Z-transform electromagnetic transient analysis in

power systems
Prof. W.D. Humpage, B.Sc, Ph.D., K.P. Wong, M.Sc, Ph.D., T.T. Nguyen, B.E. and D. Sutanto, B.E.

Indexing terms: Power transmission, Transforms

Abstract: The paper develops a new electromagnetic transient formulation for multiconductor power trans-
mission lines based on the z-transform. Beginning with the basic relationships of propagation in the frequency
domain, the successive steps of z-transform analysis are developed leading to recursive sequences for solution
in the time domain. The case of lossless propagation is considered initially in introducing z-transform analysis
in its application to transmission-line electromagnetic transient evaluations, following which a general formu-
lation is derived into which series-path line parameter frequency-dependence is directly reflected. Test analy-
ses based on a z-transform model complemented by a simplified equivalent-source representation and applied
to a 160 km, 400 kV untransposed single-circuit transmission line confirm the high inherent accuracy and low
computing-time requirements of the method.

List of Principal symbols when it is used in an intermediate transform step between

the frequency domain and the time domain, it can lead to a
v, i = vectors of voltage and current variables strikingly direct and effective solution method. In their
= matrix of propagation constants most basic terms, the underlying proposals for the formu-
Zo = surge impedance matrix lation emerged from a recognition of the close corre-
F = matrix of forward impulse responses spondence between the exponential form of the z-trans-
M\,M2 = modal transformation matrices form and that of the principal transmission-line responses
Vs = vector of source voltage in the frequency domain. From that starting point, the
RS,LS = resistance and inductance matrices in source paper develops z-transform electromagnetic transient analy-
model sis for a multiconductor power-transmission line, and it
At = trapezoidal rule step-length investigates its principal properties and computing time
tn = current time step in recursive sequences requirements. Companion papers 15 ' 16 further develop the
TV = total number of time domain solution points proposal, first by extending the present formulation to one
m — ratio of wave transit time to step interval for multinode networks, then by examining in detail the
z = z-transform parameter inherent accuracy and stability of general multinode analy-
Vectors of voltage and current variables at the switching sis, and, finally, by closely evaluating the relative merits of
end and receiving end of the line are identified by sub- z-transform and time-convolution analysis methods.
scripts s and r, respectively.
Vectors of conductor variables are distinguished from those 2 Scattering equations in the frequency-domain
of modal variables by superscript p.
Solving the second-order equations of propagation in a
single transmission line leads to
1 Introduction
= [exp{-
Following the early work on travelling-wave,1;2 frequency-
domain, 3 " 5 and time-domain6;7 forms of electromagnetic 0)
transient analysis in power systems, interest has for some l
= Zo (cj)[exp{-\(cj)x}]A(u)-
time now focused on time-convolution methods. 8 " 13 In
principle, these draw in a preparatory phase on frequency- [exp{AM*}]*(co)] (2)
domain concepts and inverse Fourier-transform evaluations
from which transmission-line impulse responses are formed, In eqns. 1 and 2, vx(co) and ix(u) are vectors of modal
following which electromagnetic transient analysis itself is voltage and current variables in the frequency domain 17>18
wholly in the time domain. Some advances in a particular at a point on the transmission line distance JC from the end
form of convolution analysis have been made in recent designated as the switching end. Z0(cS) is the modal surge
work,14 but the present paper does not seek to extend or to impedance matrix, and the terms [exp( — "k(co)x}] A (OJ)
contribute further to this pattern of development. The and [exp {A(GJ);C}] B(GJ) correspond to forward and back-
primary purpose of the present paper is that of proposing ward components of propagation respectively.
a new electromagnetic transient formulation which has its On successively adding and subtracting eqns. 1 and 2
basis in the z-transform. No previously published work has
suggested the possible use of the z-plane in this area of vx (w) + Z o (w)/ x (w) = 2 [exp { - A(w)*}] A (GO) (3)
applied analysis in power systems, but the paper shows that,
vx (w) - Zo (to)/, (co) = 2 [exp {A(GJ)X}] B{OS) (4)
Paper 988c, first received 3rd January and in revised form 14th July
Prof. Humpage, Mr. Wong, Dr. Nguyen are, and Mr. Sutanto was, At the switching end, for which x = 0
with the Department of Electrical & Electronic Engineering,
University of Western Australia, Crawley, Western Australia 6009. vs (w) + Z o (GO)I, (<o) = 2 A (GO) (5)
Dr. Sutanto is now with GEC Australia, Horsley Rd., Milperra, New
South Wales 2214 vs (co) - Z o (w)i8 (OJ) = 2B (w) (6)
370 IEEPROC, Vol. 127, Pt. C, No. 6, NOVEMBER 1980
0143-7046/80/06370 + 09 $01-50/0
Using A(cS) and B(u>) from eqns. 5 and 6 in eqns. 3 and 4 General forms for F(z) and Z o (z) are given in Section
gives 3.3, and the detailed derivation of these is developed in
Section 5, but it is useful first to consider the special case
= [exp{-X(w)x}]
for which losses are discounted. F(z) and Z0(z) are then
[».(w) + Z 0 (w)/ (7) easily formed, and the steps involved provide a basis from
which the more general nonlinear case may be derived.
= [exp{X(w)jf}]
(8) 3.2 Case for which losses are discounted

