IEEE TRANSACTIONS ON POWER APPARATUS AND SYSTEMS, VOL. PAS-88, NO.

1, JANUARY 1969

Structure in the Computation of

Power-System
APR 22 1969
TECHNICAL LIBRAP

Nonlinear Dynamical Response

assumption: that all loads have a consta; QMMI|S voltage-frequency characteristic. The closed-form integration of the system differential equations requires that the combined algebraic and differential equations of the system be reduced to the form
X1
X2 =

fk(Xl,X2,

...

,xk)
,Xk)
(1)

= f2(xi,x2,

* *

INTRODUCTION T HIS PAPER presents a general analytical solution to the problem of simulating power-system nonlinear dynamical performance when both transient and subtransient effects are modeled in the synchronous machines. The general solution is then examined to reveal a unified computational structure that allows a useful insight into the relationship between the nonlinearities, such as dynamic saliency in the synchronous machines, and the alternative impedance and admittance formulations of the transmission network equations. The state space approach commonly found in modern control theory is used throughout the paper, and Appendix II gives a derivation of the synchronous machine transient and subtransient reactances from a state space point-of-view as an alternative to the usual approach based on operational impedances.
STATEMENT OF THE PROBLEM The simulation of power-system dynamical response involves the simultaneous solution of three mathematical problems: 1) the solution, either actually or implicitly, of the linear voltage-current equations of the transmission network 2) the integration of the nonlinear differential equations of the synchronous machines and control loops 3) the satisfaction of the nonlinear boundary conditions imposed by loads having nonconstant impedance characteristics relating voltage, current, and frequency. Many computer programs presently in use include iterative techniques for the solution of each of these problems [1]; while others [2], [3] use a closed-form matrix technique for the first but resort to an iterative equivalent phasor diagram approach to handle the synchronous machines. One program has been described [41 that treats the differential equations as such but still resorts to an iteration to couple the network and machine

where fi,f2, * *,f are closed-form expressions or algorithms giving the instantaneous time derivatives of the state variables Xl,X2, * **, xk in terms of the instantaneous values of these state variables only. Thus the core of the problem is that of arranging the network equations and the nonlinear synchronous machine equations into a form that can yield such closed-form derivative computation algorithms.
ALGORITHMS FOR COMPUTING TIME DERIVATIVES Elemental System Equations The transmission network is described by the partitioned complex matrix equation

VLt)L
or by its inverse

rN1

=-ZNN ZNLL LiN

LZLN ZNL JLiL

(2)

[iN1 i

L YLN

YLL JLVLj

YNL IFvN]

(3)

Each of these sets of equations may be expanded to a set of twice the number of real equations of the form
[VD1

.Q

R[

-Xi, T Xr RR,R X

iD1
tQ1

(4)
VDr
L-Qr

equations. Our aim here is to develop algorithms that will incorporate closed-form solutions wherever possible and minimize the number of iterative loops. It will be shown that it is possible to achieve a completely closed-form simulation algorithm by making only one

Rrl -X71| Xr, Rrl.

:Rrr -Xrr
Xrr

Rrr_, iQr

iDr

Paper 68 TP 666-PWR, recommended and approved by the Power System Engineering Committee of the IEEE Power Group for presentation at the IEEE Summer Power Meeting, Chicago, Ill., June 23-28, 1968. Manuscript submitted February 7, 1968; made available for printing April 10, 1968. The author is with the General Electric Company, Schenectady, N. Y.

where VD, VQ, iD, iQ are the projections of the voltage and current vectors onto a set of synchronously rotating reference axes (D,Q). The equations of each machine refer to axes (d,,q7) rotating in synchronism with the rotor of the machine, and so the currents and voltages of the machines are related to the network by the transformations
[VDr
VQj
J

[cos
sin

-sin

FVdil

Si

Cos

[jVgJ
J

(5)

LiqJ

Fidtl

L-sin3i

cos Si sin ai]

FiDil
LiQi J

(5a)

IEEE TRANSACTIONS ON POWER APPARATUS AND SYSTEMS, JANUARY 1969

OF THE

TABLE I NETWORK SOLUTION PROBLEM

IN

POWER-SYSTEM DYNAMICAL SIMULATION

Assumptions
General case

Admittance Impedance Formulation Formulation ZNLiL + [ZNN + T(R + PQX,)Tt]iN = TKWr - YNLVL + [1 + YNNT(R + PQ X")Tt]iy

