Professional Documents
Culture Documents
2, MarchIApril 1975
TVTI1 = U (71
Paper T 74 4354, recommended and approved by the IEEE Transmission &
Distribution Committee of the IEEE Power Engineering Society for presentation at where t h et r a n s f o r m a t i o nm a t r i c e s Tv and TI a r ef r e -
the IEEE PES Summer Meeting& Energy Resources Conf., Anaheim,Cal., July 14-
19,'1974. ManuscriptsubmittedJanuary 30, 1974; madeavailableforprinting quencydependent. However, f o r a verywiderangeof
May 1,1974. f r e q u e n c i etsh e sm
e a t r i c eps r o vtebope r a c t i c a l l y
constant8' 15' I tnh ce a s oehf o r i z o n t alli n etsh e y Now t h ei n c i d e n t modal v o l t a g e s( 1 2 )a t h er e c e i v i n g
arejustslightlydifferent fromthe a,B, 0 transforma- end are known.
t i o n m a t r i c e s : mode 1 i s e s s e n t i a l l y e q u a l t o m d a w i t h A s t a t ev a r i a b l ep r e s e n t a t i o no ft h eo p e r a t i o n si n
a s l i g h t e f f e c t from mode 0 , and viceversa,while mode t h i s s t e p is d e s c r i b e d i n Appendix4.
6 remainsunaltered.
Since the transformation matrices arc constantequa- Step4.Incident modal currents.Moda1 wave c u r r e n t s are
tions(6)arealsovalidinthetimedomain.Therefore, r e l a t e d t o u n i t s t e p modal wave v o l t a g e s by a s i n g l e ex-
modal (i.e. independent)
instantaneous
voltages and ponentialresponsefunction(Fig.2). I t S v a l u ea t t = -+
c u r r e n t s canbeobtainedby: corresponds t o
v = T-l v
i
m
=
v1
TI
Yc ( w - + O ) =E = 0 (13)
I ft h ei n p u t s u i n ( 5 ) a r et h e modal i n c i d e n t V O l -
tagesv'of(12),thentheincident modal currentsbe-
m
come simply,
i; = ySm (v; - xm) (15)
where,ofcourse, x
m
is theupdatedvalueaccording to
equation (5).
A state variable presentation of this step i s given
i n Appendix 5 . Steps 3 and 4 canbe combined into a
singlestaterepresentation,givenin Appendix 6.
AT:
T - t r a v e l time a t l i g h tv e l o c i t y
-a d d i t i o n a lt r a v e lt i m e i'ml
Step 3.
Distortion of modal waves. Fig. 1 indicates
t h r e e b a s i c componentsof a d i s t o r t e d modal wave v o l t -
age,resulting rrom a f i t t i n g of exponentialcurvesto
thecalculated or experimentallyobtaineddata. Due t o
i t s s h o r td u r a t i o n ,t h et h i r d component hasnopracti-
c a ls i g n i f i c a n c e ,a c c o r d i n gt os e v e r a lt e s t s . The sec-
ond component is importantinlongdurationprograms,be-
causeof i t s l a r g e time c o n s t a n t I. at s s u r e st h a t
charge i s l o s t on a l i n e w i t h no conductance t o ground.
no 0 -t
Both components have the form o f equation (2). Four
constantsresultforeach mode: F i g . 2 . Incident modal wave c u r r e n t due t o a unit step
madal i n c i d e n tv o l t a g e .
(m = 1,2,3)
where t hsee c o nsdu b s c r i pi nt d i c a t et hs e component Step 5 . Equationsrelating vm t o in. The followingfre-
( f i r s t s, e c o n d ) . Note t h a tf o r any mode m theampli- quency domain e q u a t i o n sa r es t r a i g h t f o r w a r d :
tudesofthe two components add up t o u n i t y :
I = 1' + I"
From (16)
Yc v = I' - I"
(m = 1 , 2 , 3)
From (17) and (18)
The p r o c e d u r e i n t h i s S t e p 3 consists of .:he follow- Yc v + I = 2 1 '
ing :
Denote
I n t e r p r e t u i n eqn. (5) as v!Ide of (9) and x a s one j = Y c V
o ft h e components, f o re a c hm J e ? Update x using ( 5 ) and compare t o
and t h u so b t a i nt h er e s u l t a n t modal incidentvoltages
v ' = +x x
I ' = Yc vi
m ml m2
562
Obviously,the same algorithm-inthetime domain- r e - 5. C a l c u l a t e l i n e a r r e l a t i o n s h i p s between modal
lates j tov,asthe onedescribed at t h e p r e v i o u s s t e p , l i n e v o l t a g e s and c u r r e n t s ,
r e l a t i n g i ' t o v;, The equivalentofequation (15) i s 6 . Convert modal v o l t a g e - c u r r e n tr e l a t i o n s h i p s
m
i n t o r e l a t i o n s between phase quantities.