Eqn. 8 now rearranges to When losses throughout the transmission line are dis-
counted, ZQ (CO) is constant and real, and
vs(oj)-Z0(co)is(co) = [exp{-X(w)jt}]
exp{ — = exp( — j (16)
[vx(cj)-Z0(cS)ix(cj)] (9)
In eqn. 16, L and Care the total series-path inductance and
On setting x = I and using: shunt-path capacitance of the line, respectively, and VLC
is the wave transit time.
vr = vx(x = l) (10) We now introduce the z-transform parameter:19
ir = -ix(x=l) (11) z = exp(jojAt) (17)
eqns. 7 and 9 become
Choosing the time interval At in eqn. 17 to be an integer
va(cj)-Z0(cj)i,(cS) = [exp{-X(w)/}] submultiple m of the wave transit time in eqn. 18
(12) At = y/TC/m (18)
i>r(w)-Zo(w)ir(o;) = [exp{-X(co)/}] In this case

(13) exp{ — X(to)/} = z~ (19)

Having eliminated terms in exp{X(co)jc} by the rearrange- Eqn. 19 achieves directly the requirement in z-transform
ment of eqn. 8, only the forward impulse response, exp analysis of expressing the forward impulse response in
{ — X(w)/}, operating on the forward characteristics vs(oo) terms of z. Subject to losses being discounted, therefore
+ Zo(oo)is(cS) and vr(co) + Z0(co)ir(u)), is retained in F(z) = z (20)
eqns. 12 and 13. Although derived in a vector/matrix form
for the three modes of propagation taken together, the The constant surge impedance in the frequency-domain
form of these equations applies to each individual mode transforms to the constant Z o in the z-plane. Using this,
into which transient propagation is decomposed. In devel- and F{z) from eqn. 20 in eqns. 14 and 15 gives
oping z-transform analysis, it is the latter form, the scalar
one, that is most convenient to use. The development vs(z)-Zois(z) = z" (21)
applies identically to each mode, and, when this is com- zv(z)-Z 0 / r (z) = z- (22)
plete, the three modes are recombined in the formulation
on which a solution algorithm is based. On achieving a Using eqns. 21 and 22, the step of transforming from the z-
solution, it then remains to transform from modal to plane to the time domain can now be achieved in analytical
corresponding conductor variables in which form a solu- form. This is one of the advantages of the transform se-
tion is finally given. quence of the present work. The step corresponding to
numerical inverse Fourier transform evaluations in standout
3 Z-transform analysis convolution methods is one for which an analytical form is
available in z-transform analysis. On taking inverse z-trans-
3.1 Transmission-line equations in the z-plane forms of both sides of eqns. 21 and 22:
Taking inverse Fourier transforms of both sides of eqns.
12 and 13 leads to a time-convolution formulation. We vs(tn)-Zois(tn) = vr(tn_ (23)
here avoid this, and propose instead to transform first to vr(tn)-Zoir(tn) = vs(tn_ (24)
the z-plane and then, subsequently, from the z-plane to the
time domain. In this way, the z-plane is used as an inter- In the recursive sequences of eqns. 23 and 24, n identifies
mediate step between the frequency domain and the time the current time step: each time step is of duration At.
domain. When electromagnetic transient propagation is initiated at
Expressed in the z-plane, the transmission-line equations time t = 0
in the scattering form of eqns. 12 and 13 are
tn = nAt (25)
vs(z)-Z0(z)is(z) = (14) Similarly
vr{z)-Z0(z)ir(z) = F(z)[v.(z)+Z0(z)it(z)] (15) tn_m = (n-m)At (26)
In eqns. 14 and 15, F(z) and Z0(z) are, respectively, the If, in a particular application, the step interval, At, is
forward impulse-response and the surge impedance function chosen to be one fifth of the wave transit time, so that
in the z-plane. These are the z-plane counterparts of the m = 5, the previous value terms on the right hand side of
impulse response and the impulse characteristic impedance eqns. 23 and 24 are those formed at five time steps prior to
function of time-convolution analysis. the current one.
Ih'LPROC, Vol. 127, Pt. C, No. 6, NOVEMBER 1980 371
 The analytical form of the inverse transform by which Using the rational-fraction form of F(z) from eqn. 27 in
transformation is made from the z-plane to the time eqns. 34 and 35 gives
domain importantly avoids the consequences of truncation
in evaluating inverse Fourier transforms by numerical
methods as arises in frequency-domain analysis. Here,
truncation gives rise to Gibbs-type oscillations super-
= - y az
bo ;=o
imposed on solution variables which are then filtered from
solution using the Lanczos sigma factor. In z-transform
analysis, there are no corresponding components in solu- (36)
tion, and, therefore, additional forms of filtering do not
arise. From another aspect, the continuous form of trans-
mission line series- and shunt-path parameter distributions
-(m+y)
are fully preserved in the recursive sequence of eqns. 23 and = — > d;Z
24. It is from a combination of time-domain recursive series
which are not at any stage truncated and line parameter dis-
tributions which are maintained in them, that the main pro- (37)
perties of z-transform analysis derive.