Transient saliency included, [ZNN + T(R + PQ X")T']iN = TKTr [1 + YNNT(R + PQ X")T']iN = YNNTKPT nonlinear loads neglected (C) (D) Transient saliency neglected, ZNLiL + [ZNN + (R + PQ X")]iN = TK'Pr - YNLVL + [1 + YNN(R + PQ X")]iN = YNNTKW, nonlinear loads incluLded (E real) (F real) ZNLiL + [RNN + R +j(XNN + X")]iN = TKWr -YNLVL + [1 + (GNN + JBNN)(R + jX")]iN = YNNTKer (F complex) (E complex) Transient saliency neglected, [ZNN + (R + PQ X")]iA = TKWr [1 + YNN(R + PQ X")]iN = YNNTKI, nonlinear loads neglected (G real) (H real) [1 + (GNN + jBNN)(R + jX")]iN = YNNTKW, [RNN + R + J(XNN + X")]iN = TKWT (H complex) (G complex)

~~~~~

YNNTKFr
(B)

~~(A)

These transformations may also be written in the complex form

,n.kql )fdifrf,2kdl

**

(VDi + jVQi) (idi + Avj)

(COS 4i + j sin at) (Vdi + jVqj)

*kdji

fkqj

Efdt

bit

a'

yt 2

(51b)

for i= 1,22

n.

vN = Tvm

- = T tiN

(6)

General Solution The direct and quadrature axis currents and voltages appearing in (2)-(5) are not state variables, and so they must be eliminated from the final equations used to calculate the derivatives of the flux linkages. Eliminating Wm. and iri from (7)-(9) gives
Vn= PQ XMi XRi-l'Wi
- [Ri + PQ(XSi -

where T is either a real diagonal matrix of 2 X 2 submatrices or a complex diagonal matrix, depending on whether the real or complex form of the network equations is used. No notational distinction is made between the real and complex forms of the network equations and transformations since the form in use is obvious from the context of the paper. Each synchronous machine is described by Park's equations for a machine with an arbitrary number of rotor windings
Vm1 = PQWm - Rim

XMiXRi-'XMi1Aim1-

(16)

(7)
(8)

Wmin =

- XS im1

r= XRiir-

XM1i im

(9)
(10)

RF1pWr'
pas

=
=
=

ir' + (l/XafdOi)Er?
woai

Now it is shown in Appendix II that the matrix (XS - XM XR1 XM') is a generalized characteristic of a synchronous machine, which in the special case of no rotor circuits (except the field) is a diagonal matrix of the transient reactances, and which is a diagonal matrix of the subtransient reactances in the case with rotor circuits in addition to the field. We shall henceforth refer to this matrix as the dynamic reactance matrix since it is apparent from (16) that it applies under all conditions and not only under the so-called transient and subtransient conditions. Using the dynamic reactance matrix as derived in Appendix II allows (16) to be written as

(11)

vmi

PQ XMi XRi-1 'rF - (Ri + PQ Xil)ini.

(17)
(18)

Tmipai

Ti

WmUQPim'

(12)
(13)

where Ki = PQ XMi XRi-1. fori = 1,2,-2 ,n. We can now drop the machine number i and write the set of The matrices appearing in these equations are given in Appen- n equations (18) as the single equation (19) dix I for the case of a machine with j rotor windings on each axis and a field winding on the direct axis. (19) VN = T K 'r- T(R + PQ X")TtiN The governor and voltage regulator will not be considered in detail in this paper, but it is noted that they are described by where each of the matrices is now a diagonal matrix of the submatrices appearing in (18). Equation (19) may be substituted equations of the form directly into the appropriate rows of (2) or (3) to yield the two G11 pz1 = hi(a,zt) (14) general solution equations (A) and (B) of Table I. Next eliminate im1 and 0rt from (5), (9), and (10) to yield for the governor, and equations for the derivatives of the synchronous machine flux (15) linkages. G2ipEr1 + G31 py' = gi(Er' a, pa, yi) PWT = RFi-'(XRi-l XMItTjtiN? + XRi- Wr') + 1 Er . for the voltage regulator, where i = 1,2, * *,n. Xafdoi Examination of (7)-(15) shows that the state variables of the system are (20)
-

Vti = Vm iVm

Then using the real matrix transformation (5) gives VNt = T KiI- T1(Ri + PQ Xi")TitiNi

UNDRILL: POWER-SYSTEM NONLINEAR DYNAMICAL RESPONSE

SET UP INITIAL CONDITIONS CONSTRUCT CONSTANT MATRICES

rI

= _.. ALL . ., JA VARIABLE CONSTRUCT AND CONSTANT MATRICES

However, at each integration step the right-hand sides and all coefficient matrices of (A) and (B) can be computed in a closedform manner before the network solution is commenced. Therefore, after the coefficient matrices of (A) or (B) have been computed, it is possible to construct [from (A) and (2) or (B) and (3) ] network solution equations
S
[[ZNN

SOLVE APPROPRIATE EQUATION FROM TABLE I FOR 1N

+ T(R + PQ

X")T'I

ZNL1FiN1

FTKWr1

L
or

ZLN

ZNNJLiLJ

VL J

(22)

CALCULATE GENERATOR FLUX LINKAGE DERIVATIVES FROM EQUATION (20)

RECONSTRUCT
VARI ABLE

[1 + YNNT(R + PQ X")Tt] OliN

liL]

MATRIC ES

[YNN

