The process of self excitation in induction generators

J.M. Elder, B.E., J.T. Boys, M.E., Ph.D., and Prof. J.L. Woodward, M.A.Sc, B.E., Fellow IPENZ
Indexing terms: Synchronous generators, Asynchronous generators, Induction motors Abstract: The phenomenon of self excitation in induction machines is examined and a physical interpretation of how self excitation occurs is presented. The important parameters are shown to be the shaft speed, the residual magnetism, the reduced permeability at low magnetisation, and the size of the capacitor connected to the machine. The results are important in applications where induction machines are used as stand-alone generators for a power system, as the reliability of self excitation in such cases must be very high.

List of principal symbols

a C Im ir,is Ir, Is k Lm Lr,Ls

m P rr, rs Voc

= factor determining variation in amplitude with time of asynchronous currents, 1/s = terminal capacitance, F = magnetising current, A = instantaneous rotor, stator current, A = rotor, stator complex current, A = a constant proportional to the strength of the remanent magnetism in the rotor = magnetising inductance, H = rotor, stator leakage inductance, H
= generalised angular velocity = rotor electrical speed, rad/s, pu rotor, stator resistance, "2 = open-circuit voltage produced at rated frequency,

co co0 cor

V = asynchronous electrical stator frequency, rad/s = rated frequency, rad/s = synchronous electrical frequency proportional to rotor speed, rad/s = circuit reactances, Q,

Introduction

The phenomenon of self excitation in induction machines is nowadays well known. In most practical circumstances, however, such self excitation is undesirable as it can cause severe overvoltages [1,2] thereby stressing the insulation of the machine, or it can cause torque and machine speed fluctuations [3, 4] which detract from the performance of the machine and may cause significant overheating. With increased use of power-factor correction equipment, methods for determining the precise operating limits, before self excitation occurs, have been developed and refined [36] In these situations, self excitation occurs in an already excited machine operating initially under nominally correct conditions. On the other hand, the conditions under which self excitation can occur in an unexcited machine, spun up to speed from rest with only the residual magnetism of the rotor to induce any initial currents, are not well understood but are of increasing importance with today's trends in power-generating equipment. De Mello [7] discusses the concept of induction generators as the principal generation source in a power system and lists a number of advantages. The machines are of intrinsically simpler construction, they are more robust, and offer potential
Paper 2292 B (Pi), first received 7th April and in revised form 7th October 1982 The authors are at the Department of Electrical Engineering, University of Auckland, Auckland, New Zealand IEEPROC, Vol. 130, Pt. B, No. 2, MARCH 1983

improvements in power-system dynamic performance. In addition, the fault current level drops to zero as the exitation collapses with the terminal voltage. Casel and Knitterscheidt [8] discuss the use of induction generators as high reliability generators for critical areas such as fire-fighting equipment. In many of these applications, an induction generator must operate in a stand-alone application, and consequently it is imperative that self excitation both occurs and can be maintained. This paper discusses the mechanism by which self excitation occurs and discusses how the terminal voltage can, under favourable conditions, start from the relatively insignificant voltage casued by remanent magnetism in the rotor and build up to the full rated voltage of the machine. The physics of the process is developed by considering the machine to behave at first as a synchronous machine with a weak permanent-magnet rotor and then as an asynchronous machine as the terminal voltage rises to some useful value. For the theoretical development, the two possible states of the machine are represented by the simple equivalent circuits shown in Fig. 1. The machine starts as a permanentmagnet alternator (Fig. la) and switches to an asynchronous machine (Fig. Ib), the transition being dependent on the machine parameters, the rotor speed and the size of the terminal capacitors. The circuit parameters are all assumed to be constant except for the magnetising inductance which varies in the way shown in Fig. 2 (this curve was obtained experimentally as outlined in Appendix 7, on the machine described in Section 3). The drop in inductance at high magnetising currents is well known and determines the final steady-state voltage. The importance of the low inductance at low currents is less well recognised, but it is this feature

