# CALCULATION OF VOLTAGE PROFILES ALONG TRANSMISSION LINES

L. Marti (M)
Ontario Hydro Canada

H.W. Dommel (F)
The University of British Columbia Canada

ABSTRACT
A method for calculating transient voltages and currents at a large number of intermediate points along a transmission line is presented. In the family of EMTP-type programs, these "profile" calculations are normally made by explicitly connecting a number of short line sections to create intermediate nodes where voltages and currents can be monitored. This approach is time-consuming, computationally intensive and can be inaccurate if the number of intermediate nodes is large. The model presented here allows the accurate evaluation of voltages and currents at an arbitrary number of equally-spaced points along the line. The frequency dependence of the line parameters is taken into account. It is computationally fast and easy to use. 1. INTRODUCTION Switching surge voltage transients are an important factor in insulation coordination studies of overhead transmission lines. In such studies, the overvoltages are normally obtained at the sending and receiving ends of a line. In most instances, overvoltages are most severe at the ends of the line, due to the combined effects of incident and reflected waves. There are situations, however, where overvoltages at intermediate points along the line are higher than those at the ends of the line. In such cases, it becomes necessary to calculate a "profile" of voltages as a function of distance, to determine the maximum overvoltages, as well as their location along the line. Such a case arises if metal oxide surge arresters are at the ends of the line, instead of pre-insertion resistors in order to limit switching surges in extra high voltage lines [I], [2], [7], [ l 11. During line energization or re-energization, a metal oxide arrester effectively limits the overvoltage at the protected end of the line below a fixed value. This moves the location of the maximum overvoltage away from the end of the line, which makes it necessary to calculate the voltage profile along the line. Profile calculations are also needed to find the sheath overvoltages in high voltage cables between solidly grounded points, or between points grounded through metal oxide

arresters. Another application is the calculation of voltages induced from adjacent energized lines along intermediate points along grounded lines which have been taken out of service for maintenance. Profile calculations have also been used to produce travelling wave movies for educational purposes [3,4,5].These movies show transient phenomena as a function of time and space. The traditional way for obtaining voltage profiles in programs such as the EMTP is to calculate the voltages at intermediate nodes, which are created by connecting a number of shorter line segments together [ I l l . This is time consuming, especially when frequency dependent line models are used. If the line is broken down into a large number of sections, the accumulation of errors degrades the accuracy of the simulation. Another disadvantage of the traditional method for calculating profiles is the fact that the step size of a transient simulation must be smaller than the travel time of the shortest distributed parameter line. If the time step has to be reduced to accommodate the shorter line segments, CPU time, as well as storage requirements for past history terms for all the lines would increase. The method for profile calculation presented here addresses the major disadvantages of the traditional method. It is computationally faster, more accurate, and it does not tie the step size of the entire simulation to a value that depends on the spatial resolution of the profile. Since it is based on the JMARTI [6] line model, it takes the frequency dependence of the line parameters into account. This method was first implemented in 1982 as the Frequency Dependent Profile Model "FDPROFILE" in the University of British Columbia version of the EMTP. In this implementation, FDPROFILE calculations were imbedded in the time step loop of the EMTP. Because of the recent interest in profile calculations, the original implementationfor balanced lines has been extended to handle untransposed lines. A "stand-alone" version of the FDPROFILE program, which is independent of the particular version of the EMTP used, is under development.
2. CALCULATION OF VOLTAGES AND CURRENTS

AT INTERMEDIATE POINTS
The FDPROFILE model needs the voltages and currents at the endpoints of the line at each time step of the simulation as input parameters. These voltages and currents at the