1TK

r1

(23)

CALCULATE ARMATURE EQUATION (21). CALCULATE ALL REMAINING DERIVATIVES FROM EQUATIONS (11)-(15)
FLUX LINKAGES FROM

I NTEGRATION

NUMERICAL

STEP

NEW SYSTEM

Fig. 1. Flow chart showing basic steps dynamical response simu lation.

in the power-system

from the value of the previous solution and hence require at least two computations to detect convergence. Since the com-

whose coefficient matrices are constant for the integration step. These equations can be solved by a nodal iterative technique which determines the required values of the load bus voltages VL to satisfy the nonlinear boundary conditions. Note that only one level of iteration is required because it is not necessary to reiterate the synchronous machine solutions after each iteration of the network solution. Even though convergence of the iterative solution of (22) or (23) would usually be fast because of the small changes in network condition from step to step of the simulation, each solution would normally involve at least two passes through the computation of iv from (22) or (23). This is because any significant change in the network conditions will change the value of iN putation of iN is the only task occurring in each derivative
evaluation which involves
a

Finally, eliminating ini and i,' from (5), (E3), and (9) yields Wmi = XMiXRi-Wri Xi"TitiN'.

large nondiagonal matrix, the avoid-

(21)

ance of the iterative solution offers a significant improvement in computational efficiency. Further, avoiding the need to re-

Equations (A) or (B), (20), and (21) co institute the crux of the construct the variable matrix product [ZNN + T(R + PQ X")TtJ algorithm for computing the state varialble derivatives, because or [1 + YNNT(R + PQ X")T'] at each step of the simulation once (A) or (B) has been solved the mae ~hine derivatives follow offers still more improvements in computational speed. directly from (20). Equation (21) then grives the armature flux linkages which are in turn the required in]put data for computing Simplification of the General Solution the derivatives of the state variables of the rotational and exThe principal simplification of the general is to replace the citation systems. nonlinear boundary conditions at each load bus by a constant The flow diagram for the dynamic re sponse simulation pro- impedance load. The load buses then be eliminated from the cedure using this general solution algoritl im is given in Fig. 1. network equations, causing the vectors VL and iL to disappear The computation of the state variable dlerivatives is a straight- from (A) and (B). This leaves the two simpler equations (C) and forward closed-form process except for tlhe solution of the alge (D) given in Table I. These equations be solved directly braic equation (A) or (B). Therefore, deetailed consideration of in closed fornm by constructing the left-hand side coefficient the derivative computation algorithms is concerned mainly matrix and solving the resultant set of simultaneous linear with the properties of (A) or (B). tions. The resulting computation of the derivatives of the machine flux linkages now involves no iterative loops, and yet still treats both transient and subtransient effects within the synSolution of the Network Problem The transformation matrix T is an e xplicit function of the chronous machines. The penalty for the accurate modeling of the synchronous state variables 61, 52, * * * ,b, and hence the machines including dynamic saliency is the need to reconstruct inspectivel. But, (, the matrix and (B) are the vectors iL and iN or VL anLd iN, respectively. since VL and iL are not state variables, thi,ey are unknown at the [Z + T(R + PQX")Tt] start of each derivative computation. T'herefore, (A) and (B) anvia+;nnc ThobZ have more unknowns than independent siansrata vLne epuJruue equation. required additional equations are the ro'ws of (2) or (3) corre- or sponding to load buses, together with tlhe nonlinear boundary [1 + YNNT(R + PQ X")TI] conditions on iL and VL, and hence a closEad-form solution of (A) or (B) is not normally possible. at each step of the numerical integration.
case
can

as

can

equa-

oly

IEEE TRANSACTIONS ON POWER APPARATUS AND SYSTEMS, JANUARY 1969

Now observe that the matrix (R + PQ X") is a diagonal matrix of submatrices of the form

that

RX Multiplying out the matrix product [T(R + PQ X")T'] shows

rRi

xqi[ t

Xdi"

allows (A)-(D) to be written in the simplified forrns (E)-(H) shown in Table I. Further, the 2 X 2 submatrices of the matrix expressions

