Está en la página 1de 7

228 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 21, NO.

1, MARCH 2006

Constant Voltage Constant Frequency Operation


for a Self-Excited Induction Generator
Dheeraj Joshi, Kanwarjit Singh Sandhu, and Mahender Kumar Soni

AbstractIn this paper, genetic algorithm-based technique is II. INTRODUCTION


proposed to estimate and analyze the steady-state performance of
N INCREASING rate of depletion of conventional sources
a self-excited induction generator (SEIG). The study reveals that
the performance of the SEIG is greatly influenced by the operating
speed, load, and excitation capacitance. This provides an opportu-
A of energy and growing power demand has diverted the
attention of scientists towards nonconventional sources of en-
nity for the proper handling of these parameters so as to obtain the ergy such as wind and solar energy. Induction generators have
required performance characteristics. In this paper, a new tech-
been found to be very suitable for wind energy conversion.
nique is proposed to identify these parameters to achieve constant
voltage constant frequency operation for SEIG. Theoretical re- These may be operated in grid-connected or self-excited mode.
sults as obtained have been compared with experimental results It is only the self-excited operation that can meet the load
and found to be in close agreement. requirement of an isolated remote area. However, the main
Index TermsGenetic algorithm (GA), self-excited induction hurdle in the application of the self-excited induction gener-
generator (SEIG), wind energy generation. ator (SEIG) is its inability to control the terminal voltage and
frequency.
In this paper, an attempt is made to control the terminal volt-
I. NOMENCLATURE
age and frequency of SEIG simultaneously. It is hoped that
a Per unit frequency. this will draw the attention of scientists towards its application,
b Per unit speed. especially for remote windy areas.
C Excitation capacitance per phase. To estimate and analyze the steady-state performance of the
E1 Air gap voltage per phase at rated frequency. SEIG, researchers have made use of the conventional equivalent
E2 Rotor emf per phase referred to stator. circuit of an induction motor. Some researchers [1][7] used the
Ea Air gap voltage per phase = aE1 . impedance model, and a few [8][11] used the admittance-based
f Rated frequency. model for the analysis. However, it is believed that the conven-
I1 Stator current per phase. tional equivalent circuit, in the absence of an active source, does
I2 Rotor current per phase, referred to stator. not effectively correspond to generator operation. Reference
Ib Base current. [10] suggested a new equivalent circuit model (see Appendix I)
IL Load current per phase. for the representation of the induction generator. In this paper, a
Im Magnetizing current per phase. new approach is adopted to analyze the steady-state performance
IL p u Per unit load current. of the SEIG. This has made the analysis and physical interpreta-
Pout Output power. tion of the performance much simpler, more explicit, and more
R Load resistance per phase. comprehensive in contrast to earlier works where fourth-order
R1 Stator resistance per phase. polynomial was an inevitable intermediate step. Availability of
R2 Rotor resistance per phase, referred to stator. the genetic algorithm (GA) [13][15], a new and efficient tech-
s Generator slip. nique for optimizing system behavior, provides an opportunity
V Terminal voltage per phase. to predict the behavior of the induction generator in a defined
Vb Base voltage. manner. This recent technique has been proposed to control the
Vdi Per unit error in terminal voltage. terminal voltage and the frequency of SEIG. Computed results
Vpu Per unit voltage. using GA have been compared with experimental results. The
Vref Reference voltage (= 1.0 per unit). closeness of these results confirms the validity of the proposed
X1 Stator reactance per phase. modeling.
X2 Sotor reactance per phase, referred to stator.
Xc Capacitive reactance due to C at rated frequency.
Xm Magnetizing reactance per phase at rated frequency. III. MODELING OF SEIG
The steady-state operation of the SEIG may be analyzed by
Manuscript received September 1, 2004; revised April 21, 2005. Paper no. using the equivalent circuit representation [12], as shown in
TEC-00257-2004.
D. Joshi and K. S. Sandhu are with the Electrical Engineering Department, Fig. 1.
National Institute of Technology, Kurukshetra, Haryana 136119 India (e-mail: Furthermore, this network may be modified to a more prac-
dheeraj joshi@sify.com; kjssandhu@yahoo.com). tical format, as shown by Fig. 2, wherein Ea (1 + s) represents
M. K. Soni is with C. R. State College of Engineering, Murthal, Sonipat
131027 India (e-mail: mksoni123@hotmail.com). source voltage corresponding to mechanical power transformed
Digital Object Identifier 10.1109/TEC.2005.858074 to electrical power through rotor.