Fig. 1

Equivalent circuits

a Synchronous mode of self-excited induction generator b Asynchronous mode 0143-7038/83/020103 + 06 $01.50/0 103

that largely determines the initial self-excitation characteristics of the machine. This paper presents a theoretical analysis of the way in which self excitation occurs as a change in the state of the machine. The theory is complemented with several experimental results showing possible failure modes for such excitation. Finally, a start-up sequence under which self excitation is most likely to occur is presented and evaluated.
O.Ar

permanent-magnet alternator has marked saliency and must be analysed by the 2-axis method. Round-rotor analysis is used here, because although different levels of saturation along, and perpendicular to, the axis of remanent magnetism do create magnetic saliency, the effect is only significant at rotor speeds well below that which results in a synchronous transition. Possible analysis of the machine as a hysteresis generator has also been considered. However, such treatments of the hysteresis machine as those by Teare [10] and Copeland and Slemon [11] have been concerned principally with motor operation, and have considered the use of hard magnetic materials with hysteresis loops of regular shape incorporated in rotors of simple geometry. That the peak airgap flux density is not explicitly known in the case of the self-excited generator imposes a further major difficulty. From Fig. 3, the magnetising currentIm can be determined:

0.2

where cor is the synchronous electrical frequency proportional to the rotor speed, k = VOC/CJ0 (a constant proportional to the remanent magnetic flux density in the rotor) and Voc is the open-circuit voltage produced at rated system frequency co0.
1.0 2.0

Fig. 2 2

Variation in magnetising inductance with magnetising current Theoretical considerations

For the two simple equivalent circuits shown in Fig. 1, we consider the relationships between the terminal voltage, current and rotor speed.
2.1 Synchronous mode

The equivalent circuit for the synchronous mode, Figure la, has the form of a resonant circuit in which the forcing function is the remanent magnetism in the rotor, inducing synchronous currents in the stator. The amount of remanent magnetism present in the rotor is measured in terms of the open-circuit (capacitors disconnected) voltage at rated frequency. To determine the response to the forcing function, it is incorporated into the equivalent circuit as a voltage source, the magnitude and frequency of which are dependent on the rotor speed, (as shown in Fig. 3). This model of machine behaviour may be compared with the equivalent constant excitation machine discussed by Mainer [9] in his theoretical treatment of permanent-magnet alternators. However, factors which rule out a direct application of Mainer's methods in this case include the ill-defined distribution of remanent flux in the rotor, the relatively high values of recoil permeability which apply to the soft iron, and departure from linear operation. The practical

p,pu

Fig. 4 Possible operating points of induction generator in synchronous mode for three levels or remanent magnetism

j r sin u>r t

Fig. 3 Synchronous-mode equivalent circuit incorporating remanent magnetism as frequency-dependent voltage source
104

In this circuit, the normal response of a resonant circuit to a forcing function is modified by a nonlinear magnetising inductance and the variation in magnitude of the forcing function with rotor speed. Solving eqn. 1, using the values for the magnetising inductance and current from Fig. 2 and the machine parameters from Appendix 7, a set of possible synchronous operating points can be obtaind. Plotting these points produces curves of the form shown in Fig. 4, for values of remanent magnetism varying form an opencircuit voltage of 0.5 V to 0.1 V at 50 Hz. The curves rise gradually until increasing current causes a rapid rise in magnetising inductance (see Fig. 2), and an associated sudden decrease in synchronous resonant frequency which causes the curves to bend back upon themselves. If the machine speed is increased montonically from zero, the response will follow a curve of the form shown until the knee is reached (point X for instance), at which point, increasing rotor speed must cause the operating point to jump past synchronous resonance (to point Y). It must be noted that these curves, along with those for the asynchronous mode, can be shifted with respect to machine speed by simply varying the amount of capacitance connected to the machine. The range 1.0 to 1.4 per unit was convenient for the equipment used to drive the induction machine. This speed range is simply reduced by increasing the capacitance.
IEEPROC, Vol. 130, Pt. B, No. 2, MARCH 1983

2.2

The asynchronous mode

Using the circuit of Fig. \b, and following the method of Wagner [3], the current supplied by the machine to the capacitor (the stator current), is is given by is = Re(Isemt) (2)

where Re means the real part, / s is the complex stator current

and m = u+joj
Similarly, for the rotor circuit, the rotor current ir is i r = Re(Ire{m~JP)t) (3)

for m = a + jto always produces two roots that have a negative and are of no further interest here. However, the third root has an a that can be negative, positive or zero, and it is this root which determines whether the rotor current, the stator current and voltage, decrease, increase, or are sustained. Fig. 5 shows a plot of a against machine speed for two values of Lm. A resonant speed occurs for which a = 0. For asynchronous currents to grow, and therefore for self excitation to occur, the resonant speed must be exceeded. The speed at which resonance occurs can be reduced by increasing the terminal capacitance.
1.5

where Ir is the complex rotor current and p is the rotor electrical speed. The use of Kirchhoffs loop equations with standard phasor notation yields ls(rs + m(Ls + Lm) + 1/mC) + IrmLm and = 0 = 0 (4)