0-7803-3522-8196\$5.00 0 1996 IEEE

263

) and B.. (10) Let F. it is an approximation. and B. = F.... phase voltages and currents are found from the inverse relationships of equations (1) and ( 2 ) . xj = j-Ax (3) 2ZcIx = F.... TIME DOMAIN CALCULATIONS v i r t I v T .. + ZcId = Fd (1 1) Vo - ZcIo = Bo (12) F. be the forward and backward travelling functions evaluated at the sending and receiving ends of the line. In the calculation of the voltages and currents at N equallyspaced points along the line. Consider then... V. 264 . After the solution in the modal domain has been obtained. Addition and subtraction of (5) and (6) produces 2 V. the voltages and currents at a point x are given by V. .+. If the solution of the line at the endpoints is known. Therefore. V. since it is based on similar mathematical representations as the FDPROFILE model. = A F. and B. (5) (6) - = where Z. F.. Single-phase transmission line. that is..2. Because the FDPROFILE calculations take the frequency dependence of the line parameters into account. = F.. it is assumed that the modal transformation matrix [Q].. (9) With this transformation. or can be calculated from the voltages and currents using equations ( 5 ) and (6).. ZcZ. = F.. Let xi (j=1.... I I Xi c X x2 xj I Fig.. 1) AX = d N+l (4) from which the intermediate voltages and currents can be readily obtained using equations (9) and (10). the number of circuitsltowers on the same right-of-way. that is.... 3.. F. N . I . where N Voltages and currents are to be calculated at equally spaced points. and B.. = Bxd. is the characteristic impedance and A is the propagation function for a line segment of length Ax. The accuracy of this approximation depends on the geometry of the tower. and B. is real and constant.....N) be the distance from the sending end. and B. but this model has not yet been extended to overhead transmission lines.. = = AF. the single-phase line or a single mode of a multiphase line of length “d“ shown in Figure 1. There is an EMTP model which takes into account the frequency dependence of [Q] in directly buried cables [IO]. BjAx = A B(f+l)Ax (j = N-1. B. with FA. an n-phase line is then analyzed as a set of n single-phase lines in the “modal domain“.. can then be calculated from F.. I . + B.2 . The preferred host model is the JMARTI line model of the EMTP... and the type of surges [9].. are defined as the forward and backward travelling functions. + ZcZ. All overhead transmission line models presently implemented in the EMTP assume that [Q] is constant and real. are either a byproduct of the solution process of the host model. though work on this extension is in progress. in order to produce consistent results. which diagonalizes the matrix product [Yphue] [Zphase].. - B. the host line model should also take frequency dependence into account. The assumption that [Q] is constant is only strictly correct when the line is “balanced” or perfectly transposed._. otherwise... I f .. F.. AB... Phase voltages and currents are converted into modal voltages and currents with the linear transformation defined by In the frequency domain..terminals of the line must be calculated with a program such as the EMTP with a “host“ line model.

the effective length and the attenuation of the line can change significantly. d. To evaluate equations (16) and (17) at time t. regardless of the time step At' used by the host model. This step size is chosen so that the travel time T of a line section of length Ax becomes an integer multiple of At. the step size must be decreased by a factor of TIAt'. Since phase voltages and currents are output quantities not used in any additional FDPROFILE calculations. Storage requirements for the past history terms of all distributed parameter lines in the system will also increase by the same amount. the following assumptions are implicit = k=l Ck\$Ax. Since the solution of the line at the end points is provided at each time step by the host line model.287 ms. and 7 is the travel time of the fastest frequency component which propagates along the line. In the FDPROFILE model. the global time step of the entire simulation is unaffected by the number of intermediate voltages. From equation (lo). however. When the propagation function of a line segment of length Ax is used in FDPROFILE calculations or when N+l line segments are modelled explicitly. With the FDPROFILE model. ( t ) + h. the convolutions in (14) can be expressed numerically as m &A.17 ms.k(t-At) + dkfu-l)Ax. past history values of f. but rather a line that is 10% shorter for zero sequence calculations. an "internal" step size At is used. for large N. must be known at t-7-At. f. and the storage requirements of other lines in the system are unchanged. the intermediate currents in the time domain can be calculated from e.(t) [6]. where (n+l)At > T > nAt. If dAt is not an integer number. and execution time increases accordingly.&t) EMTP. while T for the 500 km line is 3.(t-z-At) have to be calculated by interpolation of recorded values at t-nAt and t(n+l)At. of the form where the poles and zeroes of (15) are real and lie in the right hand side of the complex plane. for the double-circuit line used in the examples shown in this paper.@) = zeqi .(t-At) where zq is constant and h. errors in FDPROFILE calculations do not propagate from time step to time step. Intermediate voltages at time t are then obtained directly from To calculate intermediate currents.(t-z-At) and b. The FDPROFILE model is relatively insensitive to the quality of the rational functions approximation of the propagation function A. and linear interpolation is needed to calculate phase voltages with equations (1) and (2). profile calculations are always local to that time step. Figure 2 shows the errors in the approximation .. it is also necessary to approximate the characteristic impedance as rational functions where the poles and zeroes of (19) are real and lie in the right hand side of the complex plane. In the case of multiphase line models in the EMTP. 7IAt is always an integer for all modes. the zero sequence value of z for a 50 km section is 0. On the other hand. In the solution of equations (16). The use of an internal step size for profile calculations reduces storage and CPU requirements. (17). and ek are constants which depend on the type of integration rule used.(t-At) depends on past history values of e. the step size of the entire transient simulation must be smaller than the travel time of the shortest distributed parameter line in the system. Therefore. In other words.k(t-t) (16) where ck.(t) and i. For example. and (22). and b.The propagation function for a line segment Ax is approximated with rational functions. When short line segments with travel times 7 smaller than At' are used to calculate intermediate voltages. This reduces interpolation errors in the evaluation of modal voltages considerably . In the 265 While errors in the approximation of A(o) are normally very small. interpolation errors become significant. If the magnitude of At is comparable to the magnitude of 2. the amount of error due to these interpolations is small and does not accumulate from time step to time step. it is not possible to choose At' such that dAt' is an integer for all modes and for all lines. This means that ten 50 km sections do not represent exactly a 500 km line. a different At for the solution of each mode results in voltages evaluated at slightly different points in time.