if Xd i # Xqi" (24) T(R + PQX")Tt = (R + PQ X"), if Xdi"l = xqiJ" (25) Neglecting dynamic saliency and setting Xdi" = xq/", therefore,

T(R + PQ X")Tt

f(612, *An),

diagonal of the main matrix. As a result the sets of 2n equations (A)-(D) are not generally reducible to a set of n complex equations. Setting xdi" = xgi", however, returns the on-diagonal submatrices to the skew symmetric form and allows (E)-(H) to be reduced to the alternative complex forms which are also shown
in Table I. Thus it is implied that the synchronous machines may be regarded as having dynamic impedances (R + jX") which are extensions of the transmission network in the case without dynamic saliency. This is not so when dynamic saliency is considered because the salient dynamic impedance matrices (R + PQ X") do not allow the sets of 2n real equations (A)-(D) to be reduced to the complex form of an equivalent ac network. Equations (E) and (F) must be solved by nodal iterative techniques in the same manner as (A) and (B). Equations (G) and (H) clearly allow the generator currents to be calculated directly without iteration.

[ZNN + T(R + PQ X') Tt] and [1 + YNNT(R + PQ X")T'] are skew symmetric everywhere except when they are on the

The basic computational structure applies equally well when no amortisseur circuits are represented, corresponding to treatment of transient effects only, and when an arbitrary number of amortisseur circuits are represented in order to treat subtransient effects. It is also independent of whether the impedance or admittancerformulation is chosen for the transmission network equations. The introduction of simplifying assumptions regarding the loads and the dynamic reactance matrices allows simplification of the detailed structure of the equations involved in 2) but does not change the basic three-step computational structure. The general case using (A) or (B) in 2) requires the reconstruction of variable matrices and an iterative solution at each integration step. The case without nonlinear loads requires no iterative solution but still requires the left-hand side matrix of (C) or (D) to be reconstructed at each integration step. Introducing the assumption that dynamic saliency is negligible allows the sets of 2n real equations (A)-(D) to be rewritten in the form of n complex equations (E)-(H). It is still necessary to reconstruct the matrices in these equations if the dynamic reactance matrix is not constant, but the assumption of a constant dynamic reactance makes all matrices in (E)-(H) constant. Thus when dynamic saliency and dynamic reactance variations are neglected, the matrices need to be constructed only once at the start of each simulation. The assumption of a constant nonsalient dynamic reactance matrix at the transient level would not always be acceptable; this assumption appears to be quite reasonable at the subtransient level where one or more amortisseurs are represented.

MAGNETIC SATURATION The effect of magnetic saturation is to change the values of certain elements of the matrices XM, XR, and XS at each integration step. This causes no basic change in the computational structure indicated by Table I as long as saturation effects are represented as functions of the rotor flux linkages, which are state variables. It does, however, necessitate the introduction of an extra iterative loop if saturation effects are to be represented as functions of quantities which are not state variables. COMPUTATIONAL STRUCTURE The basic characteristic of the computational structure presented here is the separation of the state variable derivative calculation into three distinct steps: 1) calculation of the instantaneous apparent network excitation vector TKWr. 2) solution of the instantaneous apparent network equations as summarized in Table I. 3) calculation of derivatives using the results of 2). These three steps are noninteracting as long as the dynamic reactance matrices of the synchronous machines are functions of state variables only. The significance of this noninteracting computational structure is that it eliminates the need for iterative subloops in the synchronous machine modeling and in coupling the machine models to the network model. This means that the only iterative loop involved at each integration step of a dynamic response simulation is the nodal iteration needed to handle the nonlinear loads.