3.3 Model with nonlinear parameter frequency- Substituting Z0(z) from eqn. 28 into eqns. 30—33 and
dependence rearranging using eqns. 36 and 37, gives
Turning to the more general case where losses are present (38)
and series-path parameters have a nonlinear dependence on
frequency, a substitution means of forming F(z), as in the vr(z)-Zlir(z) = vzr{z) (39)
lossless case through which z-transform analysis can be for
introduced clearly, is not now available. In addition, Z o (co)
is not now constant. The requirement is therefore one of = (Aoco +Zocd0)/d0 (40)
forming the z-plane functions F(z) and Z0(z) when the
nonlinearities which both contain preclude formal means of The right hand sides of eqns. 38 and 39 emerge from these
achieving this. The detailed steps of deriving F{z) and 2 0 (z) substitutions as
by least-squares minimisation methods given exp{ — \{oS)l]
and Zo (GO) in the frequency domain are given in Section 5.
These lead to the rational-fraction forms:
0 ;=0
(
2 ,djZ
j=o
F(z) = (27)
DO ;=1
;=o
"3
Ao 2 cjz-j J" I djz-^v
do y=i
(28)

;=o (41)
do y fi
where
7 "4
(29)
In eqn. 27 m has the same significance as that in the lossless
case, and is referred to in the general case of Section. 5.2. + ^f 1 dj
Given that the coefficients and scaling constants of eqns. -j {Bs(z)}
27 and 28 are available from the procedures of Section 5,
we can now develop z-transform analysis based on them. In + T 1 ds
this, it is convenient to form the forward and backward +» {F8(z)}
characteristics at the sending and receiving ends using " 0 ;=1
1
F
MO = r I. ajz-<
siz) = vs(z) + Z0(z)i8(z) (30) °o j o
F t (z) = vr(z) + Z0(z)ir(z) (31) ~~ I bjz-J {Br(z)}
Bs(z) = v.(z)-Z0(z)i.(z) (32)
1 "4
Br(z) = V r (z)-Z o (z)/ r (z) (33)
~T I djz-j M
The characteristics are then inter-related by d0 ;=1

Bs(z) = F(z)Fr(z) (34)