j = Y,(22)(v - x) 7. Combine with similar equations of a l l o t h e r
components a t t h e same l i n e endandsolve l?r
Accordingly,the time domain equivalentof (19) becomes a l l phase voltages and t h e n f o r t h e respect-
i v ec u r r e n t s .C a l c u l a t er e f l e c t e dv o l t a g e s v"
j + i = 2i' (23) and r e t u r n t o 1.
or, with(22)
EXPONEhTIAL FITTING BY A NEWTON-RAPHSON
ALGORITHM
Ys(vn
- x ) + in = 2iA
n
124)
As mentioned i n t h e p r e v i o u s s e c t i o n , t h e modal u n i t
where t h es u b s c r i p t n hasbeenadded, as a time count, s t e pr e s p o n s eo ft h el i n e i s approximated bya sum of
r a t h e rt h a ns u b s c r i p t m whichmighthave.beenused to two exponentials whose amplitudes add up t o u n i t y :
s t r e s s t h e modal c h a r a c t e r o f t h e v a r i a b l e s i n ( 2 4 ) . - B ( t--ttoo,)-a ( t
v ' = a(1-e I. + (1-aI(1-e
. . . 1 (31')
or with shorthand notations:
v' = 1 - aM - -
(1 (31")
a)N
S u b s t i t u t i o no f (5) (with v r e p l a c i n g u) i n t o (24) where -a ( t - t o )
yields ~ = e
Y, (1 - X ) vn + in = 2 i j l + Y, ( a xn-l + Ir v ~ - ~w +~ ~ - ~ )
N = e- B ( t - t o )
t 0 = additionaltravel
time (=AT i nF i g . 1 )
m
(251
or, emphasizingagainthe mode number m, r a t h e rt h a n The fourparametersa, a, B, to canbecalculated
the time count n, using a least square error minimizing technique
ym vm + im= i
Pm (26)
where
where yi r e p r e s e n t d i s c r e t e l y known v a l u e s f o r t h e mcdal
u n i ts t e pr e s p o n s e .
i = 2iA + y (a + ~ J V ~ - ~v+ ~
U - ~(27")
) Equation (33) i s e q u i v a l e n t t o a systemofnonlinear
Pm
equations
565
CONCLUSImS a = -ah
e
IntheconvoluLion
x(t) = / h(r) u ( t -T) dr (37)
0 are c o n s t a n t s .
x, u, and h a r e t h e o u t p u t , i n p u t and the unit impulse
r e s p o n s er, e s p e c t i v e l yI. hf ( r ) i s exponential, i . e . Appendix 2: Derivationof(5)for First OrderInput
= be-a'
h ( rApproximate
) (38) u of (4) by eaoxrfpdirreesrst s i o n
u - u
then i t hasbeen shown t h a t *t
U(nh - T) I n-l T + U (A. 5)
n
x ( t ) = e-at
X(t - At) + / h ( r )u ( t -T)dT (39) S u b s t i t u t i o no f (A.5) i n t o (4) and i n t e g r a t i o n by p a r t s
0 where
produces
equation
(5), a I as before,;
ah
but now
Equation(39)provesthattheconvolutionx(t)canbe
obtained recursively by using the previous convolution
A = -(1; - 2)
x(t -
At) and averylimited amount ofinputcontained
in the integral of the right hand side of (39).
5. The method produces a simple Norton-type line equiv- y ( t ) = /H(T) U(t - T)dT (A*8)
0
a l e nct ,o n s i s t i nogf a constant
admittance and
a
permanently updated current
source.
