299

PROCEEDINGS OF THE IEEE, VOL. 70, NO. 3, MARCH 1982

(1) with respect t o x. Thus

REFERENCES

[ 1 ) T. Koga, Multivariable bounded-reality criterion,

Rep. PIBMRI-

1381-67, Polytechnic Institute of Brooklyn, 1967.

[ I ] L. V. Ahlfors, Complex analysis. New York: McGraw-Hill,
1966.
[ 31 T. S. Huang, Stability of two-dimensional recursive filters,
ZEEE Trans. Audio Electroamat., vol. AU-20, pp. 158-163,
1972.
[ 4 ] Ph. Delsarte, Y. Genin,and Y. Kamp, An equivalence between
bounded multivariable functions anda class of bounded singlevariable functions,ZnZ. J. ConZrol, vol. 34, pp. 383-389, 1981.

Here, xrepresents any one of the parameters,radius R, spacing D,

height H,of the conductor above the earth, frequency f, but not the
length 1. Here VO and IO are assumed constant and known. The sensitivity of any eigenvector with respect to x is

The sensitivity matrix aZo/ax is obtained from

Sensitivity Analysis of a Multiconductor Transmission Line

C. S. INDULKAR, PARMOD KUMAR,

AND

D. P. KOTHARI

A b m c t - I n this paper, the theory for the sensitivity analysis of a

multiconductor line is d e s c r i i , and the sensitivities of the receivingend voltage and power with respect to the various line parameters
have been evatuated.

Thus, the normalised sensitivity of the receiving-end voltage is given

by

1. PARAMETER
SENSITIVITY
MODEL
The multiconductortransmission line equations are [ 11

- Zo Sinh ($Z)tZo
I1 = Cosh ($Z)tIo - Sinh ($2),YOVo

Vl= Cosh ( $ 2 ) Vo

(1)
(2)

where VI is a column vector representing the receivingend voltages and

Vo and Io are column vectorsrepresenting the sendingend voltages
and currents, respectively. l represents the line length. If Z and Y are
the series-impedance and shunt admittance matrices of the multiconductor line, then the propagation matrix ($I) can be evaluated from the
eigenvalues and eigenvectors of ZY as follows:

QZ = E
, Yn = Q , K ~ Q - I

(3)

where h and Q are, respectively, the eigenvalue and eigenvector matrices.

The characteristic impedance matrix (ZO) is evaluated from the propagation matrix as follow:
z o = $-z.
(4)
The sensitivity matrix az/ax of the series impedance with respect to
a generic parameter x is composed of the partialaxderivatives of the individual elements of Z matrix with respect to x. The sensitivity matrix
aY/ax is obtained from the sensitivity matrix aPc/ax of the potential
coefficient matrix PCas follows:

avll

av21

aVNl

(1 1)

Since the receiving-end complex power SI is equal to {(Vl)tZf}, its

sensitivity with respect to a parameter x is given by

where

aI; = a cosh ($Z);

ax

ax

ax

asinh ($Z);

yo*6

Io -

- Sinh ($2);- aaxYo* V,*. (13)

The sensitivity matrix of cosh ($Z)?
tuting h* instead of A in it.

a cosh ($Z)t as
ax

is determined from (14) by substi-

= -Cosh (h1121) S- + S

a Cosh (kI2z)
ax

S-

as-

+ S cosh (h1I2Z)-

ax

(14)

e;.

Once aZ/ax and a Y/ax are obtained, the sensitivity matrix of P, i.e.,
ZY is obtained from

where S =
The sensitivity matrix of sinh ( $ 2 ) ; is obtained from (14) by substituting sinh ( $ 2 ) in place of cosh ( $ 2 ) and h* for A. Further,

Since SI = Pl +jQ1, the normalised sensitivities of the active and reactive

powers at thereceiving end are
This sensitivity matrix is used to determine the sensitivity of the eigenValues hk Of P as follows [ 21 :

where qk and S k are the eigenvectors, respectively, of P and Pt, associatedwith the eigenvalue hk,andthe
asteriskindicates the scalar
product of two vectors.
11. SENSITIVITY
OF RECEIVING-END QUANTITIES

The sensitivity vectoraVz/axof

the receivingend voltage VI with

respect to a parameter x of the system is obtained by differentiating
Manuscript received July 16, 1981.
The authors are withthe
Department of Electrical Engineering,
Indian Institute of Technology, New Delhi, 110016, India.

