S.S. Murthy, Ph.D., O.P. Malik, Ph.D., and A.K. Tandon, M.Tech.
Abstract: An analytical technique using the 'Newton-Raphson' method is presented to identify the saturated
magnetising reactance and the generated frequency of a self-excited induction generator for a given capaci-
tance, speed and load. The technique is shown to be very efficient in analysing such systems under steady state.
Computed results are compared with the experimentally obtained values on a laboratory machine, and a.
reasonable correlation has been observed. Effects of various system parameters on the steady-state perform-
ance have been studied, and the results presented provide guidelines for optimum design of such systems.
1 Introduction 2 Theory
If an appropriate three-phase capacitor bank is connected In the analysis presented in this Section, the following
across an externally driven induction machine, an EMF is in- assumptions are made:
duced in the machine windings due to the excitation provided (a) Only the magnetising reactance is assumed to be affected
by the capacitor. This phenomenon is termed the 'capacitor by magnetic saturation, and all other parameters of the equi-
self excitation', which can be exploited to operate the induction valent circuit are assumed to be constant. Self excitation
machine as a generator. The induced voltages and currents results in the saturation of the main flux. As the value of the
would continue to rise, but for magnetic saturation in the magnetising reactance xm reflects the magnitude of the main
machine which results in an equilibrium state being reached. flux, it is essential to incorporate in the analysis the variation
A capacitor self-excited induction generator offers certain of xm with the saturation level of the main flux [4]. Passage
advantages over a conventional synchronous generator as a of the leakage fluxes occurs primarily in the air [4], and thus
source of isolated power supply. Reduced unit cost, brushless these fluxes are not affected to any large extent by the satu-
rotor (squirrel cage construction), absence of a separate DC ration of the main flux.
source and ease of maintenance are among the advantages. (b) Leakage reactances of stator and rotor, in per unit, are
Owing to changed emphasis on energy problems, development taken to be equal. This assumption is normally valid in
of suitable isolated power generators, driven by unconventional induction-machine analysis [4].
energy sources such as wind, biogas, etc., has recently assumed (c) Core loss in the machine is neglected, although the
greater importance [1]. The development of static power con- analysis can be easily extended to account for core loss.
vertors [2, 3] has facilitated the control of self-excited gener- (d) MMF space harmonics and time harmonics in the induced
ators, regarding output voltage and frequency. voltage and current waveforms are ignored. This assumption is
Steady-state analysis of such generators is of interest, both valid in well designed machines. Further, the experimental
from the design and operational points of view. Knowing the waveforms of generated voltages exhibited negligible time
parameters of the machine, it should be possible to determine harmonics.
its performance for given capacitance, speed and load con-
ditions. If the terminal voltage and frequency are known, as in
the case of a machine connected to an infinite busbar, the
prediction of performance is straightforward [4, 5 ] . However,
in a capacitor-excited generator used as an isolated power
source, both the terminal voltage and frequency are unknown
and have to be computed for a given speed, capacitance and
load impedance. The analysis is complicated owing to the Fig. 1 Equivalent circuit of the induction generator with load
magnetic saturation in the machine and the need to choose
suitable parameters corresponding to this saturated condition. The steady-state equivalent circuit of a capacitor self-
A procedure to identify the saturated parameters and the excited induction generator with a resistive load connected at
generated frequency for a given load using the Newton- its terminals is shown in Fig. 1, where:
Raphson method is presented in this paper. Using these Rs, Rr = per phase stator and rotor (referred to stator)
identified values the steady-state performance can be easily resistance
calculated from the equivalent circuit. The procedure is shown x
is> xir =
P e r P na se stator and rotor (referred to stator)
to be simple, comprehensive, efficient and well suited for leakage reactance
digital simulation. A computer algorithm is developed to pre- xm = magnetising reactance
dict the performance using the proposed method. Simulated xc = per phase capacitive reactance of the terminal
results are compared with the experimental values obtained on capacitor C
a laboratory machine under similar conditions. RL = load resistance per phase
Having confirmed the validity of the analytical technique, (all reactances referred to above relate to the base frequency/)
the effect of various system parameters on the steady-state F,v = p.u. frequency and speed, respectively
characteristics is studied. The results are presented graphically h> In II = stator, rotor (referred to stator) and load current
and discussed in order to provide guidelines for the designers. per phase
Vt> vg ~ terminal and airgap voltage, respectively
Paper 2193C(Pi, P9), received 22nd April 1982 The process of transient voltage build-up under self excitation,
Dr. Murthy and Mr. Tandon are with the Department of Electrical and ultimately reaching steady state, can be explained through
Engineering, Indian Institute of Technology, New Delhi 110016, India. the operational equivalent circuit [6], which results in a
Dr. Murthy is presently visiting, and Dr. Malik is with, the Department characteristic polynomial in the time-derivative operator d/dt.