Consequently it
canbeeasilyintegratedinprogramsbased onnodal whereH(t) is theunitimpulseresponse(inverseFour-
system representation. ier transform of H(u)).
I f H(w) r e s u l t s from s t a t e equations
k=Ax+Bu
(A. 91
ACKNOWLEDGEMENT y = C x + D u
i.e.
The a u t h o r s w i s h t o e x p r e s s t h e i r a p p r e c i a t i o n f o r ~ ( w )= C(jw-u)-l B + D [A. 10)
f i n a n c i a l s u p p o r t by the National Research Council of
Canada. then it is c l e a r l y an a l g e b r a i cf u n c t i o no f w and t h e
unitimpulseresponse i s a matrix of exponential func-
Appendix 1: DerivationofEquation ( 5 ) f o r Second t i o n so f t . In such a s i t u a t i o nt h ec o n v o l u t i o n (A.8)
OrderInput can be performed analytically, except for thelasttime
step.
Approximate u of (4) by a second order expression
Fornotationalconveniencereduce (A.8) t oa form
U(nh T) E at2 + yT + Un (A. 1) e q u i v a l e n tt o( 3 ) :
-
m
This is e q u i v a l e n t oe q u a t i o n( 4 ) C
p l e t ei n t e g r a t i o n from h t o
placedby a s i n g l et e r m ,
-
. l e a r l yt h e
i n (A.12) hasbeenre-
e - & x ( t - h ) i, n( A . 1 5 ) .
com-
Or
Appendix 6.StateVariableFormulationofSteps 3 & 4
i no t h e r words, t h e last s t a t ex n - 1 i n (4)transmjts The equationsofAppendices 4 and 5 canbe combined
a l lt h ei n f o r m a t i o no b t a i n a b l e byan i n f i n i t ec o n v o l - i n t o a s i n g l e s e t of s t a t ee q u a t i o n s .S i n c ee a c h mode
u t i o n e, x c e p tf o r one singlestepofconvolutions in i s independent,thefollowingequationsarevalidfor
( 4 ) , whichhasbeen e a s i l y r e d u c e d t o an a l g e b r a i c fonn any mode and t h e mode number will n o t b e s p e c i a l l y i n -
in (5). dicated. Thus, f o r any mode, theequivalentofequa-
Appendix4. StateVariableFonrmlationof Modal Wave t i o n s (A.17) is
Distortions(Step3)
'1 -al ' 1 + b l 'ideal
A s t a t ev a r i a b l ef o r m u l a t i o n o f Step 3 i s based on
equationsofthetype(A.9), or (A. 20)
52 = -a2 '2 + b2 'ideal
v ' = c1 + c2
i=Ax+Bv'
ideal,
(A. 16)
v' = cx
m where,according t o (11) ,
r .
P 1
1 1
(A. 17")
According t o (A.24) t h es t a t ev a r i a b l ed e s c r i p t i o n
of the distortion ofeach modal wave r e q u i r e s t h e know-
ledgeoffiveconstants:al, a , b l , a and y A total .
of 10 Constants will be require2 €or a symmeSrica1 l i n e
and 15 c o n s t a n t sf o r an untransposedlinemodel.
w i t ht h ec o n s t r a i n t s (11).
Appendix 7 . StateVariableFormulationofLineDistor-
Appendix 5. StateVariableFormulationofIncident tion(Steps3,4,5)
Modal CurrentCalculation(Step 4) Equations (A.24) y i e l dv ' and i ' f o r each mode. I t
is now p o s s i b l et o add Step 5 and produce a complete
The g e n e r a ls t a t ev a r i a b l ef o r m u l a t i o n of(A.9)can state equation formulation of t h e d i s t o r t i o n e f f e c t s o f
beapplied t o Step 4 ofthegeneralalgorithm: t h et r a n s m i s s i o nl i n e .
k = A x + Bv'm Compare equations(20)and (21), and w r i t et h e eq-
uivalent of (A. 22) for the product (20) :
1 (A. 18)
i = C x + D v '
m m
56 7
6, = -a c4 + av Appendix 8. D e t a i l s of theExponentialFitting.
(A.251
j = -YsSq + YSV Consider the following shorthand notations:
S u b s t i t u t ej from (A.25)and i f from (A.24) i n t o( 2 3 ) .