A. Sensitivity with Respect to Line Length

The required expressions for determining the normalised Sensitivity
of the receiving-end voltage and current with respect to the line length
are
=! ! !

az

Sinh ( $ 2 ) $Vo - Zo Cosh ( $ l ) t $ t I o

_
a
-- Sinh ($~);$;IO
- Cosh ( $ 2 ) ? $ ?
az

0018-9219/82/0300-0299$00.75
0 1982 IEEE

YtVo.

(18)
(19)

PROCEEDINGS OF THE IEEE, VOL. 70, NO. 3, MARCH 1982

300

TABLE I
SENSITIVITYVALUES FOR A TRANSMISSION
LINE

I
III.

APPLICATION TO A SAMPLE SYSTEM

of the theory, a220 KV horizontal line with sending-end active and reactive powers of 150 MW
and 50 W A R with line length 300 Km is considered. The spacing D ,
height H,radius R , and resistance R of the conductor are assumed to
be 6.096 m, 7.62 m, 1.27 cm and0.93 X lo4 O/m, respectively.
Table I summarises the results obtained at 60Hz.
The reactive power at the receiving end decreases due to increase of
spacing because of the increase of line reactance and decrease of the
shunt susceptance. However, the voltage magnitude and therefore the
regulation increases with increase of spacing. The height of the conductor and its radius influence the reactive power only. Summarising,
both receiving-end voltage and power are most sensitive to length and
frequency and least to height. Of all the quantities, the reactive power is
affected the most with the change in the values of the parameters.

In order to demonstrate the utility

REFERENCES
[ 1] L. M. Wedepohl, "Application of matrix methods to the solution
of traveling-wave phenomena in polyphase systems," Proc. Inst.
Elec. Eng.,vol. 110, pp. 2200-2220, Dec. 1963.
( 2 1 S. Cristina and M. D.'Amore, "Propagation on polyphase lossy
power lines-A new parameter sensitivity model," IEEE Trans.
Power App. Syst., vol. PAS-98, pp. 33-44, Jan./Feb. 1979.

Fig. 1. Block diagram representation of an nth-order system.

where h 1 , kl , h 2 , . . are scalar quotients. The minimal realization thus
obtained is then used to derive the reduced-order models. The relation
connecting the state variables of the original system with those of the
model is also developed, which makes it possible the use of such models
for deriving suboptimal control policies [6].

11. MODELREDUCTION PROCEDURE

Consider a linear time-invariant system represented by

g(s) is then expanded in the modified Cauer form by constructing the

following table:

MODIFIEDROUTHTABLE

State-SpaceModels Using Modified Cauer Continued Fraction

R.PARTHASARATHY

AND

SARASU JOHN

.. .. .. ..

A6muct-This letter presents a procedure for obtaining reducedader

models in statespaceformulation, using m o d i f d Cauer continued
fraction as the basis.
The proposedprocedure is amenable to programming on a digital
computer.

I. h R O D U C T l O Y
In the past decade, the problem of deriving reduced order models
for high-order plants has received considerable attention of research
workers [6]. In [ 11, Shieh and Goldmanpresented a procedure for
obtaininga minimal realizationemploying the first,second, and the
mixed Cauer forms. In this letter, we employinstead the following
modified Cauer form [ 21 :

Thedenominatorandnumerator
coefficients of (1) form the fust
two rows. The successive rows are evaluated using the relations [ 31 :
aj+l,k=aj,k+l -hjbi,k+l

and

1
g(s) =

hl

where

+'
1

kl

and
k . = bj,n+l - j
,
a j + l , n + l -i

Manuscript received May 28, 1981;revised November 18, 1981.

R. Parthasarathy is with the Department of Electrical Engineering,
Indian Institute of Technology, Madras, India.
S. John is with the Department of Electrical Engineering, Karnataka
Regional Engineering College, Sreenivasa Nagar, Mangalore, India.

j=1,2:..,n.

A block diagram interpretation of this form of continued fraction is

given in Fig. 1. From the figure arealization for (1) can be directly
written as

0018-9219/82/0300-0300$00.75 0 1982 IEEE

Lj=Hv+Lu y = D p u

(3)