You are on page 1of 8

Electrical Power & Energy Systems Vol. 18, No. 1, pp.

65-72, 1996 Copyright 1996 Elsevier Science Ltd

ELSEVIER

0142-0615(95)00tt62-3

Printed in Great Britain. All rights reserved 0142-0615/96$15.00+ 0.00

Some clarifications in the transient energy function method


M A Pai, M Laufenberg and P W Sauer Department of Electrical and Computer Engineering, University of Illinois, 1406 W Green St, Urbana, IL 61801, USA

The purpose of this paper is to clarify the evaluation of path dependent integrals in the energy function method for stability analysis in power systems. In the literature these are handled in an approximate manner through straight line approximation leading to closed form analytic expressions of the energy functions. This may not always be accurate. Here it is compared "with the trapezoidal method of integration along the .faulted trajectory as originally proposed by Athay et al. The paper also emphasizes some tutorial aspects for explaining the PEBS and the BCU method. Keywords: transient energy function, power system stability

used when detailed models are considered or if they are applied to reduced order and hence smaller systems. In structure preserving energy functions (SPEF) (or the sparse transient energy function (TEF) method) path dependent integrals exist due to voltage dependent real loads or contributions from the electrical variables of the machine 3. A variation of the potential energy boundary surface (PEBS) method which obviates the need to compute the post-fault stable equilibrium point (s.e.p.) is also proposed. This might reduce the computational burden in a quick screening of contingencies.

I. I n t r o d u c t i o n Historically, the energy function method for multimachine power system stability analysis used the classical machine representation with the internal node model and neglecting transfer conductances. Under these conditions we get a mathematical :model which is conservative and either the Lyapunov based method or the first integral method gives an equivalent energy function. As transfer conductances represent the effect of constant impedance loads, ignoring them gives erroneous results with respect to critical clearing times. Various efforts at approximating the transfer conductance terms analytically have been made, the most popular among them being the straight line approximation of the faulted trajectory 2. A somewhat obscure, but not so obvious approximation, is the assumption that the post fault s.e.p, is the same as the pre-fault s.e.p, for computing this integral. This point is also clarified in this paper. It is also shown that the method due to Athay et aL l of using the trapezoidal method is the correct one, as the path of integration is known from the faulted trajectory. While it may be true that using the straighl; line approximation does not introduce significant errors while using classical models for large systems, it may, if the energy function method is
Received 14 June 1994; revised 11 May 1995; accepted 1 June 1995

II. M a t h e m a t i c a l model 4-7 With the usual notation, the mathematical model for an m machine system with constant voltage behind reactance representation and constant impedance load approximation is given in the Centre of Inertia (COI) notation as:
Oi ~--'03i
Mi Mg ~i : Pmi - Peg - ~ PCOl

(1)

6=f.(O)

i= 1,2,...,m

(2)

The right-hand side in equation (2) has different parameter values (i.e. Gij and Bij values) in computing Peg and Pcol for the faulted period (0 _< t _< tot) and the post-fault period (t > td). The energy function for the post-fault system is constructed as:

1 m :" --~"~ g i ~

2g__~

m fOi - ~_~ I fi(O) dOi /=1 JO:

+ v ? e ( o ) :A VTO T

(3)

where 0 i and c?1 are the variables from the faulted trajectory. In the absence of transfer conductance terms Gij(i C j), the expression for Vpe(O ) can be expressed analytically in a closed form 4'5. Otherwise the Gij terms

65

66

Clarifications in transient energy function method." M. A. Pai et al.

contribute a path dependent term as follows l'

rule as:

VpE(O) = - - Z ei(Oi -- Os) - - Z


i=l

m-1 ~.~ I i=1 j=i+l

Cij (cos Oij -- COS 0/~)

Iij(o) =

D,j[cos(O? - 0;) + cos(0 s - 07) 1


x [(0 + 0;) - (0 s + 07) ] (9)

_ [o,+o,
J~+7 Dij cos 0q d( Oi + Oj)

]
(4)

= Veos + VMA6 + VD

where Cij ~-- EiEjBij , Dij = EiEjDij and Pi = P m i IEil2Gii . In computing equation (4), 0 i is obtained from the faulted trajectory and 0~ is the post-fault s.e.p. We focus on the evaluation of the terms on the right-hand side of (4). The third term is obviously path dependent.