of Electrical Engineering, University of Calgary, 2500 University Drive
NW, Calgary, Alberta, Canada T2N 1N4 If one of the roots of this polynomial, for a given C, v and RL,
260 0143-7046/82/060260 + 06 $01.50/0 IEEPROC, Vol. 129, Pt. C, No. 6, NOVEMBER 1982
has positive real part, self excitation would occur. As the voltage The Newton-Raphson method requires the initial guess of the
builds up, the resulting airgap flux drives the machine into values of the unknowns, say xmo, Fo. Because F v and xm is
saturation, which gradually decreases the saturated magnetising less than unsaturated, xm, xmo, and Fo can be chosen as
reactance to a value that results in the positive real part of the
root being reduced to zero. This signifies the steady-state xmo = xm (unsaturated), I
(9)
condition being reached, and this saturated value of xm has to Fo = v
be used in the steady-state analysis.
Therefore, in Fig. 1, both the steady-state values of xm and Now/0 =f(xmo, Fo) = 0 and^o = g(xmo,Fo) = 0.
F are unknown and have to be determined for the given capaci- After the first iteration, xm and F will assume the values
tance, speed and load, to calculate the steady-state response xmo + h andF o + k in the process of satisfying eqns. 3 and 4;
using this equivalent circuit. The proposed procedure is ex- the increments h and k are given by [7]:
plained in the following text.
From Fig. 1, the loop equation for the current Is can be -fo
written as (10)
= 0 (1)
where This iteration process can continue till the desired accuracy is
reached, i.e. when \f(xm, F)\ < e and \g(xm, F)\ < e. A
F-v simple computer algorithm can be developed incorporating the
R.JX
RjF-jxjF U F-v
above procedure and steady-state values of xm and F can be
+ j(xm+xlr) computed for any value of C, v and RL.
. Having thus determined F and xm, the next step is to
(2) calculate the airgap voltage Vg and the terminal voltage Vt
Under steady-state self-excitation Is = 0. Therefore, from eqn. (Fig. 1). For this purpose, one can use information regarding
1, Zs = 0, which implies that both the real and imaginary parts the variation of xm with the quantity Vg/F which relates to
of the right-hand side of eqn. 2 would be separately zero. Sub- the airgap flux. The information required can be obtained ex-
stituting x^ =xlr=xl, this simplifies to the following two perimentally by driving the induction machine at synchronous
nonlinear simultaneous equations with xm and F as unknown speed corresponding to the line frequency, i.e. F= 1, and
variables: measuring the magnetising reactance for different input
voltages at line frequency. A curve of Vg/F against xm can be
f{xm, F) = (Clxm + C2)F3 + (C3xm + C,)F2 ^ plotted using the experimental results. From this curve, Vg/F
for the above steady state saturated xm can be obtained.
(CsXm+C6)F+(ClXm+C8) =0 (3) Knowing F, the airgap voltage Vg can be computed.