Thisyields
(A. 27')
or, i n m a t r i x n o t a t i o n ,
y s v + i = ys cc (A. 28")
With t h en o t a t i o n si n (A.32) t o (A.34) t h ec o e f f i c i e n t s
Equation (A.28 ) can be w r i t t e n , s i m i l a r l y t o (5) i na of equation(35) become:
f o n anuropriate for numerical updating:
5, = a Sn-l + L ' v!ideal, + " Vldealn-l + " Vldealn-2+
+N'vfdealn-2 + , Y " V ~ -+
~ N"v,-~ ) 3. J. P . Bickfordand P . S. Doepel,"Calculation of
Switching Transients with Particular Reference to
(A. 31) LineEnergization",Proc.I.E.E.,Vol.114,April
or, simply,equation(26)inwhichthe mode number m 1967,pp.465-477.
hasbeenreintroduced.
The purpose of this appendix has been to produce a
simplematrixformulation, by equations(A.28),forthe
whole
process of
terminal wave d i s t o r t i o n . However,
computationally,thestep-wisesolutionoutlinedinthe
main t e x ta p p e a r st o be t h e more straightforwardap-
proach.
-= -aMi + BNi
atO
a Bi12)
-= aMi(a(ti - to) - 1)
atO
(A. 34)
-= (l-a)Ni(B(ti - to)-l)
atO
aBj4)
-= - a a2Mi - (1 - a) B2Ni
atO
4. A Semlyenand E. Wagner, "Beitrag zum genauren Ber- Discussion
echnenderSchaltspannungenin Hochspannungs lei-
tungen nach dem Bergeron-Verfahren", Efektrotech- John G.Anderson (General Electric Company, Schenectady, N.Y.): A
n i s c h e Z e i t s c h r i f t (ETZ-A Vo1.90,1969,No.18,pp.436 great deal of effort has been applied in recent years to the calculation
-440. of frequency dependent response of transmissionlinesin both the
frequency and time domains, and Prof. Semlyen and his coworkers have
made a significant contribution in attaining what appears to be a good
5. H . B. Thore'n and K. L . Carlsson,"ADigital Computer acceptable approximation to theusual convolution approach. They have
Program f o r t h e C a l c u l a t i o n of SwitchingandLight- also made a very constructive addition to the network solution methods
ningSurgeson Power Systems", IEEE Trans.on Power so well advanced by Herman D o m e 1 in recent years. Several questions
ApparatusandSystems,Vol. PAS-89, February1970, come to mind as one reads this interesting paper:
pp. 212-218. 1. The utilization of modal analysis in the computation of attenu-
ation and distortion of travelingwaves on multiconductor lines has
6. J . K . Snelson,"PropagationofTravelling Waves on been highly successful in radio noise and carrier current propagation,
TransmissionLines - Frequency Denendent
Parame- and it is natural that its utility in switching-surge analysis be continued.
ters", IEEE Trans.on Power Apparatus and Systems, However, many of the serious switching surges on power systems are
Vol. PAS-91, Jan/Feb.1972, pp. 85-91. due to faults onthe lines themselves, and whereas the modes are
normally independent, they interact with one another atthe fault. Now
7. R. G. Wasley
and S. Selvaninayaeamoorthy, "Forward modes are to travelingwaves what symmetrical components are to
and Backward ResponseFunctions for Transmission power frequencies, and at the fault equivalent circuits in symmetrical
LineTransientAnalysis", IEEE Paper No.T73435-5. components or Clark components have been useful to represent various
fault conditions. Have theauthors given any consideration to the
development of equivalent modal circuits to represent various line fault.
8 . A . Ametami, "Refraction
Coefficient Method for conditions? What modal procedures do they find most useful to rep-
Switching-SurgeCalculationsonUntransposedTrans- resent the multi-fault conditions that can occur in the middle of a long
mission
Lines", I E E E Paper No. C73444-7. . line?
2. In any digital computation of a network, the consistency of
9. A. Semlyenand M. C . Kwok, "CalculationofSwitching numerical stability is always paramount. This is one of the reasons why
Transients onOverheadLineswithFrequency Depend- trapezoidal integration is popular. Although it is not themost accurate
e n t Parameters, Part I-Theory", I E E E Paper No.