(42)
BJz) = F(z)FJz) (35) "o ;=i

372 IEEPROC, Vol. 127, Pt. C, No. 6, NOVEMBER 1980

 z "4
T id}Br(tn.j)
d0 y=i ]Z {lr Z

1 "4 As in the lossless case, n identifies the current-time step,

and in the time-series eqns. 43—46
tn = nAt (47)
3.4 inversion to the time-domain. tn-j = (n-j)At j = 1,2,...,R (48)
The final step in the development of the formulation is that f „_„,_,- = (n-m-j)At j = \,2,...,R (49)
of returning from the z-plane to the time domain by inverse
z-transform operations. On taking inverse z-transforms of the value of m in eqn. 49 is at choice in analysis.
both sides of eqns. 38 and 39 From the derivations of F(z) and Z0(z) in Section. 5,
the largest value of k typically required in the individual
(43) series from which F(z) and ZQ(z) are formed is four. In
that case, the formulation draws on the previous values at
iAtn) = Vzr(tn_j) (44)
solution steps (n — 1), (n — 2), (n — 3), and (n — 4), and
The previous value terms in eqns. 43 and 44 are in turn also (n—m—1), (n—m—2), (n — m — 3), and
given by (n —m— 4), in forming solution values at step n. It will be
seen that the final time-domain series of eqns. 43 and 44
are of the same form as those in the lossless case: the
1 "» considerably greater degree of detail in the general case,
Vzsitn-i) = - I ajFr(tn-m-j)
°0 y=0 particularly that arising from the nonlinear frequency
dependence of line parameters, is reflected into the
previous-value calculations.
"2

3.5 Vector form

Although it is convenient to develop z-transform analysis
I 4 on the basis of a single mode and therefore in scalar form
T I. djV8{tn.j) throughout, the derivation and the final form of the time
" 0 ;= 1 domain sequences is clearly identical for each mode. Having
now completed the steps of the formulation for the z-
transform line model, we now adopt, from eqns. 43 and 44,
(45) the vector/matrix notation:
^ • 1 %('„-,)
" 0 ;= 1
v s (f n )—Z/i s (O = vzs(tn.j) (50)

Zoc "4 w r ( r n ) - Z , / r ( r n ) = vzr(tn.j) (51)

— y t
0 y=i 4 Composite model in the time domain
4.1 Source representation
When the source from which circuit-breaker switching in
do ;= 1 line energisation takes place is represented in the simplified
equivalent form often used in switching overvoltage studies,
and the voltage variables at the switching end of the line in
1 "l vector vf (t) are related to specified source voltages in
Vzr(tn-i) = — I ajFs(tn-m_j) Vf(t)by
(0 (52)
J "2
in which Rs and Ls are matrices of equivalent source model
do j= i resistance and inductance parameters, respectively. The
superscript notation in eqn. 52 identifies phase co-ordinates
| "4 whilst the transmission-line formulation leading to eqns.
T Z0 50 and 51 is throughout in modal variables. In Appendix
10.1, it is shown that, transforming to modal variables, so
that the source model can be used directly in conjunction
with the line model, and, using the trapezoidal rule with
do j=i (46) step length At, leads to
vs(tn) = Vs(tn) —Zsis(tn) + vp(tn_\ ) (53)

do j=i Now expressed in modal variables, the sequence of eqn. 53

-rIdMt n.j) is one that can be combined directly with the line model
" 0 y= 1
in the time domain.