0885-8969/$20.00 2006 IEEE


JOSHI et al.: CONSTANT VOLTAGE CONSTANT FREQUENCY OPERATION FOR A SELF-EXCITED INDUCTION GENERATOR 229

modeling (constant frequency model or CFM) may be used to


control the generated frequency of SEIG using speed as a con-
trol parameter. However, generated voltage may be controlled
by proper handling of excitation capacitance [3], [8].
Once air gap voltage E1 at rated frequency is known, then
the performance of the machine, using the equivalent circuit
representation as shown in Figs. 1 and 2, may be obtained as
follows:
Stator Current:
aE1
Fig. 1. Per phase equivalent circuit representation for SEIG. I1 =  2 2
. (6)
R1L + X1L
Terminal Voltage:

V = I1RL2 + XL2

aE1 RL2 + XL2
=  2 2
. (7)
R1L + X1L
Load Current:
V
IL = . (8)
R
Fig. 2. Modified per phase equivalent circuit representation for SEIG.
Output Power:
Pout = 3IL2 R. (9)
where
RXc2 IV. GENETIC ALGORITHM
RL =
a2 R2 + Xc2
Different from conventional optimization methods, the GA
aR2 Xc was developed based on the Darwinian evolution theory of
XL = 2 2 . (1)
a R + Xc2 survival of the fittest. It has produced good results in many
Analysis of the circuit [12] in Fig. 2 results in the following: practical problems and has become a powerful tool for solv-
ing nonlinear equations. The outstanding advantage of the GA
a2 s2 + a1 s + a0 = 0 (2) is that it does not need to differentiate the objective function
where and constraints of the problem or to require a good initial esti-
mate. The GA manipulates strings of binary digits and measures
a2 = a2 X22 R1L each strings strength with a fitness value. The main idea is that
 2  stronger strings advance and mate with other strong strings to
a1 = R2 R1L 2
+ X1L
produce offspring. Finally, one string emerges as the best. An-
a0 = R1L R22 other important advantage is that it offers parallel search, which
can overcome local optima and then finally find the globally
R1L = R1 + RL
optimal solution.
X1L = aX1 XL . (3) The mechanics of the GAs are elementary, involving noth-
ing more than copying strings, random number generation, and
  2   swapping partial strings. A common GA is mainly composed of
R2 R1L 2
+ X1L three operators: reproduction, crossover, and mutation. GA for
Xm = . (4)
sa2 X2 R1L + aR2 X1L this particular problem has the following components:
1) genetic representation for potential solutions to the
Equation (2) provides two possible values of slips, out of problem;
which only the lower one is relevant for the generating mode and 2) a way to create an initial population of potential solutions;
where the relation between generated frequency and operating 3) evaluation function that plays the role of the environmental
speed is rating solutions in terms of their fitness. This is because
the population undergoes a simulated evolution at each
b = a(1 + s). (5)
generation. This role of an environment helps relatively
Equations (2), (4), and (5), along with the magnetization curve good solutions to reproduce, while relatively bad so-
for the induction machine, may be used to determine the speed lutions die;
(required to generate a particular frequency) and terminal volt- 4) genetic operators then alter the composition of children.
age for given excitation capacitance and load. Therefore, such The multidirectional search is performed by maintaining
230 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 21, NO. 1, MARCH 2006

Fig. 3. Variation of excitation capacitance with load.