SIMULATION RESULTS To illustrate the differences between the FDPROFILE model and conventional profile calculations. the performance ratio for a 10-point profile is 0. errors do not accumulate from time step to time step. other than the number of intermediate points. Since at any given time step f. errors are distributed more or less uniformly through all intermediate points. (17) and (22). the intermediate voltages and currents are calculated inside the time step loop of the EMTP. Figure 3 shows the one-line diagram of the .(t) and bd(t) are calculated from equations (9). a simulation with the FDPROFILE model can be 2 to 8 times faster than a simulation with explicitly segmented lines.(t) .z. Since the characteristic impedance is the same for the FDPROFILE model and the host model. The only additional information required. and system configuration.15. and there are 2 convolutions with a(t) for each intermediate voltage. possibly with host models other than JMARTI. IMPLEMENTATION AS A POST-PROCESSOR The main advantages of implementing the FDPROFILE model inside the EMTP time step loop are computational speed and simplicity of the interface.(t) = vd(t) + zeqid@) v.(o) is roughly the same and that only intermediate voltages are requested. To initialize past history terms in (16). magnitude errors accumulate exponentially and errors in the evaluation of T accumulate linearly with N.(t) per section For N=10. then there is the additional overhead of one convolution with z. of the magnitude of A(o) for a 50 km segment (dashed trace) and the effective errors of ten 50 km segments (solid line). the energization of a double-circuit line has been simulated using the EMTP-based implementation. In FDPROFILE calculations for voltages. since the calculation of F. and management of the large amount of output quantities produced. A quantitative estimate of the relative computational speed of the FDPROFILE routines can be obtained from the number of numerical convolutions evaluated per time step. ' dT ri ~ o 2 4 io 4.. '1" 1 '2 'A''i T 2 Id" i T 2 'd" i 6 2 T r l T r l .solid: ten 5 km sections: dashed: one 50 km section. it makes sense to move the profile calculations to a postprocessing program. 3. Also. profile calculations require extra memory for storage of past history terms. IMPLEMENTATION IN THE EMTP In the implementation of the FDPROFILE model in the UBC version of the EMTP. is the approximation of A(o) for a line segment of length Ax. However. A post-processing version of the FDPROFILE model is under development. Even though the approximation of A(w) is remarkably good.i.(t) + hd(t-At) (27) = . A post-processing program also has the advantage that it can be used in more than one version of the EMTP. In versions of the EMTP where memory is limited.-2 N + 4 r = N+1 line sections 4N + 4 In the explicit connection of N+l sections there is one convolution with a(t) and one convolution with y. The implementation as a post-processor will be somewhat slower than its EMTP-based counterpart because of the additional convolutions for calculating f. and 0.h.(t) for both sending and receiving ends. If the intermediate currents are also requested.77 if voltages and 266 fJt) b.00 i currents are requested (assuming that the simulation time step is smaller than the travel time of the shortest line segment).54 if only voltages are requested. steady-state conditions at the endpoints prior to time zero must also be known. that is. the advantages in terms of compatibility justify the small loss in performance. 2: Relative error for the approximation of A(o). From the voltages and currents at the endpoints.(t-At) Both types of implementation are equally accurate as long as voltages and currents are known with full precision. Depending on the amount of output requested.(t) and bd(t) are intermediate variables readily available from the JMARTI line model. sending and receiving end voltages and currents at each time step are required.00 i r = N-point profile 3N -~ N + l line sections 4N + + 4 4 -5. 5. In order to be compatible with most versions of the EMTP and. (10) and (21) Frequency (Hz) Fig. On the other hand. While the FDPROFILE model is also affected by the accuracy of the approximation of A(w). equation (18) can be used directly. it does not have to be approximated again. then the performance ratio is N-point projile .(t) and bd(t). there is an additional averaging effect. for a total of 4 convolutions per section. f. and B. start at opposite ends of the line. 4 convolutions account for the solution at the endpoints. Actual execution speed is much higher than what these ratios indicate. Assuming that the order of the approximations of A(o) and Z.