This is a good approximation if 0 s is close to 0 . Fouad and Stanton 8 recognized this fact in their work and called it the V correction term. If one uses the potential energy boundary surface (PEBS) method 9, then even if the post-fault network is not equal to the pre-fault network, this term can be subtracted out of the energy function, i.e.

v(o,m) = VKE( ) + V E(O) -- V E(O )

(10)

Hence the potential energy can be defined with 0 as the datum as:

III. Different methods of evaluating the path dependent term


Define 0i+0s Iij = J07+07D ij cos Oij d( Oi + Oj)

f E(o)

vpe(o ) =

_
(5)
= i=l 0

lOldO,- ,I>l
fi(O) dOi
m i=l m-1 m F

doil
(11)

II1.1 Analytical approximation In the bulk of the literature, by assuming a straight line path of integration Iij is approximated analytically as:

= - ~-'~Pi(O~-O)- z

[Cij(cosOij-cosO~)

i=1 j=i+l [

Ii;

= (O, - Os) + (O; - O ~ (0 i 0 s) -- (Oj ~. Dij(sin Oij -- sin OSj)

(6)

f0,% ] - I DqcosOijd(O i + Oj) j0?+~o

Theoretically this is incorrect and the actual path needs to be taken into account as pointed out in Reference 1. The straight line path is used only in computing V~r while using the controlling u.e.p, method. In computing the potential energy term (4), it is shown that if the PEBS method is used to compute V~r,then it is possible to avoid computing the post-fault s.e.p, altogether by proper initialization of Vee(O). These two issues are now discussed.
111.2 Initialization of method

VpE(O) and its use in

the PEBS

While integrating the faulted trajectory in equations (1) and (2), the initial conditions on the states are Oi(O) = 0 and ~bi(0) = 0. In the energy function, the reference angle and velocity variables are 0/~ and ~bi(0) = 0. Thus at t = 0, we evaluate VeE(O) in equation (3) as:

At the PEBS crossing 0", lJ'eE(O* ) gives a good approximation to Vcr. To detect the PEBS crossing, we can monitor the quantity f T ( O ) . (0 -- 0 s) where f(O) refers to the post-fault parameters and 0 s is the post-fault s.e.p. The PEBS crossing 0* is the point where this quantity 1 changes from negative to positive. This requires a knowledge of 0 s. The method proposed by Kakimoto et al. 9 detects the PEBS crossing as the point where the potential energy Vee reaches a maximum value. Hence one can directly monitor _Vee and thus avoid having to monitor the dot p r o d u c t f r(0) (0 - 0s). This leads to an important advantage of not having to compute 0 s at all. In fast screening of contingencies, this could result in a s!gnificant saving in computation. The technique of using Vee(o) in the PEBS method has not been widely used so far in the literature and merits further investigation.
111.2.1 Trapezoidal a p p r o x i m a t i o n

VpE(O) = - ~-'~ [ff fi(O) dOi i=1 ai


m
=

(7)

rn 1 ~__~
e , ( - s) -

For t > 0, instead of computing Iij , by equation (6), we compute it by the trapezoidal rule as (letting n -k A t , k >_ 1):

- Z

i=1

i=1 j=i+l

Iij(n ) = Iij(n - 1) + 1Dij[cos(Oi(n ) - Oj(n))


"}- COS(0/Q"/ -- 1) - Oj(n - 1))]

F X [c,j(cos
I..

- cos o b)

[Oi(n) + Oj(n) - Oi(n - 1) - Oj(n - 1)],


n _> 1 (12)

-- d0/~+~ OijcosOijd(O i + Oj)] S0+0j [

= K (a constant)

(8)

If the post-fault network is the same as the pre-fault network, then K = 0. Otherwise the value of K in (8) should be included in the energy function. The path integral term in (8) is evaluated using the trapezoidal

with Iij(O) = 0. This initialization assumes that the constant K in (8) has been evaluated using (9) and added to the potential energy function in (11). Equation (12) is used in Reference 1 for the path dependent integral without adding equation (9). Thus as Oi(kAt), ~i(kAt), (k >_ 1) are evaluated from the fault-on trajectory we obtain V(O, Co) as (letting

Clarifications in transient energy function method." M. A. Pai et al.