Fig. 4. Variation of generator speed with load.

a population of potential solutions and encourages infor- Now, GA processes the population of strings by applying
mation exchange between these directions; the genetic operators (crossover, mutation), causing multidirec-
5) values for various control parameters that the GA uses tional search. Thus, a new population is created called new
(e.g., population size, probabilities of applying GA). generation that has different chromosomes. Performance of
each chromosome is evaluated by using a performance index
(Vdi ):
V. GENETIC ALGORITHM-BASED MODELING OF SEIG
 
Operating speed and excitation capacitance have been real- Vref Vpu
Vdi = . (10)
ized as two control variables to handle the generated frequency Vref
and terminal voltage for SEIG. To use GA to solve the given
problem, the chromosomes, consisting of control variables Minimum value of this index then corresponds to the optimum
(i.e., C and b), are to be coded in string structures. Here, set of control variables, such as speed, excitation capacitance,
concatenated, multiparameter, mapped coding is used. Control and load resistance in SEIG. All control variables are optimized
variables are mapped to a specified interval (Amin , Amax ), according to their fitness value calculated by this performance
where Amin and Amax are the minimum and maximum values index. If the required goal is not achieved, the chromosome of
of the variables. This interval is also termed as bounds and the current generation will go through operations, crossover,
can be suitably assumed. To construct a multiparameter coding, and mutation. After a number of generations, the program
each coded variable has its own bounds as shown in tabular converges. GA-based optimization formulation is shown in
form (see Appendix II). the flowchart (see Appendix II).
JOSHI et al.: CONSTANT VOLTAGE CONSTANT FREQUENCY OPERATION FOR A SELF-EXCITED INDUCTION GENERATOR 231

TABLE I
COMPARISON OF COMPUTED AND EXPERIMENTAL RESULTS

TABLE II
COMPUTED RESULTS FOR CVCFM

Fig. 5. Constant voltage constant frequency model.

Application of GA, along with CFM, results in the various the operating speed of SEIG (i.e., the minimum cut-off speed
combinations of C, b, and R, with Vdi 1% and a = 1. The must be near and above the synchronous speed). Fig. 6 shows
results so obtained may be plotted as shown in Figs. 3 and 4. the variation of terminal voltage with load current for constant
Mathematical treatment of these curves using MATLAB frequency variable voltage operation. A fall in terminal voltage
yields the following expressions: with load strikes the need of voltage controller. In this paper, it
C = 0.5466IL2 p u + 4.4162IL p u + 20.9227 is the GA controller.

b = 0.0051IL3 p u 0.0016IL2 p u + 0.0783IL p u + 1.0038 measured value calculated value
%Error = 100.
measured value
(11)
where
Furthermore, Table II shows the computed results when the
IL constant voltage constant frequency model (CVCFM) is applied
IL p u = .
Ib on machine 1 with unity power factor load. CVCFM results in
Equation (11) may be used to preestimate the control param- constant voltage and constant frequency operation, with termi-
eters C and b for any load, which results in a constant voltage nal voltage as 0.995 pu and generated frequency as 1 pu. This is
constant frequency operation of SEIG, as shown in Fig. 5. possible only with simultaneous control of two variables (i.e.,
operating speed and excitation capacitance). It is observed that
as the load increases, there is a continuous increase in the exci-
VI. RESULTS AND DISCUSSIONS
tation capacitance and operating speed, to maintain the terminal
Table I shows the comparison of computed results on ma- voltage and frequency.
chine 1 (see Appendix III) using CFM and GA based constant Fig. 7 shows the variation of C and b with load to maintain
frequency model (GACFM) with the experimental one GACFM constant voltage and constant frequency for 0.8 pf load. These
has been used to estimate the values of C and b with a = 1 and results may be obtained by modifying the load branch in Fig. 1.
with minimum voltage error with respect to experimental volt- The closeness between experimental and computed values con-
age taken as reference value. It is observed that the computed firms the validity of CVCF modeling using GA. From Figs. 3
results are found to be in close agreement with experimental re- and 7, it is observed that the slope of C changes with pf. It is due
sults. This confirms the GA-based selection of parameters. It is to additional reactive power requirement of load. Fig. 8 shows
observed that to obtain rated frequency supply, operating speed the experimental variation of terminal voltage and frequency.
must be always greater than 1 pu. This will pose a limitation on The results are found to be very close to 1 pu.
232 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 21, NO. 1, MARCH 2006

Fig. 6. Variation of terminal voltage with load.