IEEPROC, Vol. 127, Pt. C, No. 6, NOVEMBER 1980 373

4.2 Combining source and line models. We therefore propose to use
In forming a composite model for a single source/line 0(w) = 0 n (co) + 0,(co) (59)
interconnection, vs(tn) in eqn. 50 is replaced by the in which the subscripts n and / refer to the nonlinear and
expression for vs{tn) in eqn. 53 to give linear components, respectively, of the |3 function.
(tn) - [Z, + Zs] is(tn) = v2 _,) - vp (54) Following the derivation in the lossless case of Section
3.2, we set:
In solution, is(tn) in eqn. 54 is solved from
= u>mAt (60)
Using
(55)
Xn(w) = (61)
In the case of energising an open-circuited line, ir(tn) = 0 the forward impulse response can be formed from
for all n and so, from eqn. 51
exp{ — X(GJ)/} = exp{ — \n(oj)l}e\p(— jcomAt) (62)
*>r(O = Vzr{tn-j) (56)
The second exponential term on the right hand side of eqn.
Using eqn. 56, the receiving-end voltage solution is cal- 62 is in a form into which the z-parameter can be substi-
culated entirely from previous values.
tuted as in the lossless case. The requirement is then one of
forming exp (—dn(Kn(co)l), first in the frequency domain,
4.3 Transformation to phase-variables
and then, by transformation, in the z-plane. For this purpose,
On achieving a solution for vr{tn) for n = 1, 2, . . . ,N, we propose to use
conductor voltages are finally formed from
exp{-X n (w)/} = , , , ? (63)
v?(tn) = Mxvr{tn) (57)
in which Mx is the modal/phase-variable transformation for
matrix for voltage variables. In general, Mx is frequency S = (64)
dependent, but the consequence of using it at a single fre-
quency in final transformation, as in eqn. 57, has previously Taking the modulus of both sides of eqn. 63:
been evaluated,14'20 and has been shown to lead to only
low-order error. For frequencies in excess of about 1-5 kHz, |exP{-XnM/}| = (65)
the real part of Mx becomes almost constant, and is much
greater than the imaginary part. In evaluating eqn. 57, the
real part of Mx is therefore used. Providing that the fre- In Fig. 1 is shown the |exp{ — X n (cj)/}| function. Given
quency at which it is calculated is greater than 1 -5 kHz, the this, numerical values for gx and g2 may easily be found
precise frequency selected for this is not critical: in the using a least-squares minimisation procedure, although the
present work it is 5 kHz. second-order form of eqn. 63 leaves a residual low-
frequency error. Rather than increase the order of the poly-
nominal in eqn. 63 to improve matching, we here propose
to eliminate low-frequency error by the notch-filter
5 z-plane responses and functions response:
5.1 Principal steps in forming F (z) and Zo (z)
Both the forward impulse response, exp( — A(co)/}, and the I+/3S+/4S2
modal surge impedance function, Z0(co), are available in
the frequency domain and can be formed numerically for a The filter characteristic required to counter low-frequency
given range of frequencies. Rational-fraction forms of these error is first formed from the error between the calculated
can then be found by a least-squares matching procedure. response of Fig. 1 and that corresponding to eqn. 65.
Known transformations between the frequency domain and Filter coefficients f\, f2, fz and/ 4 are then found by curve
the z-plane are then available finally to form the required fitting to this error function. The complete forward
impulse response, F(z), and the related surge impedance impulse-response function is then represented by
function, Z o (z), in the z-plane. Completing this process (1 + / i s + / 2 s 2 ) e x p ( — j com At)
provides the ah bh ch dh Ao and Zoc coefficients of the exp{ —
general formulation of Section 3.3. The procedure is closely
similar in forming F{z) and Zo (z) but, for clarity and com- (66)
pleteness, the two cases are here developed separately.

5.2 Forward resp onse

On separating out the real and imaginary parts oc(co) and
/3(o/) of the propagation constant X(co), a numerical evalu-
ation of j3(co) confirms that it can be represented very
closely by a combination of one linear and one nonlinear
function of frequency. Given 10 20 30 UO 50
frequency,(rad/sec) 10
X(co) = a (58) Fig. 1 Variation of forward-response magnitude with frequency
374 IEEPROC, Vol. 127, Pt. C, No. 6, NOVEMBER 1980
This now transforms to the z-plane using of the line, the fault level is 20000MVA and closure
m coincides with a voltage peak in phase V . For the purposes
exp( — icom^t) = z~ (67) of comparison, solutions are shown in Figs. 3 & 4 from the
and z-transform model developed in the paper and from a full
frequency-domain algorithm. For high accuracy in the
' z-\ solutions against which comparisons are made, 500 fre-
s = At (68)
quency points have been used in frequency-domain
analysis, with a frequency separation of 200rad/sec
Substituting from eqns. 67 and 68 in eqn. 66 then gives the covering the frequency range 0—105 rad/sec, and with a
forward impulse response in the z-plane: step length of 20/us in evaluating inverse Fourier trans-
F(z) = forms. Line parameters are re-evaluated at each frequency
point. This appears to provide the most reliable reference
^4) solution available at the present time. The z-transform
2 solutions of Figs. 3 and 4 are for a time step of 50 jus.
b2z~ + b3z~ Closely similar accuracies to those of the solutions in
(69) Figs. 3 and 4 are achieved for other switching modes and
The relationships between the a and b coefficients in eqn.
69 and the / and g constants of eqn. 66 follow from the
substitution and are collected together in Appendix 10.2.