C72 method it is numerically quite stable, and incidentally avoids the fast
563- 5. eigenvalue problem. Is the quadratic approximationused by the authors
equally stable?
10.D. E. Hedman, "Distortion of T r a v e l l i n g Waves on 3. h e to corona formation during switchingsurges, a reallife
transmission line taken on a non-linear characteristic - although this is
Three-phaseTransmissionLines", IEEE Paper No.C 73 usually ignored because of its complexity and uncertain status. Have the
409-0. authors given any .consideration to the representation of this non-
linear effect, particularly as it might change the coupling between
ll.K. G. Fehrle,
"Digital Computer Calculations of conductors orcircuits?
Switching
Surge Wave Fronts on EHV T r a s m i s s i o n 4. Could the authors touch briefly on possible means of attain-
Lines", IEEE Paper No. C 73-150-0. ment of variable time steps in their convolution approximation? Vari-
able time steps - available in some integration procedures such as the
12.A. Ametani, ' ! S t r a t i f i e dE a r t h Effects on Wave Prop- Gear algorithm - can result in a significant reduction in computer time
atation-Frequency-Dependent Parameters", IEEE Paper at practically no loss in accuracy. This Was one of the difficulties with
NO. T 74081-6. standard trapezoidal integration - the time step could only be changed
by expensive recomputations.
13.5. A. Robertand K . D. Tran, !'New D i g i t a l Simula- 5. Finally, one of the most tedious and expensive problems with
t i o n f o r Ground-Mode SwitchingSurgeResponse of EHV digital power system transient analysis has been the lack of any good
TransmissionSystems", IEEE Trans. on Power Appara- hillclimbing technique to efficiently search out the maximum voltages
t u s andSystems,Vol. PAS-88, May 1969,pp.597-603. that can occur. Usually the programmer has had to either adopt a shot-
gun and random switching procedure wherein a greatmanycasesare
14. K . D . Tran and J . A. Robert,"DigitalSimulation computed,or elsesearch outthe derivativesbysome cutandtry
routine. Each new method that comes along should be carefully evalu-
and Analysis of
Surges on PolyphaseTransmission ated for itscapability to provide maximum voltages and currents atleast
LineswithEarthReturn!', IEEE Trans. on PoumAppa- computational cost. How do the authors propose to seek out maximum
ratus and Systems,Vol. PAS-91, March/April1972, voltage or current levels with the digital procedure they propose?
pp.452-458.
Manuscript received August 1,1974.
15.A.Dabuleanuand A. Semlyen, "odellingofTrans
p o s i t i o n s andDouble CircuitTransmissionLinesin
Switching SurgeCalculations",
Submittedfor
the
1974 I E E E Summer Power Meeting. R H. Brierley and J. K. Snelson (Ontario Hydro,Toronto, Ontario,
Canada): The authors are to be commended for developing this method
16.A. Semlyen and A. Dabuleanu, "A System Approach t o of reducing the computing time required for transient calculations on
AccurateSwitchingTransientCalculations Basedon frequency dependent transmission lines. The greater efficiency of this
S t a t eV a r i a b l e Component Modelling",'submittedfor method comes from making full use bf the speed of convolving two
t h e 1974 IEEE S m e r Power Meeting. functions when one of the functions is an exponential or a series of
exponentials. This essential fact had beenrecognizedbyMeyer and
17.L. M. Wedepohl, "ApplicationofMatrix Methods t o Dommel (ref 1) and Umoto and Hara (ref 2) but neither of these authors
theSolutionofTravelling-Wave Phenomena i n Poly- had developed the method to the degreeofusefulness and computa-
tional efficiency described in this paper.
phaseSystems",Proc. IEE Vol.110, . Dec. 1963,pp As the authors areaware Ontario Hydro has incorporated their
2200-2212. subroutines into the B.P.A. "TransientsProgram"describedinrefer-
ences 1 and 2 of the paper. In the course of this work additional infor-
18.A. Semlyenand A . Dabuleanu,''Improvingthe Effic- mation came t o light that may prove useful to others.
iency of AccurateSwitchingSurgeCalculations", It is important for any line representation to be accurate down to
I n t e r n a t i o n a l E l e c t r i c a l , ElectronicsConference low frequencies when transients must be studied in the presenceof
and Exposition,Toronto,October1973,Paper No. 60 Hz load or fault currents. To obtain accuracy at low frequencies the
73071. exponential series must fit the time response for a long period of time.