1.0

CC=-O.15

cr=o.o

cr=o.2

/ r (r r + (m -jp) (Lr + Lm)) +

Solving eqns. 4 and 5 for Ir gives

-jp)Lm

(5)
05

[{m 2 C(LS + Lm) + mCrs + l}{-rr-(m-jp)(Lr =0 m (m-jp)CLl]Ir

Assuming/,. =0, and simplifying eqn. 6 in terms of m, yields

+ Lm)}
(6)

m3C[(Lr +

Lm)(Ls+Lm)-L?n]+m2C[(Ls+Lm)rr
+ Lm)rs] (7)

1.0 P,pu

2.0

m [Cr8rr + (Lr + Lm) -jpC(Lr {rr-jp(Lr + Lm)} = 0

Fig. 6 Contours of a representing exponential growth and decay of asynchronous currents in the induction machine having parameters given in Appendix 7

At this point, the equation may be normalised by converting to per-unit values. Then, if p is expressed as a fraction of coo, we have from eqn. 7: -Xl] +(m/co 0 ) 2 Xm)

[(Xs + Xm)rr + (Xr + Xm)rs -jp{(Xr+ (Xs + Xm) -Xl}

+ (m/coo) [rsrr + Xc(Xr + Xm) + Xm)} = 0 (8)

- jp{Xr + Xm)r8] + Xc [rr -jp(Xr

The roots of this cubic equation define the conditions under which asynchronous machine operation is possible. Solving

From the values of magnetising inductance shown in Fig. 2, contours of machine current and speed for different values of oc can be obtained, as shown in Fig. 6. The steadystate operation of the asynchronous machine must be at a point of resonance; i.e. a = 0. As can be seen, at a particular speed there are two operating points which satisfy this criterion, A and B. Point A represents the point at which the machine is fully excited. Point B, however, represents an unstable condition, in that any change in speed will cause the machine to drop out of resonance, an increase in speed tending to cause self excitation with resonance being regained at point A, whereas a decrease will cause the machine to drop back to the synchronous mode. It is point B which is of interest, and Fig. 7 shows the curves expanded about this area.
a=0005 / or=0.0075

cr=0.0025

Lm=

150 =0.01

Q02

=0.24

,100

cf=-O.OO! 50

2.0

-0.02

1.0

1.2 P.PU

1.4

Fig. 5 Plot of a against machine speed for two values ofmagnetisng inductance, Lm = 0.4 Hand 0.24H IEEPROC, Vol. 130, Pt. B, No. 2, MARCH 1983

Fig. 7 Contours of a for area of machine operation critical to initiation of self excitation. Positive a corresponding to growth of asynchronous currents leading to excitaton 105

2.3

Interaction between modes

The interaction between the synchronous and asynchronous modes can be seen by combining the contour a = 0, for asynchronous operation, with the curves for synchronous response as shown in Fig. 4. For a given value of remanent magnetism, a machine started from rest will have a capacitor current that follows the appropriate synchronous curve (A) up to a knee point (X). If the speed is further increased, the current must jump discontinuously to point Y, which is within the area where asynchronous operation is possible. Under these conditions, an asynchronous component will grow rapidly and the terminal voltage will increase to the appropriate value. The fixed rotor poles due to remanence are destroyed and the synchronous component of current decays to zero. If the synchronous operating point is close to point X, only a small disturbance is required to initiate self excitation.

At a slightly lower ramp rate (curve b), self excitation is almost achieved but fails as the synchronous machine loses some remanent magnetism. Self excitation then occurs at a higher speed. At a still lower ramp rate (curve c) there is considerable interference between synchronous and asynchronous components, resulting in substantial current fluctuations until self excitation occurs. At extremely slow rates (curve d), the remanent magnetism is steadily reduced until finally self excitation occurs. This demagnetisation was verified by halting and reversing the ramp in machine speed. The machine's response did not retrace curve d but dropped onto a curve corresponsing to a lower remanent magnetism.
200 200

\ 1.0
P.PU

1.0

c
700

\
\ \

200

1 >\ 's
1.0

V p

1.0

Fig. 8 Vector diagrams a Corresponding to point X in Fig. 4 for circuit shown in Fig. 3 b Corresponding to point Y