67

kAt = n):
m W/

_ (0u - Os) +__(0~ --Oj_) Dq{sin(O~ - 0~) (e~ el) ( e ? - e:)


S'~lllP=l +p[(O u - e s ) - (O?-- OjlJJlp=O

V(O(n), &(n)) = ~ Z , Miw2(n) - Z Pi(Oi(n) - 0)


i=l

- ~:

m-1[
i=1 j = i + l
m--I m

i=1

Cq(cosOij(n)- cos0~j)
"]

(eu o:) -- (o?

(e? - o:) + (o2 - o,;)

0.s) Dij (sin 0~ - sin 0/~)

(16) +K

-j

Iij(n 1

(13)

where Iij(n) is given by equation (12). 0 ~ is used in computing K and is also needed to compute V~ discussed' in the next section. Equation (13) is equivalent to V(O, &) in equation (4) and is used in the following sections.

V. Numerical example
The well known three-machine nine-bus system n was chosen to illustrate the various observations in Sections III and IV. Figure 1 is the single line diagram. Several tests were done on this system to illustrate the two methods (PEBS and BCU) as well as the approximation techniques involving Dij terms. The first example case is the *7-5 contingency, this consists of a fault on bus 7, which is cleared by the opening of line 7-5. Using the step-by-step integration method, this contingency was found to have a critical clearing time of approximately 0.201 s. All of the plots

IV. Computation of Vcr


IV.1 Controlling u.e.p, method If the controlling u.e.p, method I is used, there is a need to compute 0 s. An efficient way of computing the controlling u.e.p, is through the boundary controlling u.e.p. (BCU) method 1'14. In this method, one continues to integrate after the PEBS crossing using the following reduced set of equations of the post-fault system.

Gcn. 2

I.oad C

Gen. 3

Oi=Pmi-Pei----~

Mip

co1~=f.(O)

i=l,...,m

(14)

with the initial condition 0(0) = 0", until II f(0)ll is minimum. This is the minimum gradient point (MGP) Ou. To get the exact u.e.p, we solve f(O) = 0 by the NewtonRaphson method with ~u as the initial guess to get 0 u. In most cases 0 u is sufficiently close to 0 u. Having obtained 0 ", symbolically denote V~, = V(0 u, &u) = Vee(O~), as o3u = 0. However, the critically cleared trajectory involving both the faulted and postfault system to evaluate Vpe(0 u) are not known, as t~ is not known! Hence various methods have been proposed to compute V?E(O~). Among them the straight line approximation from 0 s to 0 u is the most convenient one 1'2. VeE(O u) is given by
?n

_@ m

Load A

Gen. I

Figure 1. The three-machine, nine-bus system

VpE(OU) = -- Z
i=1 m-I

ei(ou - Os)
m

~
i=1

~
j=i+I

[ c . ( c o s O ~ - cosOb)]

1.5

+(o u o:1-(o?

(0u - Oi~) + (0~ - 0~) Dij(sin 0~. - sin 0~)

o:1

1
v

(15) The third term of equation (15) is derived as follows. Assume a ray from 0s to 0/u and then any point on the ray is 0, = 0s +p(O u - OS), (0 <_p < 1). Thus d(0g+0j) = dp(0 u - 0s + 0/u - 0~). The path dependent term in equation (4) is now evaluated at 0 u asl:

0.5
<

v~(o")

= [(o? - o h + ( o 7 -

o:)1

D,.:cos

"10

012 014 016 018

1 112 114 116 118 Time (sec)

{(0 -

0:) +p[(0? - 0h - (0y - 0:)]}dp

Figure 2. Generator angles of three-machine system, tcr = 0.19 s

68
30 25
20 15

Clarifications in transient energy function method: M. A. Pai et am.


t 1 J I I I $

gl0 ==

Ps <
(1)

-%

0:2

014

016 ' 018

Time (sec)