Fig. 7. Variation of excitation capacitance and speed with 0.8 pf lagging load.

Fig. 8. Variation of terminal voltage and generated frequency with 0.8 pf


lagging load.

VII. CONCLUSION
In this paper, an attempt is made to control the terminal volt-
age and frequency of the SEIG. The CVCFM is proposed to
achieve the constant voltage constant frequency operation. GA,
which is a modern approach and efficient technique, is used to
optimize the performance index, thus resulting in the preesti-
mation of control variables to achieve the required performance
characteristics. Closeness between computed and experimental Fig. 9. Equivalent circuit representation.
JOSHI et al.: CONSTANT VOLTAGE CONSTANT FREQUENCY OPERATION FOR A SELF-EXCITED INDUCTION GENERATOR 233

results proves the validity of GA modeling. It seems to be very APPENDIX III


effective and useful for making the SEIG feasible for remote
The details of the induction machine 1 used to obtain the
windy areas. experimental results are as follows:
Specifications: Three-phase, four-pole, 50-Hz, star con-
APPENDIX I nected, squirrel cage induction machine 750 W/1 HP, 380 V,
The usual equivalent circuit representation for a three-phase and 1.9 A.
induction motor is shown in Fig. 9(a). This circuit may be re- Parameters: The equivalent circuit parameters for the ma-
drawn as shown in Fig. 9(b) and (c). In Fig. 9(c), the sink voltage chine are
is given by E2 (1 s); E2 being equal to I2 (R2 /s + jX2 ). For
R1 = 9.5 , R2 = 8.04 , X1 = X2 = 8.84 .
generator operation, the sink voltage becomes source voltage
with slip as negative. The corresponding equivalent circuit is
shown in Fig. 9(d). Base Values:
Base voltage = 219.3 V
APPENDIX II
Base current = 1.9 A
Base impedance = 115.4
TABLE III Base frequency = 50 Hz
LOWER AND UPPER BOUNDS
Base speed = 1500 r/min.
Air Gap Voltage: The variation of air gap voltage with mag-
netizing reactance at rated frequency for the induction machine
is as given:
Xm < 169.2 E1 = 512.69 2.13Xm
179.42 > Xm 169.2 E1 = 891.66 4.37Xm
184.46 > Xm 179.42 E1 = 785.79 3.78Xm
Xm 184.46 E1 = 0.