CONCLUSIONS This paper has presented a general approach to the simulation of multimachine power-system dynamics with detailed treatment of subtransient flux linkages and subtransient saliency in the synchronous machines, and has described a set of simplifications which all exhibit a certain structure in common with the general approach but which offer computational economies. Emphasis has been placed on minimizing the number of iterative loops involved; it has been shown that the only essential iterative loop is that needed to handle nonlinear loads. Appendix II gives a derivation of the transient and subtransient reactances which is independent of the concept of operational impedances and demonstrates that the structure of the synchronous machine analysis within the multimachine problem is independent of the number of rotor circuits considered. The use of a unified state space approach has allowed a simple and comprehensive analysis of the power-system dynamics problem and gives an insight which is a valuable alternative to the well-known approach using operational calculus.
APPENDIX I MATRICES IN ELEMENTAL SYSTEM EQUATIONS cos 3, -sin Al_ sin a1 cos An -sin An cos 61 T L

sin An
PQ =[

COS bnz

UNDRILL: POWER-SYSTEM NONLINEAR DYNAMICAL

RESPONSE

i/wOrfd
l/worAl

where

l/workql
l/workdj

Xd,

Xafd2Xkkcd
Xakq2

RF =
Xkkq

XffdXkkd

2XafdXfkdXakd
-

Xf kd2

Xakd2Xffd

1/wOrk01

XM

[Xa fd Xakdl 0
0
Xffd
Xf kdl
Xfkdl
Xkkdl Xkd2l

... Xakd
..

0 X

Gq1

Xakcql ... Xakcfij

Yet again, these expressions are exactly those given in [5], [6] for the "subtransient reactances." Hence, it may be generalized in the case of a machine with an arbitrary number of amortisseur circuits on each rotor axis that a set of characteristic reactances appearing at the generator terminals is given by the expression
X"
=

*
*

Xfkdj * .. Xkdl j
.

XS - XM XR-1 XMt = [

xq"]

Xkd2j

XR =

Xfkdj

Xkcdjl
0

XkkAdj
Xkkql
*''
. .

XAqlI
Xkq2j

Xkq2l .

X,tqjl
XS

.
0

XA;kqj-

These direct and quadrature axis reactances as given by the expression (XS - XM XR-1 XM') are referred to as subtransient reactances wherever the number of amortisseur circuits j is greater than or equal to one. However, since the same expression yields the transient reactances in the case where there are no amortisseur windings, we shall refer to the matrix X" as the "dynamic reactance matrix" of the system. The phenomenon of unequal numerical values of the diagonal elements of X" is referred to as dynamic saliency in the general case. In the particular cases of j = 0 and j > 1 we refer, respectively, to transient dynamic saliency and subtransient dynamic saliency.
'

LIST OF SYMBOLS state variables of a k-dimensional system X1, * * Xk vector of generator and load terminal voltVN,VL ]. QP= ages referred to network axes rotating at L-1 o0 synchronous speed ZNN,ZNL,ZLN,ZLL partitioned parts of the network nodal imAPPENDIX II pedance matrix partitioned parts of the network nodal adYNL, YLN, YLL YNN, AND MATRIX DYNAMIC SALIENCY DYNAMIC REACTANCE mittance matrix Consider the case when the amortisseur circuits are neglected. VD,VQ,iD,iQ direct and quadrature axis components of In this case (8) and (9) become, for each machine: generator voltage and current referred to network axes [Xd i[d] Xa [ifd] [d direct and quadrature axis components of Vd,Vq,li,iq generator voltage and current referred to generator axes *fd = lXffd70I [i [Xafd ] [.i generator rotor angle = [Td ntQ ] 2-vector of generator d- and q-axis armature IF. and the matrix expression flux linkages = [Tfd,Tkd,k7cq I vector of generator field, d-axis amortisseur, Ir (XS -XMXR XM') and q-axis amortisseur winding flux linkages vectors of currents corresponding to Pmx, Wr im, multiplies out to give 2-vector of generator terminal voltages Vm [Vd,Vq] corresponding to Wm [Xd - Xafd2/Xffd 01 (XS - XM XR-1 XMt) vector containing generator open-circuit Er = [Efd,010] field voltage generator rotor speed in per unit of rated But it will be recalled from [5] that (xd - Xafd2/Xffd) and Xq are a speed just the expressions for the apparent direct and quadrature axis synchronous speed, radians per second reactances at the machine terminals at the instant of application wo Tm generator inertial constant, seconds of a disturbance, namely the "transient reactances." shaft torque input to generator, per unit Again when a single amortisseur winding is included on each Tl vector of state variables describing the prime circuit, the matrices appearing in (8) and (9) are those given z mover system of a generator in Appendix I, and multiplying out the matrix product shows vector of state variables describing the exthat y citation system of a generator unsaturated value of generator d-axis armaXafdo XS-XM XR-1 XM =[0 0] ture to field winding mutual reactance

[ Xd

L0 xq

[Xad

lr