112

114

116

1'.8

Figure 3. G e n e r a t o r a n g l e s of t h r e e - m a c h i n e system, tc, = 0.21 s

were obtained using the MATLAB program 12 with the Power System Toolbox 13. Figure 2 is a plot of the generator angles with a clearing time of 0.19 s, which shows that the system does indeed remain stable. Figure 3 is the same contingency cleared at 0.21 s, where the generator angles are unstable. Next, tcr was calculated using the PEBS method. This method looks for the value of the machine angles at the point at which the dot productfr(O). (0 - 0 s) discussed in Section IV. 1 crosses from negative to positive. This dot product is plotted in Figure 4, and shows that the PEBS is crossed at t = 0.356 s. Once the PEBS crossing is found, the potential energy is found at the same point, and is defined as the critical energy of the system, V~r. The calculations of Vt, os and VMnG in equation (4) at this point are trivial. For the calculation of Vz~ one could use either the trapezoidal rule (TRAP) of equation (12) or the straight-line

approximation (SL) of equation (6) for the evaluation of the path-dependent integral. As the path of integration is available, it is more accurate to use the actual path in the evaluation using the trapezoidal rule rather than to assume a straight-line path. To examine how this stright-line approximation affects the PEBS method, VD was calculated both ways. For the *7-5 contingency, the SL method results in a Vet(O*)= Vcr= 1.2853. The TRAP method gave VpE(O* ) = Vcr ~- 1.3269. Figure 5 shows an example of the faulted energies on the system. Note that Vccr for the TRAP method is very close to the maximum of Vet in Figure 5. This is very typical when using the PEBS method. A quick way to estimate tcr from the graph is to draw a line from the peak of VeE parallel to the x-axis until it intersects VT"or. The time at this intersection is tcr. Using the SL results for Vcr, tcr =0.196s, whereas tcr--0.199s using the TRAP method. The results for this contingency were very similar for both methods. The loading of the system was increased to 150% of the nominal case, with all the excess load being taken by the slack bus, and the PEBS method was used again. The loading was then pushed to 200% of the nominal, again with the extra load taken by the slack bus and using the PEBS method. The results are shown in Table 1. In all cases the trapezoidal rule is more accurate than the straight-line approximation. The difference between the two methods becomes more pronounced as the system becomes more heavily loaded. At the highest loading the difference in V~r for the two methods is about 10%. The critical clearing time is also more accurate using the trapezoidal method, as expected, and the difference in tcr between the two methods is also greater with higher loading. The increase in tcr with increased loading is solely due to the fact that the extra load is taken up by the slack bus. The purpose here was only to show the difference between the TRAP and SL methods. If increased load is shared by the non-slack buses, the difference between the TRAP and SL methods still exists and as expected increased loading will decrease tc~.

2.5 . . . . . . . 2

........................ i. . . . . . . . . . . . . . . . . .
. . . . . . . . . . .

!.............. :

':

1.5 .................................. ~o i

: . . . . . . . . i .............. :
: ...... ! i

1 ........

......

-0.5 -1
: i ! I I

-1"50

0.05

0.1

0.15

0.2
Time(se, c)

0.25

0.3

0.35

..... t
0.4

Figure 4. Z e r o c r o s s i n g of t h e d o t p r o d u c t in t h e PEBS m e t h o d

Clarifications in transient energy function method." M. A. Pai et al.

69

. . . . . . . . . . . . .

................................................................... i ....................... :::>...

i
4 .................................................................................................... ~.................. ~ i
. . . . . . . . . . . . . . . . . .

i/'"
3

i
i ..................

e,. [.~

:(a)...:(~
i
/

..................

........................ : /" i

. . . .

................

.................

0.05

0.I

0.15

0.2 Time(scc)

0.25

0.3

0.35

0.4

Figure 5. (a) VTOT and (b) VpE using the trapezoidal rule for "7-5

T a b l e 1. C o m p a r i s o n

of SL and TRAP

using the PEBS

method,

to,

in s e c o n d s

System 1.0

loading

Straight-line approximation Vcr = 1 . 2 8 5 3 -- 0.196s Vcr = 1 . 9 7 1 9 = 0.272 S V~r ---- 2 . 5 9 6 3 = 0.360S

Trapezoidal Vcr = 1 . 3 2 6 9 = 0.199S

rule

Actual

(step by step)

tcr = 0 . 2 0 1 s

tcr
1.5

tcr

tcr
2.0

Vcr = 2 . 1 1 6 5 tc~ = 0 . 2 8 0 S Vcr = 2 . 9 5 9 2 = 0.377 S

tcr tcr

= 0.279s

= 0.387 S

tcr

tcr

1.2 1.15 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.1 ........................... ,............... ,. . . . . . .

~ i

~ ........................ i........................................ i i

i .......... :

1.05
O

Z
0.95 0.9
......................................................................................................................... ! ............

0.85 .....................................................................................................................................