REFERENCES
[1] S. S. Murthy, O. P. Malik, and A. K. Tandon, Analysis of self-excited
induction generators, Proc. Inst. Elect. Eng., C, vol. 129, no. 6, pp. 260
265, 1982.
[2] G. Raina and O. P. Malik, Wind energy conversion using a self-excited
induction generator, IEEE Trans. Power App. Syst., vol. PAS-102, no. 12,
pp. 39333936, 1983.
[3] A. K. Tandon, S. S. Murthy, and C. S. Jha, New method of computing
steady-state response of capacitor self-excited induction generator, IE(I)
J.-EL, vol. 65, pp. 196201, 1985.
[4] N. H. Malik and S. E. Haque, Steady-state analysis and performance of an
isolated self-excited induction generator, IEEE Trans. Energy Convers.,
vol. EC-1, no. 3, pp. 134139, 1986.
[5] N. H. Malik and A. H. Al-Bahrani, Influence of the terminal capacitor
on the performance characteristics of a self-excited induction generator,
Proc. Inst. Elect. Eng., C, vol. 137, no. 2, pp. 168173, 1990.
[6] L. Shridhar, B. Singh, and C. S. Jha, A step towards improvements in the
characteristics of self-excited induction generator, IEEE Trans. Energy
Convers., vol. 8, no. 1, pp. 4046, Mar. 1993.
[7] T. F. Chan, Steady-state analysis of self-excited induction generators,
IEEE Trans. Energy Convers., vol. 9, no. 2, pp. 288296, Jun. 1994.
[8] L. Quazene and G. McPherson, Analysis of the isolated induction gener-
ator, IEEE Trans. Power App. Syst., vol. PAS-102, no. 8, pp. 27932798,
Aug. 1983.
[9] T. F. Chan, Analysis of self-excited induction generators using an iterative
method, IEEE Trans. Energy Convers., vol. 10, no. 3, pp. 502507, Sep.
1995.
[10] K. S. Sandhu and S. K. Jain, Operational aspects of self-excited induction
generator using a new model, Elec. Mach. Power Syst., vol. 27, no. 2,
pp. 169180, 1999.
[11] I. A. M. Abdel-Halim, M. A. Al-Ahmar, and M. Z. El-Sherif, A novel
approach for the analysis of self-excited induction generator, Elec. Mach.
Power Syst., vol. 27, pp. 879888, 1999.
[12] K. S. Sandhu, Iterative model for the analysis of self-excited induction
generators, Elec. Power Comp. Syst., vol. 31, no. 10, pp. 925939, 2003.
[13] D. E. Goldberg, Genetic Algorithms in Search, Optimization, and Machine
Fig. 10. Flowchart for optimizing the control variables of SEIG. Learning. New Delhi, India: Pearson Education, 2001.
234 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 21, NO. 1, MARCH 2006

[14] D. Kalyanmoy, Optimization for Engineering Design: Algorithm and Ex- Mahender Kumar Soni was born on November 14,
amples. New Delhi, India: Prentice-Hall, 1995. 1950. He received the B.Sc.Engg. and M.Sc.Engg.
[15] K. F. Man, K. S. Tang, and S. Kwong, Genetic algorithms: Concepts and degrees from Kurukshetra University, Kurukshetra,
applications, IEEE Trans. Ind. Electron., vol. 43, no. 5, pp. 519533, Oct. India, in 1972 and 1975, respectively, and the Ph.D.
1996. degree from Kurukshetra University, in collaboration
with Indian Institute of Technology, Delhi, India, in
Dheeraj Joshi was born on July 11, 1978. He re-
ceived the B.E. (electrical) degree from the Univer- 1988.
Currently, he is a Professor and Director
sity of Rajasthan, Jaipur, India, in 1998 and the M.E.
at C.R. State College of Engineering, Murthal,
(power apparatus and electric drives) degree from the
Sonipat, India. His areas of interest are control,
University of Roorkee, Roorkee, India, in 2000. He
is pursuing the Ph.D. degree at the National Institute microprocessor-based instruments, and networking.
Dr. Soni received the Best Paper Award from the System Society of
of Technology, Kurukshetra, India.
India in 1987.
Currently, he is a Lecturer in the Electrical En-
gineering Department, National Institute of Tech-
nology, Kurukshetra, India. His areas of interest
are artificial intelligence, power electronics, and
electric drives.

Kanwarjit Singh Sandhu was born on December


21, 1957. He received the B.Sc.Eng. (electrical),
M.Sc. (electrical), and Ph.D. (electrical machines)
degrees from Regional Engineering College, Kuruk-
shetra University, Kurukshetra, India, in 1981, 1985,
and 2001, respectively.
He joined the Electrical Engineering Department
of Regional Engineering College, Kurukshetra, as a
Lecturer in January 1983. Currently, he is an Assis-
tant Professor in the Electrical Engineering Depart-
ment, National Institute of Technology, Kurukshetra.
He has many publications in the area of induction generators. His areas of in-
terest include electrical machines, wind energy conversion, and power systems.

También podría gustarte