200

The effect of this transition on the synchronous machine can be seen from the vector diagrams in Fig. 8 (for circuit, see Fig. 3). Below resonance, the load angle 6 is such that The MMF Fx produced by the current flowing, reinforces the remanent MMF F2 to produce the resultant MMF Fo. While above resonance, the current acts to reduce the remanent magnetism. The change in phase of the current between the two operating points is of the order of 150 degrees. For the resonant-frequency line to be crossed and self excitation to occur, the synchronous machine must sustain this large change in phase effectively a large loadangle swing.
3 Experimental observations

10

Fig. 9 System responses Curves a to d show response to reducing ramp rates in machine speed. Curve e is the response to a very low level of remanent magnetism Ramp rates, rad/s 2 : (a) 1.0, (6)0.82, (c) 0.44, (d)0.12

Under most operating conditions, the remanent magnetism will persist for a sufficient time to guarantee self excitation when the change in state occurs the rate of change of rotor magnetism will be limited by the highly conductive rotor. If, however, the transition is attempted at very slow speeds, it is considered likely that problems may occur. To investigate the consequences of slow load-angle swings, a number of experiments were performed. These experiments involved accelerating the induction machine from rest at different ramp rates. Fig. 9 shows typical system responses for a 3-phase, 4-pole, 2.25 kW induction machine for rotorspeed ramp rates ranging from 1.0 rad/s2 down to 0.12 rad/s 2 . The responses are plotted as machine current against rotor speed with the contour a = 0 shown (see Appendix 7 for comment on the position of this contour). Curve a, at the highest ramp rate, shows a typical response where the initial synchronous build up leads to a jump through synchronous resonance causing the machine to self excite. 106

There appear to be two mechanisms by which the remanent magnetism of the machine can be altered. At very slow ramp rates, the demagnetising armature reaction caused by the load angle is sufficient to keep the machine below the area of possible asynchronous operation up to a speed of approximately 1.24 pu in this case. At higher ramp rates, asynchronous currents do appear but may 'beat' or interfere with the synchronous currents. Such interference may cause the operating point to vacillate in and out of the area of possible asynchronous operation (curve c). Yet again, the asynchronous component may degauss (either partially or completely) the rotor of the machine as it drives it through many cycles yet fails to self excite if the collapse of the synchronous component is faster than the growth rate of the asynchronous one (curve b). In practice, of course, all these effects are present at the same time and it is difficult to distinguish between them in a real case. At high ramp rates, no interference effect is observed, as the rate of growth of the asynchronous component is too high. The machine response for an almost completely demagnetised rotor is shown in Fig. 9 curve e, with self excitation
IEEPROC, Vol. 130, Pt. B, No. 2, MARCH 1983

occurring at a speed of approximately 1.44 pu. Even with a completely demagnetised rotor, self excitation will occur if the speed exceeds that resonant speed determined by the minimum value of magnetising inductance and the value of terminal capacitance. Demagnetisation of the rotor was achieved by using a critical ramp rate which produced the maximum reduction in remanent magnetism after the machine failed to excite. The load angle effect alone cannot produce this level of demagnetisation as, at extremely low ramp rates, self excitation occurs at speeds below 1.24 pu.
4 Practical implications

the machine's rotor and bearings must be rated for the higher speed) (d)by adding sufficient terminal capacitance to reduce the resonant speed to below the rated machine speed. In the last two methods, care must be taken to avoid large overvoltages when the machine excites. In large machines, these two methods may not be practical owing to the large reduction in magnetising inductance at very low currents. Typically, the ratio between the inductance in the essentially linear operating range and the inductance for very small currents is of the order of 5 :1 for the iron alone. The presence of the airgap reduces the effective ratio. Thus a small machine may have a ratio of 2 :1, whereas large machines will approach 5:1. With high ratios, very large capacitance or very high machine speeds would be required to ensure self excitation by either of the last two methods. Experimental work in this area is sometimes made rather confusing in our experience by initially unexpected behaviour. One example is that the small transient resulting from disconnecting and reconnecting the capacitance to a nonexcited machine is often seen to be sufficient to start self excitation. Also small slip-ring induction machines may not excite owing to the discrete voltage drop at the brush contacts. 5 Conclusions

A major problem in starting stand-alone induction generators is that of guaranteeing self excitation when a machine having an unknown amount of remanent magnetism in the rotor is accelerated from rest. An understanding of the mechanism of self excitation, as presented in this paper, permits recommendations to be made for improving the reliability of start-up.