0.8 0.5

0.6

0.7

0.8

0.9
Time(see)

I.I

1.2

1.3

1.4

1.5

Figure 6. Plot of the minimum gradient point function

70

Clarifications in transient energy function method." M. A. Pai et al.

1.4

1"6 t
/
,

.......

'
. . . . . . . . . . . . . . . . . . . .

,:
............. ~:',~ . ......... i................ ~............ ~ ................

......i i.::! ........ .... !."'" "'"-.iV ..... tt(.) !:


,::
......... : ..............

i
::
-

..... :
. ...............................

N, , VtotCo) "'-'~.

e~

0.8~

0.6t ........./..!../

.......................................................\ .....

....

0.42/...... ,~':! ....


0 0.2 0.4 0.6
Time

i i............ i ~i'ii "" "/ ........


0.8 1.2 1.4 1.6

(sec)

Figure 7. Vror and VpE for critically cleared system using (a) straight line approximation (b) trapezoidal rule
The BCU method was also tested for differences between the straight-line and trapezoidal evaluations of the energy function. The first step in the BCU method is identical to the first step in the PEBS method, in that the 0* is found. The next step is to integrate the post-fault gradient system equation (14) using 0* as the initial condition, until a minimum of ll f(0)ll is found. This is the minimum gradient point (MGP). Figure 6 shows the results of this integration for the *7-5 contingency at nominal loading. The potential energy is then evaluated at this M G P using the straight-line approximation (equation (15)). The trapezoidal rule is not an option here, because the exact path from 0 s to 0 u is not known Using Figure 6, the M G P was found at t = 1.384s, and using equation (16) the critical energy Vcr was calculated to be Vpe(Ou) = 1.3198 = V~r. As this is close to the values obtained by the T R A P and SL methods (Table 1), ter should be between 0.196 and 0.199s. The BCU method is the same as the PEBS method until 0* is reached. Before the fault is cleared there is not much difference in the energies computed by the SL and T R A P methods. However, large variations show up in the post-fault period. It is only when the SL approximation is used to calculate energies after tcr that large differences occur, lcr is of the order of 0.2 s and the PEBS crossing is closer to 0.4 s. In Figure 7, the difference in energies of the two methods is quite large after the fault is critically cleared In fact, the SL method shows that the total system energy increases after the fault is cleared, when in reality this is

1.4

1.2

0.8

o.6

0.4

o.21
O'

0.05

O.1

O.15

0.2 Time(see)

0.25

0.3

0.35

0.4

Figure 8. Potential energy of the line-cleared system

Clarifications in transient energy function method." M. A. Pai et al.

71

2.5

. .

".

. .

/ "

1.5

0.5

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

Figure 9. Potential energy of the self-cleared system


impossible. VToTmust decrease after the fault is cleared in a system that has damping. The trapezoidal method of computing Iq correctly reflects this. This could be of significance if the energy function is used for voltage dip calculations. The path integral terms initially are a small percentage of the potential energy but increase to about 30-40% near the PEBS crossing. However, large variations show in in the post-fault period.
V.1 The energy function at t = 0

Figure 8 is the *7-5 contingency, and Figure 9 is the *7 contingency. The only other way VpE could start at zero is if 0 is substituted for 0 s in the potential energy equations. This procedure is correct only when the PEBS method as discussed in Section III.2 is used.
V.2 Post-fault energy Another point of clarification concerns the total system energy after the fault has been cleared. In the literature, the post-fault energy of the system is always portrayed as a constant. However, in a system with damping, the energy must decrease over time. Note that in Figure 10, which portrays the three-machine system critically cleared from a *7-5 contingency, the energy definitely decreases over time. This obvious decrease in system energy is not so apparent in a larger system. Figure 11 is a plot of the energy function for the ten-machine system with a *30-34

In Section IlI.2, the initialization of VpE at t = 0 was discussed. In most energy function plots the value of VpE at the time when the fault is applied is shown to be typically zero. This is only true under specific circumstances, namely, when the post-fault system is identical to the pre-fault system, i.e., when the fault is self-clearing. Figures 8 and 9 show the difference between a fault that is switched out, resulting in a different system, and a fault that is self-cleared, resulting in the same system.

Vtot 1 ..................................... 5
i ......... ....................... ! ...................