1 50. energization with arrester. dashed: two 250 km sections.50 p. . -1. One three-phase circuit is energized.0 46.2. .0 48. in a 10-segment case. 1.( Time (ms) Fig 5: Midpoint voltages. I 1. . 2.00:.2 . .( 2. Solid: FDPROFILE model. .0 .0 . . . I . -1. .0 5 . . gains in computational speed also increase with the number of intermediate points and the complexity of the rest of the system. . I .OOl.00- Fig. . .0 . . 0 0 . Solid: IO-section model. / I 1 2. 2 -1. . However. dashed: two 250 km sections. . Errors in the traditional approach become more significant at higher frequencies (see Figure 2) and when the number of intermediate points increases. the FDPROFILE model produces more accurate results that the traditional approach. . Figure 6 shows the profile of maximum overvoltages for both the energized and the grounded circuit. 40.0 44. I . and voltages are calculated at 50 km intervals.0 200. energization with arrester. The reference solution has been obtained by connecting only two 250 km sections (dashed trace). I . 40.0 500. .0 46. I I .0 Time (ms) lo. 42. -2. .0 400. while the other circuit is grounded at both ends. . . . . . . 1 1 . \ / v . Distance (Ian) Fig 6: Profile o f maximum overvoltages -0. 267 FDPROFILE simulation was only 20% faster than the traditional approach. I . Figure 5 shows the results obtained with a nine-point FDPROFILE model. 0 1 . . I .0 6. Figure 4 shows the voltages measured 250 km from the source when ten 50 km line segments are connected to obtain the intermediate voltages (solid trace). 3: Test circuit for line energization with arresters test system. I . . 8.0 . . .0 rl u m m O. The energized circuit is terminated with surge arresters to limit voltages to 1.u.00. . . It can be seen from these results that even -Z.50h Time (ms) 2.\. . ooT-- m 1.0 42. .00.50- -1. The line is 500 km long. . It can be seen from this simulation that the FDPROFILE is virtually identical to the reference simulation solution. . I I I .0 4.0 300.00 0. the Time (ms) Fig 4: Midpoint voltages. . For this particular example. . 48.5c-j . 0 100.0 44.50: -3 & - - & - I & - Y 0 -0. I .