g(xm,F) = (DlXm+D2)F2+(D3xm+D4)F + With Vg, xm, F, xc, v, RL and machine parameters known,
calculation of the terminal voltage Vt and the load current is
D5 = 0 (4) straightforward using the equivalent circuit of Fig. 1. Expres-
where sions for the respective variables are summarised below:
1.0
\
dO.75 - <
<^---K=1.0
>"
0.5
Q25
1.5
OJO 0.5 1.0 1.5 2.0 2.5
1.25
Fig. 9 Effect of magnetising reactance
xm = K X xm (nominal) for the same Vg/F
1.0
1.75r
0.75 1.5
-U=0.7
02 5 -0.5
id) Effect of magnetising reactance: From the preceding
Sections, it is apparent that the magnetisation characteristic,
indicated by the variation of xm with airgap flux, has an im-
portant bearing on the load characteristics. Both the saturated o.O 0.5 1.0 1.5 2.0 2.5 30
and unsaturated magnetising reactance can be changed by
appropriate alteration in design. The variation of saturated
Fig. 10 Effect of speed
xm can be approximately represented by a linearised equation
similar to eqn. 12. To study the variation of saturated xm, gc/v2 = constant = 0.747
V t, F
264 IEEPROC, Vol. 129, Pt. C, No. 6, NOVEMBER 1982
much affected by xm. Using increased xm, the connected such an analysis, as the solution can be obtained in three or
capacitance can be decreased for the same Vt. A designer has four iterations, requiring very little computational time, to an
to compare the economy of choosing a larger frame or a higher accuracy of 10~ 4 . Simulated results on a laboratory machine
valued capacitor for the desired output voltage. compare quite favourably with the experimentally measured
values. The operating characteristics are governed by the
4.1.3 Effect of speed: Self-excited induction generators have magnetic saturation, which has been incorporated by the use
been suggested as suitable for power generation using wind of experimental data relating the variation of magnetising
energy [1]. However, in such a situation the prime mover reactance with the airgap flux.
speed changes over a wide range. To provide suitable regulators, Families of characteristics showing the effect of various
the effect of speed on the output voltage and frequency should system parameters are presented. As these curves are presented
be determined. Further, the minimum capacitance to cause in per unit values, they may provide guidelines for appropriate
self excitation at any speed is approximately proportional to design of the system. Whereas small changes in stator resistance
the inverse of the square of the speed. At decreased speed, the and leakage reactance do not affect their performance, the
capacitance has to be increased. Fig. 10 shows the family of rotor resistance, magnetising reactance, speed and terminal
load characteristics for speeds v = 0.7, 0.8, 0.9, 1.0, 1.1 and capacitors are found to be very sensitive parameters in deter-
1.2 p.u., by keeping the value ofgc/v2 constant equal to 0.747. mining the terminal voltage and frequency, in addition to their
The terminal voltage increases almost linearly with speed for variation with load.
the same power output; e.g. at Pout = 1.0, Vt almost doubles
if the speed is changed from 0.7 to 1.2 p.u. Output frequency 6 References
is directly proportional to speed. By the parallel family of
curves, it is apparent that both the voltage and frequency 1 WATSON, D.B., ARRILLAGA, J., and DENSEM, T,: 'Controllable
regulation are almost the same at all speeds. d.c. power supply from wind-driven self-excited induction machines'.
Proc. IEE, 1979, 126, (12), pp. 1245-1248
2 ARRILLAGA, J., and WATSON, D.B.: 'Static power conversion
5 Conclusion from self-excited induction generator', ibid., 1978, 125, (8), pp.
743_746
An analytical technique using the N&wton-Raphson method 3 NOVTONY, D.W.,% GRITTER, D.J., and STUDTMAN, G.H, 'Self
has been described to predict the steady-state behaviour of excitation in inverter driven induction machines', IEEE Trans. 1977,
capacitor self-excited induction generators. A computer PAS-96, pp. 1117-1125
4 ALGER, P.L.: 'Induction machines (Gordon and Breach, 1970)
algorithm using this technique has been developed. It identifies 5 SAY, M.G.: 'Alternating current machines' (Wiley, 1976)
the values of saturated magnetising reactance and the output 6 MURTHY, S.S., BHIM SINGH and TANDON, A.K.: 'Dynamic
frequency for the given capacitance, speed and load. The models for the transient analysis of induction machines with asym-
steady-state equivalent circuit can then be used to compute metrical connections, Electr. Mach. and Electromech., 1981, 6, pp.
the performance. The procedure is shown to be well suited for 479-492
7 GROVE, W.E.: 'Brief numerical methods (Prentice Hall, 1966)