5.3 Surge impedance function in the z-plane

At high frequency, the surge impedance tends to a con- -120-
070 1-40 210 280 3 50 420 \49O,
stant, real value denoted in Section. 3.3 by Zoc. The vari- 5 60 6 30
ation of | Z 0 ( C J ) — Zoc\ with frequency is shown in Fig. 2. -360i
time, ms
700
The principal steps of Section 5.2 are now followed in
fitting to this calculated response. A first-order form is in -600 J
this case satisfactory in the initial least-squares fitting.
To this is added a notch filter, so that, overall, we propose Fig. 3 Sending-end phase 'a' voltage transient in simultaneous
energisation
e2s2) 160 km, 400 kV single-circuit transmission line.
7 c \ 20 000 MVA fault level at switching point
(70)
z (s) =
° frequency domain
z-transform
Substitutions from eqn. 68 in eqn. 70 when the h and e
coefficients have been found lead to the z-plane form
ClZ~ C2Z~2 +C3Z 3
)
T (71)
d0
6 30 X
The relationships between c and d coefficients in eqn. 71 700\
and the e and h parameters of eqn. 70 are summarised in time ms
Appendix 10.3.

6 Trial solutions

In Figs. 3 and 4 are shown sending-end and receiving-end Fig. 4 Receiving-end phase 'a' voltage transient in simultaneous
energisation
transient waveforms for a 160 km, 400 kV untransposed
single-circuit transmission line the principal data for which 160 km, 400 kV single-circuit transmission line.
20 000 MVA fault level at switching point
are summarised in Appendix 10.4. The switching mode is frequency domain
that of simultaneous pole-closure at one end of the line z-transform
when the other end is on open circuit. At the switching end

Fig. 5 Receiving-end phase 'a' voltage transient in single-

20 30 u40 conductor energisation
frequency , (rad /sec)10
160 km, 400 kV single-circuit transmission line
Fig. 2 Variation of surge-impedance function with frequency 20 000 MVA fault level at switching point
frequency domain
Zoc = Lt {Z0(co)} z-transform

IEEPROC, Vol. 127, Pt. C, No. 6, NOVEMBER 1980 375

for a wide range of e.h.v. transmission line constructions. The computing-time requirements of z-transform analy-
Transient waveforms for the case of pronounced sis to achieve these accuracies are exceptionally low. The
asymmetry in energising only one conductor of a multi- full frequency-domain solution against which comparisons
conductor line are shown in Figs. 5 and 6. Here, ground- are made requires about 150 sec (CDC Cyber 72/73). A
return modes are of principal importance. Fig. 6 confirms standard time-convolution solution13'20 providing similar
the solution accuracy in the case of an unswitched con- accuracy requires 1 -5 s once the impulse responses and
ductor waveform. In some earlier methods, the solution impulse characteristic impedance matrix functions have
accuracy may have had some dependence on the fault level been formed. The z-transform solution is achieved in 0 1 5 s.
at the point from which energisation is made An onerous
test condition is then one in which source parameters are 7 Conclusions
reduced to zero leading to the limiting condition in which
the fault level tends to a large value. Figs. 7 and 8 confirm From numerous trial studies on e.h.v. transmission lines of
that the accuracy of the new method remains very high different configuration, the new formulation based on the
for this test case. z-transform and the solution sequence to which it leads
combine high accuracy and reliability with low computing
time requirements. The formulation itself is a direct one
200
and the final time-domain sequences for solution lend
160
120
themselves very easily to a programming implementation.
80 No additional measures or forms of compensation are
40 required.
0 u . , .V, . , .£/, , . Although the special role of the z-plane in transforming
3 50 A20y<r, 5 60 630 / between the frequency domain and the time domain is
-40
A 90 700 central to the formulation of the paper, it does not draw
-80 time ms
-120 extensively on formal z-transform theory. The use of the
-160 z-plane is a particular one which derives from the
-200- exponential form of transmission line impulse responses in
Fig. 6 Receiving-end phase 'b' voltage transient in single the frequency domain. Given the further development of
conductor energisation the companion papers,15'16 the proposals of the present
160 km, 400 kV single-circuit transmission line. work seem to have wide scope and to offer further con-
20 000 MVA fault level at switching point tributions to electromagnetic transient analysis methods in
frequency domain
z-transform power systems.