Q3

10.2

The feasibility of using an induction machine as a standalone generator is enhanced by the treatment of self excitation presented in this paper. The actual method of starting may vary from one installation to another. Using the methods outlined above, however, the reliability of starting can be very high and advantage can be taken of the ruggedness, simplicity and low cost of the induction machine.
6 References

0.1

100

200

300

Im,mA

Fig. 10 Plot of magnetising inductance showing error bars

A reduction in the remanent magnetism can occur at reduced acceleration rates, as previously shown, if the terminal capacitor remains connected while the machine is run up. Similarly, although it is beyond the scope of this paper to examine this subject in greater detail, any load connected to the machine demonstrably reduces the ability of the machine to excite. Thus, to improve the reliability of start-up, it is recommended that, in all situations, the capacitor remain disconnected until the final machine speed is attained, and furthermore that any major load remain disconnected until the machine is fully excited (there will usually be some load imposed by the control devices connected to the generator output). In addition, the reliability of starting can be made very high by one of the four following methods: (a) by passing a DC current through the machine before it is run up to speed, sufficient remanent magnetism may be guaranteed (b)by switching in charged terminal capacitors. If the capacitors are charged to a high voltage, say rated machine voltage, the discharge current is normally sufficient to cause self excitation even with a degaussed rotor (c) by increasing the machine speed above the rated value, causing the resonant speed at low magnetisation to be exceeded, and thereby initiating self excitation (note that
IEEPROC, Vol. 130, Pt. B, No. 2, MARCH 1983

1 WAGNER, C.F.: 'Self excitation of induction motors', AIEE Trans., 1939, 58, pp. 47-51 2 SMITH, I.R., and SRIHARAN, S.: Transients in induction machines with terminal capacitors', Proc IEEE, 1968, 115, pp. 519 527 3 WAGNER, C.F.: 'Self excitation of induction motors with series capacitors', AIEE Trans., 1941, 60, pp. 1241-1247 4 LIMEBEER, D.J.N., and HARLEY, R.G.: 'Subsynchronous resonance of single-cage induction motors', IEEProc. B, Electr. Power Appi, 1981, 128, (1) pp. 33-42 5 BARKLE, J.E., and FERGUSON, R.W.: 'Induction generator theory and application', AIEE Trans., 1954, 73, Pt III A, pp. 12-19 6 KITAEV, A.V., and ORLOV, I.N.: 'On the physical mechanism of asynchronous self excitation and self oscillation', Electr. Technol. USSR, 1979, (2), pp. 19-28 7 DE MELLO, F.P., and HANNETT, L.N.: 'Large scale induction generators for power systems', IEEE Trans., 1981 PAS-100, pp. 2610-2618 8 CASEL, J., and KNITTERSCHEIDT, H.: 'Induction generators for generating and emergency power supplies of the future', Electrotech. Z. ETZ, 1981, 102, pp. 139-141 9 MAINER, O.E.: Theoretical treatment of permanent magnet alternators', Int. J. Electr. Educ, 1968, 6, pp. 417-425 10 TEARE, B.R.: Theory of hysteresis-motor torque', AIEE Trans., 1940, 59, pp. 907-912 11 COPELAND, M.A., and SLEMON, G.R.: 'An analysis of the hysteresis motor YJEEE Trans., 1963, PAS-82, pp. 34-42
7 Appendix

The machine used in the experimental investigations was a 3-phase, 4-pole, 2.25 kW induction machine. The machine parameters were rr = 2.1 S7, rs = 2.91 1, Lr = 13.5mH,Z,s = 13.5mHandC=25MF. These parameters were assumed constant. To measure the magnetising inductance Lm, the machine was driven at synchronous speed and the applied terminal voltage varied.
107

However, at very low machine currents, even with the use of precise electronic wattmeters, the measurement of magnetising inductance was extremely difficult. To obtain more accurate values for Lm, the remanent magnetism was utilised to provide the voltage required to measure Lm at very low machine currents. The machine was represented by the circuit shown in Fig. 3 with the terminal capacitor disconnected. From this circuit, the magnetising inductance was calculated by measuring the open-circuit

voltage and the short-circuit current, each test being repeated several times to ensure that a stable level of remanent magnetism had been achieved. However, owing to the limited magnitude of the remanent magnetism, this method could only produce results for machine currents below 30 mA. Fig. 10 is the plot of Lm against machine current, with the inclusion of error bars to give an indication of the uncertainty in the values of Lm and therefore in the exact position of the a contours, at low machine currents.

108

IEEPROC, Vol. 130, Pt. B, No. 2, MARCH 1983