From 1959 to 1966 he was with the Technical University Munich . 3. PAS-94. "An Interactive Package for Electric Power Engineering Education". Canadian Electric Association Meeting. January 1989. The frequency dependence of the line parameters is taken into account. January 1989. Marti.25 I I y(ft) 120. Ing.H Brunke. 10. He received the Dipl. Transient Calculations on Transmission Lines with Ground Retum Using Recursive Convolutions".1862-1874. J.S. [2] J.A. IEEE Transactions. Vol. 1. "Simulation of Transients in Underground This paper describes the calculation of voltage and current profiles on transmission lines when the solution at the endpoints is known. Semlyen and A.R. Ray. . no. 3. Atmuri. The Parameters". "Voltage and Current Profiles and Low-order Approximation of Frequency-dependent Transmission Line f Electrical Engineering. Hermann W. 147-155. S. vol.65 Wmi Diameter of ground wires 1. The calculation of the intermediate voltages and currents is done in the modal domain. Cascais. No. Earth resistivity 100 S Z m cond 1 I I x(ft) -18. MASc and PhD degrees in Electrical Engineering in 1983 and 1987. Woodruff.6. double circuit.. J. 22-27.J. Marti. Marti. A stand-alone post-processing version which is compatible with most versions of the EMTP is under development. "An Application of Metal Oxide Surge Arresters in the Elimination of Need for Closing Resistors in EHV Breakers". L. 1. degrees in electrical engineering from the Technical University Munich.Grid System Strategies & Plans Division. Yasuda. 416-424.. 131 L.05215 a m i Diameter = 1. 10991110. 3. Portland.005 Q). October 1995. Selim. 57. Oregon. Portugal August-September 1987. Legate. and joined Ontario Hydro in 1989. [ I l l J. Ontario. pp. Germany in 1959 and 1962. Salon.602".Vol.C. "Elimination of Closing Resistors of EHV Circuit Breakers". pp. McCallam. IEEE Transactions on Power Delivery. Ninth Power Systems Computation Conference. Dommel. "Accurate Modelling of Frequency-Dependent Transmission Lines in Electromagnetic Transient Simulations". 391-400. pp. 561-571. [5] S. IEEE Transactions on Power Delivery. pp. pp. The arrester in Figure 3 is modelled as a nonlinear resistance with a simple two-slope characteristic (-. 4. pp. He did post-doctoral work in cable modelling in 1987-1988. January 1982.W.R. "Application of Metal Oxide Arresters for the Control of Line Switching Transients".J. Toronto. Marti. Dr. and from 1966 to 1973 with Bonneville Power Administration . It offers an accurate alternative to the time consuming and computationally slow process of subdividing a line into smaller segments. 147-157. 223-231. J. under the assumption that the modal transformation matrix [Q] is real and constant. M. Dabuleanu "Fast and Accurate Switching The tower data for the transmission line and simulation data used in the simulation described in Figure 3 is as follows: Voltage class 230 kV. Cables with Frequency-Dependent Modal Transformation Matrices".R. CONCLUSIONS [ 101 L. IEEE Transactions on Power Apparatus and Systems.3636) dc resistance of (solid) ground wires = 2. [9] H. [8] A.. [7] R. IEEE Transactions PAS-101. and Vsat = 1. Dommel (F '79) was born in Germany in 1933. "Design Studies for the Mead-Phoenix 500 kV ac Transmission Project". BIOGRAPHIES Luis Marti (M'79) received an undergraduate degree in Electrical Engineering from the Central University of Venezuela in 1979.5. "Transmission Line Transients in Motion Movies". Canada. IEEE Transactions on Power Delivery.J. July 1988. Brandwajn. E. Ing and Dr. pp. IEEE Transactions Power System Delivery. APPENDIX [4] L. Thesis. MarchlApril 1975.H. vol.E. Department o University of British Columbia. May 1990. Lundquist.602" (thicknesddiameter = 0. from The University of British Columbia. no. V. Transactions AIEE.R. respectively. Brunke. vol.F.G. 268 The source impedance used in Figure 3 is 10 SZ. Insulation Coordination Seminar. Canada REFERENCES [I] A. where he is currently working in the Technical Support Department . MASc. The FDPROFILE model is computationally fast and easy to use. Dommel is a Fellow of IEEE and a registered professional engineer in British Columbia. Ribeiro. [6] J. Tallam. pp. Since July 1973 he has been with the University of British Columbia in Vancouver. No. pp. April 1982. July 1938. T. respectively. Number of conductoribundle = 3 Separation of conductors in bundle = 18" dc resistance of main ACSR conductors = 0. "Approximate transformation matrices for unbalanced transmission lines". Marti. D. January 1982. 4.