To maintain a good fit for a long time and at the same time retain a good
19.R. Uram and R. W. Miller, :'MathematicalAnalysis fit for the shorter times required for accuracy at high frequencies may
and Solution of Transmission-Line Transients, I- require the use of more exponential terms than used in this paper. A
Theory", IEEE Trans. on Power ApparatusandSys-
tems, Vol. PAS-83,November 1964,pp.1116-1123. Manuscript received July 24, 1974.
569
simple check on the low frequency accuracy can be obtained by deter- recursive convolutions andfor their interesting remarks and contri-
mining the effective DC parameters of the line. butions.
If the unit step response is represented by: Mr. Anderson’s questions will be answered in the order they were
posed.
--t-TO -- t-To t-To 1 ) The suggestion of modal network interconnection due to un-
Al(1-e + A 2 ( 1 - e T2) . . . A n ( ] - e T n ) balanced faults has the appeal of making complex conditions intuitively
more understandable. Computationally, however, we prefer to revert to
and the admittance response is represented by: phase quantities at line terminals or at line sections resulting by trans-
positions or faults. The fault itselfis represented by anadmittance
matrix. This ismainly due tothe fact that modal transformation
matrices are different for most system components. They are not
sufficiently simple(like transformation into symmetrical or Clarke
then, as pointed out in the paper, for a line with zero shunt conductance components) to permit direct interconnections of modal networks
the step response tends to unity and the admittance response tends to (without usingideal transformers) and ifsuch interconnections were
zero as time tends to infinity. feasible they would have significance only for the faulted part of the
This requires that: network.
2) The discusser isquite rightin his remarks concerning the
A1 + A2 . . . + An = 1 trapezoidal integration. The quadratic approach permits to increase the
time step but losses have to be considered in modelling (this is always
-
Yo y1 -y2 . . . -ym = 0 the case with a transmission line) in order to avoid possible numerical
instability.
With these conditions the effective DC resistance of the line is: 3) Corona is an important problem of switching transients and a
difficult one becauseof its non-linear nature. At the Universityof
Toronto an approximate method of solution is under investigation,
consisting of subdividing the line into linear, coronaless sections, and
placing lumped nonlinear corona models between these line sections.
4)Our present program has no provision for changing the time
Examination of this expression shows that exponential terms with small step but we believe that this feature wouldlend it better versatility,
amplitudes but very long time constants can have significant effects on Indeed, it would permit to calculate transients with good resolution and
the effective DC resistance. Representation of such terms should not be obtain rapidly steadystate solutions, as well. It would involve recalcula-
difficult with the authorsexponential series approach as adding an tion of some coefficients (which would not require much computer
extra exponential term does not greatly increase the computing time. time) and use, for instance, as inputs un-2~1n-4instead of Un-1 , ~ n - 2in
However, if the response functions are represented on a point by point eqn. (5) if At is doubled.
basis long duration responses require excessive computation time and 5 ) Maximizingovervoltages or getting statistical outputs for all
storage. possibleswitchingangle (e) combinations or evencalculatingrisksof
The short duration third exponential term in Fig. 1 is neglected by failure seem to be time consuming tasks. However, due to linearity, any
theauthors which introduces a discontinuity of slopein the step input condition6 can be obtained by a linear combination of two inputs
response and a fictitious extra delay. Comparison of Figs. 3 and 4 shows in quadrature. The latter yield outputs which are stored and then pro-
that the new method retains sharp discontinuities for longer times than perly recombined. We wish to note that statistical insulation coordina-
the phase convolution method. Could this retention of discontinuities tion is more rigorously achieved if it is based not solely on maximum
be caused by the neglect of the third exponential term? overvoltages but also on smaller values. A study on the above problem
Test studies indicate the calculation time for a 3 phase unbalanced is now partially completed at the University of Toronto and details will
line with all three modes frequency dependant is approximately twice be reported in a forthcoming paper.
