector has the same fre- implemented on PIC 18F4550 to validate the performance
quency as the grid. By feeding the produced signal into of the proposed passive islanding detection algorithm.
the RB0 of the PIC18F4550, the microcontroller is pro- Table 2 lists the prototype parameters of the single phase
grammed to be a very accurate zero-crossing detector with grid-connected PV inverter. The hardware was tested under
an interrupt-driven code. various load conditions. Equipment such as PV array
Vgrid Vgrid
400 400
200 200
0 0
-200 -200
-400 -400
Vpcc Vpcc
400 400
200 200
0 0
-200 -200
-400 -400
Iinv*10 Iinv*10
100 100
50 50
0 0
-50 -50
-100 -100
0 0.2 0.4 0.6 0.8 1 1.2 0 0.2 0.4 0.6 0.8 1 1.2
Time (s) Time (s)
(a) (b)
Vgrid
400
200
0
-200
-400
Vpcc
400
200
0
-200
-400
Iinv*10
100
50
0
-50
-100
0 0.2 0.4 0.6 0.8 1 1.2
Time (s)
(c)
Fig. 9. Simulation result for islanding detection, at (a) Q = 1, (b) Q = 2, and (c) Q = 2.5.
162 K.N.E.K. Ahmad et al. / Solar Energy 97 (2013) 155167
2
simulator, resistor-inductor-capacitor (RLC) load, pro- Qf
grammable AC source, PM6000, and LeCroy Oscilloscope C L 4
R
were used. The test was conducted from 500 W to 2 kW
power output. In the following experiments, the grid-con- Fig. 8 is a diagram of the PSIM model of the islanding
nected inverter was controlled to generate balanced real circuit.
power with the load. Various load types were used to verify
the proposed islanding detection method. Table 3 lists the 3. Results and discussion
load parameters of the experiments. The parameters of
the resistorinductorcapacitor (RLC) load were as 3.1. Disconnection time for dierent method and algorithm
specied by the IEEE Std. 929-2000 (Recommended Prac-
tice for Utility Interface of Photovoltaic (PV) Systems, To prove that the proposed passive islanding detection
2000) for testing islanding operation performance. IEEE has advantages over the implemented algorithm in
Std. 929-2000 states that a non-islanding inverter must shut TMS320F2812 DSP in terms of disconnection times, the
down within 2 s of grid disconnection for loads with a qual- corresponding comparisons were made on both passive
ity factor (Qf) equals to or lower than 2.5. The values for L islanding methods. Low cost PIC 18F4550, compared with
and C are calculated through this formula: ATMEGA and TMS320F2812 DSP with under/over volt-
age and under/over frequency islanding detection algo-
V2 rithms, were compared. The circuit connections and
P 2 algorithms of the past methods are as in Syamsuddin
R
et al. (2009). The prototype of both methods monitored
R grid voltage and grid frequency at PCC (see Table 4).
L 3
Qf x
Vpcc
400
200
-200
-400
Iinv*10
100
50
-50
-100
(a)
Vpcc
400
200
-200
-400
Iinv*10
100
50
-50
-100
(b)
Fig. 10. Simulation result for (a) RL load (b) RC load under islanding operation with Q = 2.5 and P = 1 kW.
K.N.E.K. Ahmad et al. / Solar Energy 97 (2013) 155167 163
Table 5
Disconnection times of the two methods.
Under voltage Over voltage Under frequency Over frequency
disconnection time (s) disconnection time (s) disconnection time (s) disconnection time (s)
ATMEGA and DSP based Could not be determined 1.2 2.1 52.5
algorithm
PIC 18F4550 algorithm 0.6 0.6 0.6 0.6
Fig. 11. Experiment results for R load under islanding operation, for (a) 25% (500 W), (b) 50% (1 kW), (c) 75% (1.5 kW), and (d) 100% (2 kW) inverter
output, with (M1) Vgrid, (M2) Vpcc, (M3) the inverter output current and (M4) load current.
The disconnection time of PIC18F4550 was much faster for under/over voltage disconnection. This may add some
than that of ATMEGA and TMS320F2812. The cost of lagging to the systems when detecting under/over voltage.
design and implementation for the proposed method were The proposed circuit and algorithm aim to correct the
also much less. The disconnection times were validated problem.
according to AS 4777.3-2005 standard. Table 5 compares
the disconnection times. 3.2. Simulation results
As the table shows, the proposed method is much faster
and more stable than the past method (very unstable). Simulations were performed using PSIM system. The
Although over-voltage disconnection time of the past DC voltage source was 400 V, the grid voltage 240 V, the
method passed the less-than-2-s standard, the proposed grid frequency 50 Hz, and the output lter inductance
method enabled much faster disconnection. From Syam- 5mH. The inverter was designed for 1.4 kW power output.
suddin et al. (2009), this method used TMS320F2812 The RMS current injected into the grid was 6 A. The main
DSP for reading the grid voltage at the sensor signal parameters of the simulation were as those of the experi-
through the ADC and calculates the RMS voltage value ments. The inverter was loaded by a parallel RLC circuit
164 K.N.E.K. Ahmad et al. / Solar Energy 97 (2013) 155167
with quality factors 1, 2, and 2.5. Fig. 9 shows the point of the simulation result for RC under islanding operation
common coupling voltage Vpcc, the inverter current and with 1 kW inverter output power. The Vpcc voltage is
voltage, and grid current and voltage ideally when the almost the same for both simulation and experiment results
inverter is tripped o. (see Fig. 10(b) and 12(b)). From the simulation result for
Fig. 10(a) shows the simulation result for RL under RC load under islanding condition, the value of Vpcc volt-
islanding operation with 1 kW inverter output power. age before inverter is tripped o is 160 V and 216 V from
The inverter output current is almost the same for both the experiment (see Fig. 12(b)).