IEEE TRANSACTIONS ON POWER APPARATUS AND SYSTEMS, VOL. PAS-88, NO. 1, JANUARY 1969

Xafd, Xffd, etc.

rfd, rkd, rk,

mutual coupling and self reactances of synchronous machine windings generator armature resistance resistances of synchronous machine field winding, d-axis armature winding, and q-axis armature winding number of rotor circuits represented on the d and q axes of each synchronous machine.

ACKNOWLEDGMENT The author acknowledges the encouragement and helpful comments of Dr. L. K. Kirchmayer and DD. N. Ewart of the General Electric Company.

REFERENCES [1] M. S. Dyrkacz, C. C. Young, and F. J. Maginniss, "A digital transient stability program including the effects of regulator, exciter, and governor response," AIEE Trans. (Power Apparatus and Systems), vol. 79, pp. 1245-1257, 1960 (February 1961 sec.). [2] H. E. Brown, H. H. Happ, C. E. Person, and C. C. Young, "Transient stability solution by an impedance matrix method," IEEE Trans. Power Apparatus and Systems, vol. PAS-84, pp. 1204-1214, December 1965. [3] J. E. Day and K. C. Parton, "Generalized computer program for power systems analysis," Proc. IEE (London), vol. 112, pp. 2261-2274, December 1965. [4] K. Prabhashankar and W. Janischewskyj, "Digital simulation of multimachine power systems for stability studies," IEEE Trans. Power Apparatus and Systems, vol. PAS-87, pp. 73-80, January 1968. [5] G. Kron, Tensors for Circuits. New York: Dover, 1959. [6] C. Concordia, Synchronous Machines. New York: Wiley, 1951

Security

Planning and Operation European Power Systems Part I

Supply in

of

PETER W. CASH AND EDWARD C. SCOTT

Abstract-The results of an inquiry into the security standards and practices adopted in the operation of power systems in 16 European countries are presented. A complementary paper gives similar results for standards and practices in planning [1]. The inquiry has revealed that differences in practice in system operation are comparatively small and appear to derive from the circumstances of geography, type of plant, etc., rather than from any basic difference in approach. INTRODUCTION
1 increasingly dependent on an adequate supply of electric-

ALL ASPECTS of life in industrial countries

are

becoming

This paper is in five sections: 1) General Discussion: This is a brief statement of the authors' viewvs on the security considerations that arise in system planning and operation. 2) Summary of Questionnaire: A questionnaire circulated in September, 1963, has been followed by supplementary questionnaires and correspondence. The replies form the basis for sections 3 and 4 of this paper and sections 1 and 2 of the companion paper [1]. 3) Broad Statistics of the Countries. 4) Security in Operation. 5) Summary of Security Practice in Operation.

ity. Whether the supply should have an expected continuity of say 99.9 or 99.99 percent is, in the present state-of-the-art, largely a matter for the judgment of the supply engineers. The Study Committee on Large Systems and International Interconnections of the International Union of Producers and Distributers of Electrical Energy (Unipede) decided therefore in 1962 to mount an inquiry into security practices in different European countries. The results are reported in this and a companion paper [1].
Paper 68 TP 113-PWR, recommended and approved by the Power System Engineering Committee of the IEEE Power Group for presentation at the IEEE Winter Power Meeting, New York, N. Y., January 28-February 2, 1968. Manuiscript submitted October 9, 1967; made available for printing April 17, 1968. The authors, now retired, were with the Central Electricity Gen-

erating Board, London, England.

GENERAL DISCUSSION The system-planning engineer endeavors to arrange the future equipment of the system so that a supply of electricity of the desired degree of continuity and constancy of voltage and frequency will be provided at minimum practicable total capital and operating cost. The system-operation engineer endeavors to arrange the use of the equipment that has been installed so as to provide such a supply at minimum operating cost. The two functions are similar. The planning engineer has to choose the new genierating and transmission plant to be built; the operation engineer has to choose the generators to be run, their outputs, and the way in which the transmission system will be switched. In both cases the main tasks are to forecast the power and energy requirements, to calculate the amount and geographical disposition of the generating plant needed to meet those demands, and to arrange reliable transmission facilities to the