8 Acknowledgments
The authors are grateful to the Australian Research Grants
Committee for financial support, and to the West
Australian Regional Computing Centre for running their
programs. The generous support of power-systems research
in the Department of Electrical & Electronic Engineering at
the University of Western Australia by Professor A.R.
Billings, together with his professional co-operation at all
times, is gratefully acknowledged.
1 60 200 2-40 2-80 3-20 3 60 400
time , ms 9 References
Fig. 7 Receiving-end phase 'a' voltage transient in single-
conductor step energisation 1 BICKFORD, J.P., and DOEPEL, P.S.: 'Calculation of switching
transients with particular reference to line energisation', Proc.
160 km, 400 kV single-circuit transmission line. Switching from IEE, 1967, 114, (4), pp. 465-477
zero-impedance source representation
frequency domain 2 BICKFORD, J.P., MULLINEUX, N., and REED, J.R.: 'Com-
z-transform putation of power system transients', IEE Monograph 18 (Peter
Peregrinus 1976), p.75
3 BATTISSON, M.J., DAY, S.J., MULLINEUX, N., PARTON,
K.C., and REED, J.R.: 'Calculation of switching phenomena in
power systems', Proc. IEE, 1967, 114, (4), pp. 478-486
4 BATTISSON, M.J., DAY, S.J., MULLINEUX, N., PARTON,
K.C., and REED, J.R.: 'Some effects of the frequency
dependence of transmission line parameters', ibid, 1969, 116,
(7), pp. 1209-1216
320 3 60 400 5 BATTISSON, M.J., DAY, S.J., MULLINEUX, N., PARTON,
K.C., and REED, J.R.: 'Calculation of transients on trans-
time,ms mission lines with sequential switching', ibid., 1970, 117, (3),
pp. 587-590
6 DOMMEL, H.W.: 'A method for solving transient phenomena in
multi-phase systems', Proceedings of the Power Systems
Computation Conference, 1966, Report 5.8
Fig. 8 Receiving-end phase 'b' voltage transient in single- 7 DOMMEL, H.W.: 'Digital computer solution of electromagnetic
conductor step energisation transients in single- and multi-phase networks', IEEE Trans.,
160 km, 400 kV single circuit transmission line. Switching from 1969, PAS-88, (4), pp. 388-396
zero-impedance source representation. 8 BUDNER, A.: 'Introduction of frequency-dependent line par-
frequency domain ameters into an electromagnetic transient program', ibid., 1970,
2-transform PAS-89, (1), pp. 88-95
376 IEE PROC, Vol. 127, Pt. C, No. 6, NOVEMBER 1980
 9 SNELSON, J.K.: 'Propagation of travelling waves on trans- Following transformation to the modal axes, eqn. 52
mission lines: frequency-dependent parameters', ibid., 1972,
PAS-91, (1), pp. 85-90 becomes
10 SEMLYEN, A., and DABULEANU, A.: 'Fast and accurate
switching-transient calculations on transmission lines with Vs{t) = Vs(t)-(RS+Lsp)is(t) (72)
ground return using recursive convolutions', ibid., 1975, PAS-
94, (2), pp. 561-569 where
11 CARROLL, D.P., and NOZAR1, F.: 'An efficient computer
method for simulating transients in transmission lines with vs(t) = M\lifs(t)
frequency-dependent parameters', ibid., 1975, PAS-94, (4),
pp. 1167-1174 Vs(t) = M\l Vf(t) (73)
12 SEMLYEN, A., and ROTH, A.: 'Calculation of exponential pro-
pagation step responses', ibid., 1977, PAS-96, (3), pp. 667-671 is(t) =Milif(t)
13 AMETANI, A.: 'A highly efficient method for calculating
transmission line transients', ibid., 1976, PAS-95, pp. 1545- RS =M
1549 (74)
14 NGUYEN, T.T., WONG, K.P., and HUMPAGE, W.D.: 'Impulse
sampling sequences in time-convolution electromagnetic trans-
ient analysis in power systems', Electr. Power Syst. Res. (in On substituting into the trapezoidal rule
press)
15 HUMPAGE, W.D., WONG, K.P., and NGUYEN, T.T.: vs(tn) = Vs(tn)-Zsis(tn) + vp(tn_ (75)
'Development of z-transform electromagnetic transient analysis
methods for multi-node power networks', IEE Proc. C, Gen., for
Trans. & Dist., 1980, 127, (6), pp. 379-385
16 HUMPAGE, W.D., WONG, K.P., and NGUYEN, T.T.: 'Time- Z c = RK+aLK (76)
convolution and z-transform methods of electromagnetic
transient analysis in power systems', ibid., 1980, 127, (6), pp. vp(tn-i) = V8(tn.1)-v8(tn.1)-(R8-aL8)i8(tn.l)
386-394
17 BOWMAN, W.I., and McNAMEE, J.M.: 'Development of equiva- (77)
lent matrix circuits for long untransposed transmission lines',
IEEE Trans., 1964, PAS-83, pp. 625-632 and
18WEDEPOHL, L.M.: 'Application of matrix methods to the
solution of travelling-wave phenomena in polyphase systems', a = 2/At (78)
Proc. IEE, 1965, 110, (12), pp. 2200-2212
19 GABEL, R.A., and ROBERTS, R.A.: 'Signals and Linear As in Section. 4.3, the real parts of modal matrices Mi and
Systems'(John Wiley, 1973) p. 171 M2 evaluated at 5 kHz are used in eqns. 73 and 74.
20 BERGMANN, R. Ch. G., and PONSIOEN, P.J.M.: 'Calculation of
electrical transients in power systems: untransposed trans-
mission line with frequency-dependent parameters', Proc. IEE,
1979, 126, (8), pp. 764-770 70.2 Forward-response coefficients.
10 Appendixes In relating the a and b coefficients of the rational-fraction
form of F(z) in eqn. 69 of Section. 5.2 to the / and g
10.1 Source representation. coefficients of eqn. 66 and as found by curve fitting to the
Vectors of phase variables and the associated coefficient response of Fig. 1, the following are first defined:
matrices of eqn. 52 of Section. 4.1 relate to the simplified
equivalent-source model of Fig. 9. Phase impedances are
calculated from the fault level specified at the terminals of (79)
the model, whilst the neutral impedance is calculated from
a specified ratio of source impedances in the positive- and 4/ 2
= 2 1 - (80)
zero-phase sequences. (At)