simulation and experiment results (see Fig. 10(a) and see
Fig. 12(a)). From the simulation result for RL load under 3.3. Experiment results
islanding condition, the value of the maximum inverter
output current before inverter is tripped o is 7.08 A and The grid system was rst connected to the inverter and
5.7 A from the experiment (see Fig. 12(a)). Fig. 10(b) shows then removed to create islanding. Fig. 11 gives the dierent
Fig. 12. Experiment results for (a) RL load, and (b) RC load under islanding operation, with (M1) Vgrid, (M2) Vpcc, (M3) the inverter output current
and (M4) load current.
K.N.E.K. Ahmad et al. / Solar Energy 97 (2013) 155167 165
Fig. 13. Experiment results for RLC load under islanding operation, at (a) Q = 1, (b) Q = 2, and (c) Q = 2.5, with (M1) Vgrid, (M2) Vpcc, (M3) the
inverter output current and (M4) load current.
Fig. 14. Experiment results for RLC load under islanding operation, at Q = 2.5 with (M2) the inverter output current, (M3) load current.
166 K.N.E.K. Ahmad et al. / Solar Energy 97 (2013) 155167
Load power
load
power
grid
power
inverter
power
Grid power
seconds
Fig. 15. The inverter comes on-line and gradually increases its output to 1 kW.
Table 6 grid values. It has been proved that the larger the power
Detection times for the RLC load with Q = 2, as taken by PM6000 imbalance (DP and DQ in local generation and consump-
Universal Power Analyzer.
tion in the islanded system prior to grid disconnection),
Timestamp Channel Arms Watt the larger the variations are in the frequency of the voltage
20130130155137.30 B Load 5.00 1091.88 and voltage at the PCC. Therefore, standard under/over
20130130155137.30 C Grid 2.51 182.38 frequency and under/over voltage protections are success-
20130130155137.30 D Inverter 5.35 1283.13
20130130155137.60 B Load 0.00 0.00
ful in preventing islanding in system with large enough
20130130155137.60 C Grid 0.00 0.00 power imbalances. The islanding condition for the experi-
20130130155137.60 D Inverter 0.00 0.00 ments conducted was made to have the smallest power
imbalance possible, i.e., the worst-case scenario. Fig. 13
shows the experiment results for RLC load under islan-
Table 7
ding, at Q = 1, Q = 2, and Q = 2.5. In the following exper-
Detection times for the dierent quality factors.
iments, the grid-connected inverter will be tripped o
Capacity inverter Quality factor of the RLC Detection
21.25 ms (Q = 1), 24.8 ms (Q = 2) and 55.7 ms (Q = 2.5)
output Load time
after the grid is disconnected. From all the following exper-
1 kW 1 21.25 ms
iments (see Fig. 13), the inverter (Vpcc) briey swings up to
1 kW 2 24.8 ms
1 kW 2.5 55.7 ms slightly more than + and 400 volt peak after the grid is
disconnected, and then shuts down. The detection time of
islanding operation may be dierent under dierent loads
Table 8 and dierent quality factors of RLC load. As Fig. 14
Response to abnormal voltages.
shows, (M2) and (M3), at inverter output 1 kW and R
Voltage (at PCC) Maximum trip time LC load with Q = 2.5, the inverter output current was
V < 120 (V < 50%) 6 cycles almost identical with the load current. This indicates the
120 6 V < 211 (50% 6 V < 88%) 120 cycles worst case for islanding detection, where the power-ow
211 6 V 6 264 (88% 6 V 6 110%) Normal operation
between the grid-connected inverter and the local load is
264 < V < 329 (110% < V < 137%) 120 cycles
329 6 V (137% 6 V) 2 cycles balanced.
Table 6 lists the disconnection times for when the grid
was disconnected until the inverter was tripped o. As
shown in Table 6 the disconnection time is 0.3 s (24.8 ms)
inverter outputs: 25%, 50%, 75% and 100% with pure resis- for RLC load with Q = 2. PM6000 Universal Power
tive load under islanding condition. The results allow con- analyzer was set to the minimum time setting available
clusion that the ratio of the inverters output and load for data recording (every 0.3 s). The detection time satises
consumption can aect the amplitude of the load voltage. the IEEE 929-2000 control standard, verifying that the pro-
Fig. 12 shows the experiment results for the RL and R posed islanding method is able to detect islanding eec-
C loads under islanding operation. In the following exper- tively. As shown in Table 6, the inverter current value
iments, the grid-connected inverter will be tripped o when immediately zeros out after the grid is disconnected (see
islanding is detected. Table 7).
Under islanding conditions, the magnitude and fre- According to IEEE Std. 929-2000 standard, the relation-
quency of the voltage at PCC tend to slide from the rated ship between the amplitude of the load voltage and the
K.N.E.K. Ahmad et al. / Solar Energy 97 (2013) 155167 167