thatfor a similarlineconsidered continuously transposed and non- It is gratifying to see that Ontario Hydrohas found our methodof
frequency dependant. This is approximately one fifth of the time used representing transmissionlines interesting enough to be included in
by our earlier frequency dependant representation which of necessity BPA’s “Transients Program”. We arealso thankful to Messrs. R. H.
considered the line continuously transposed, and which did not have the Brierley and K. Snelson for privatediscussions which enabled us to
desired accuracy over long time periods. improve the accuracy and versatility of the computer program. These
Another test indicated that a change from two to four exponential improvements will be indicated in our answer to their discussion.
terms for the line step responsesincreased the calculation time by As mentioned by the discussers, low frequencies are necessary for
approximately 5%. This indicates thatthe major computing effort accurate step responses at longtimes.Thisisespecially tme for the
involves the phase to mode transformation rather than the convolution modal characteristic admittance step response which, in the frequency
calculation, and can therefore be attributed to the representation of domain,presents sharp variation in the region from zero to a few hertzs.
unbalanced lines ratherthantothe representation of frequency de- This was mentioned by Meyer and Dommel in a recent paper[A] and
pendance. canbe demonstrated analyticallyeasilyby noting that in the inverse
The calculation of the step response characteristics of the line is Fourier transform only the real part of YC is needed and that Re (Yc) is
done in a separate program, much like the calculation of line impedances. proportional to thevelocity of the respective mode.
The fitting of exponential curves to the calculated response in the time Numerical results indicate that at 0 Hz the velocity is zero, but
domain is at present very expenisve of computer time. It would appear after only a few hertzs it shoots up toapproximately 200,000 km/s for
reasonable to expect that in future a program could be developed to fit the ground mode and 290,OOQ km/s for the line mode.
the calculated responsein the frequency domain with functions that This clearly indicates that the region of low frequencies has to be
would transform analytically intoexponentials in the time domain. explored in detail, with small frequency steps, if accurate step responses
This would avoid the expensive inverse fourier transform. It is also our are t o be obtained. For D50-100 Hz the frequency step used in per-
belief that errorsin fitting in the frequency domain are more significant forming the inverse Fourier Transform can be increased considerably.
and interpretable by users than an error in the time domain fit. One We alsoagree with Messrs. Brierley and Snelson that lines closed
would then know whether a particular frequency would, or would not on loads or faults have to have a very good representation at 60 Hz. As
be correctly represented. the discussers pointed out, this canbesolvedbyusing an extra ex-
ponential. We found necessary to increase the number of exponentials
REFERENCES only for modal characteristic admittance step responses, from one to
two. The first exponential fits the curve for high frequencies (or short
[ 1 ] W. S. Meyer and H. W. Dommel, “NumericalModellingof Fre- times) and the second one takes care of thetail of the curve (low
quency-&pendant TransmissionLine Parameters in an Electro- frequencies). Moreover, we introduced a constraint so that the sum of
magnetic Transients Program”. IEEE 1974 Winter Power Meeting. the two exponentials, in the frequency domain, should be equal to the
[2] J. Umoto and T. Hara, “A NewDigitalAnalysisof Surge Per- 60 Hz value of Yc. Using the notations given in the discussion, this can
formance in Electric Power Networks Utilizing the Convolution be written as
Integrale” Electrical Engineering in Japan, Vol. 91, No. 3, 197 1,
pp. 48-57. y1 +- y2 = Y c , w = 2n60 (40)
j w +tL l J’ W + l jw
t2
A. Semlyen and A. Dabuleanu: We wish to thank the discussers for their Messrs. Brierley and Snelson pointed out that increasing the num-
questions which permit us to clarify some details and perspectives of ber of exponentials from 2 to 4 has increased the computer time only
by 5%. We feel that this result indicates thatthecomputer time
Manuscript received October 31, 1974. required to calculate the current source for the Norton line equivalent
5 70
at each time step is merely a mall fraction of the overall computer time b) for fust order input
which includes such components as reading in data and storingthe
output.
We do not know exactly the answer to the question of why the
output plot of the transients calculated withphase convolution presents
A‘ = A - e ( 1 -a)
peaks more rounded than the one obtained from the recursive convohr-
tion. We consider that these differences are really small, and feel that a
more likely cause than the one indicated by the discussers would be the
fact that the numerical integration in the phase convolution method is
p‘ = p + + E (1 -a)