0.8

~4
0.6

m3 0.4 2 0.2

o;

02

04

06

08

Time(sec)

12

14

16

18

0.2

0.4
Time

0.6 (sac)

0.8

Figure 10. Three-machine system critically cleared to


show decrease in system energy

Figure 11. Ten-machine system critically cleared to show almost constant system energy

72

Clarifications in transient energy function method." M. A. Pai et al.


comments. They also wish to thank the reviewers for their useful suggestions.

contingency. In the larger system, the post-fault energy does remain reasonably constant, because of the decreasing influence of the transfer conductances as the system increases in size. Note also that the while the fault is not self-clearing, the potential energy also starts very close to zero. As the system increases in size, the removal of one line makes less difference to the system energy. This point is, however, important if an energy function is to be used after dynamical equivalencing which results in fewer equivalent machines.

VIII. References
1 Athay, T, Podmore, R and Virmani, S 'A practical method for direct analysis of transient stability' IEEE Trans. Power Appar. Syst. Vol PAS-98 No 2 (1979) 573-584 2 Uemura, K, Matsuki, J, Yamada, I and Tsuji, T 'Approximation of an energy function in transient stability analysis of power systems' Electr. Wng. Jpn Vo192 No 6 (1972) 96100 3 Padiyar, K R and Ghosh, K K 'Direct stability evaluation of power systems with detailed generator models using structure preserving energy functions' Int. J. Electr. Power Energy Syst. Vol 11 No 1 (1989) 47-56 4 Pai, M A Power system stability N. Holland, Amsterdam (1981) 5 Pai, M A Energy function analysisforpower system stability Kluwer Academic Publishers, Boston, MA (1989) 6 Fouad, A A and Vittal, V Power system transient stability using the transient energy function method Prentice Hall, New York (1992) 7 Pavella, M and Murthy, P G Transient stability theory of power systems: from theory to practice John Wiley, New York (1993) 8 Fouad, A A and Stanton, S E 'Transient stability analysis of a multi-machine power system Parts I and II' IEEE Trans. Power Appar. Syst. Vol PAS-100 No 8 (1981) 3408-3424 9 Kakimoto, N, Ohsawa, Y and Hayashi, M 'Transient stability analysis of electric power systems via Lure type Lyapunov functions Parts I and II' Trans. lEE Jpn Vol 98 No 5/6 (1978) 10 Chiang, H D 'Analytical results on direct methods for power system transient stability analysis' Control Dyn. Syst. (ed. C T Leondes) Vol. 43 Part 3 Academic Press (1991) pp 185274 11 Anderson, P M and Fouad, A A Power system control and stability Iowa State University Press (1977) 12 Little, J N MATLAB: User's Guide The Mathworks Inc. (1991) 13 Chow, J H and Cheung, K W 'A tool box for power system dynamics and control engineering education and research' IEEE Trans. Power Syst. Vol 7 No 4 (1992) 1559-1564 14 Chiang, H D, Wa, F F and Varaiya, P P 'A BCU method for direct analysis of power system transient stability' IEEE Trans. Power Syst. Vol 9 No 3 (1994) 1194-1208

VI. Conclusion
In this paper several points of the transient energy function method have been clarified. The first is a discussion on the validity of the straight-line approximation for computing the path dependent integral. It was shown that it is more accurate to use the trapezoidal rule whenever possible. At higher loadings the T R A P method is more accurate than the SL method. The trapezoidal method also exhibits much better results when calculating the post-fault potential energy. This could be very important in voltage dip calculations. In the BCU method after the controlling UEP is computed, a straight-line approximation is necessary to compute Vcr. Another clarification concerns the potential energy t = 0. The point was made that the only way the potential energy could be zero at the start of the fault is if the preand post-fault systems are identical. This is not the case if any line switching occurs; hence, this should be reflected in the computation of the energies. If the maximum of the potential energy referenced to 0 is used to compute gcr , then there is no need to compute 0 s in the PEBS method. The third clarification is regarding the post-fault system energy. Although often it is shown as constant, that is not necessarily the case. In a small system with damping, it is easy to see that the energy decreases with time. With larger systems, because of the decreased effect of the transfer conductances, not only do the energies appear to start from zero, they are almost constant or slightly decreasing after the fault is cleared.

VII. A c k n o w l e d g e m e n t s The authors would like to thank the National Science Foundation for its support through its grant NSF ECS 91-19428. They also wish to thank Professor K. R. Padiyar of I.I.Sc., Bangalore, India for his useful

You might also like