Afl
= l - ^ - ++ (81)
At (At)

+ (82)
At (At)

(83)

+ (84)
Zsc
At (At)

_
t
At (At)2
(85)

h = 2|1 — _2 (86)

2gl
- i 11
At (At)2
(87)
Fig. 9 Simplified equivalent source representation
IEE PROC, Vol. 127, Pt. C, No. 6, NOVEMBER 1980 377
 Then: and
a0 = t2 d0 = u s u n (99)
=
a.\ t\ 4- 2t2 dx = u4u1 + us u6
a2 = to.+ 2ty +t2 (88) d2 - u4u6 + u3 un (100)
a3 = 2? 0 + U d3 = u3 u6
=:
QQ to
10.4 Transmission line data in test analyses.
and:
Fig. 10 shows the tower configuration and conductor
^o =
^s ^8 spacings for the 400 kV transmission line on which the test
analyses of Section 6 have been based. Other basic line
^i = tsh+Uh data are collected together in Table 1.

b2 = t5t6+ U h + t3 t 8 (89)

£3 — U t6 + t3 tn

b4 = h t6 6 78

10.3 Surge-impedance func tion coefficien ts o;o

From the coefficients of the Z0(z) function in eqn. 70 in - • - 9 98

o!o
Section. 5.3, we define:
o'o
(90) 4153 !—-~8 15 i 3124
° At (At)2 ^-O;O

(91)
0 3048--hr
"I 20-88

0 3048 12 04
_ 2ex 4e2
Ul (92)
~ l~~Al + (A~ty
/ / / / / / /
(93)
At (At)2 Fig. 10 400 k V transmission-line conductor spacings
upper = phase 'a'
middle = phase ' b '
u4 = 2 II ^2 (94) lower — phase ' c '
spacings in metres
. 2e,
(95)

= 1 + (96)
At
Table 1 : Basic transmission line data
2hy
= 1 — (97) Number of circuits 1
At Number of conductors per phase 4
Number of earth-wires 1
Conductor position symmetry none
The c and d coefficients of eqn. 71 are now given by: Conductor resistivity, S7m 3-2X 10"
Earth-wire resistivity, S7m 2-69 X 10"
c 0 = u2 Conductor strand diameter, cm 0-32
Earth-wire strand diameter, cm 0-32
c, = lit +u2 (98) Geometric mean diameter for 4-conductor bundle, cm 30-94
Outer diameter of earth-wire, cm 2-86
c2 = u0 + ut Number of effective strands in phase conductors 54
Number of effective strands in earth-wire 54
c3 = u0 Earth resistivity, SIm 200

378 IEEPROC, Vol. 127, Pt. C, No. 6, NOVEMBER 1980