IEEE Transactions on Power Systems, Vol. 4, No.

3, August 1989 1293

TJSING A PERSONAL COMPUTER T O TEACH POWER SYSTEM TRANSIENTS

R. M. Nelms, G. B. Sheble', Steven R. Newton, and L. L. Grigsby

Department of Electrical Engineering
Auburn University, AL 36849

ABSTRACT - This paper presents a state variable technique One popular computer program utilized by the electric
for teaching power system transients using a personal utility industry is the Electromagnetics Transients Program
computer. Transmission lines are divided into a series of (EMTP) [I], which is based on the ideas of Dommel 12-41.
pi-section segments. Each segment consists of a series Simple equivalent networh,,are derived for all components in
resistance and inductance and a shunt conductance and the system. Nodal equations are formulated for the system
capacitance. Using this line representation, a state model is represented by the equivalent networks. The transient is
formulated for the power system using the capacitor voltages calculated through repeat solutions of these nodal equations.
and inductor currents as the state variables. The state Since EMTP is not user friendly, the student can easily become
equations describing the system are transformed to a set of frustrated with it. In addition, EMTP requires an extensive
linear difference equations through the use of trapezoidal background in power system analysis and will overwhelm the
integration. The state variables are updated by solving this set beginning student. Glover and Sarma [5] present a technique
of equations. Nonlinear elements such as surge arresters may similar to that used in EMTP for the solution of transient
be included in the analysis. The technique presented here problems. Their computer program is much more user friendly
utilizes the student's knowledge of network theory and is easily than EMTP and, therefore, is much easier for the student to
implemented on a personal computer. use. The classical lattice diagram technique [5,6,10] may be
used in a study of transients. This technique uses an
KEYWORDS: education, personal computer, state variables, approximation to the travelling wave equation. The student
power system transients needs to understand the concepts of reflection coefficients and
impedance matching, which are not network concepts. In
INTRODUCTION addition, it is very difficult to incorporate resistive effects.
Electromagnetic transients in power systems may be The lattice technique appears to be more difficult to
produced by switching actions, faults, or lightning strikes. implement on a computer than the concepts behind EMTP [3].
Students may investigate the effects of transients by using a The technique presented in this paper utilizes the
tiansient network analyzer or a digital computer. A transient student's background in network theory. It requires the
network analyzer requires the construction of a power system formulation of a state model for the power system. Therefore,
analog. Computer-based transient analysis techniques the student only needs a course in network theory as a
necessitate developing a mathematical model to describe the prerequisite. Many network textbooks such as [7] present
transient behavior of the power system. These techniques are techniques for deriving state models for networks. The state
especially attractive for student investigation of transient variables are selected to be the capacitor voltages and inductor
phenomena because of the availability of low cost personal currents in the network. The state equations describing the
computers. This paper presents a state variable technique network can be written in the form
which can be used to teach a senior elective course in power
system transients using the personal computer.
& = A s + Bu

where is a vector of state variables, U is a vector of inputs,

89 WM 044-9 PWRS A paper recommended and approved
by the IEEE Power Engineering Education Committee of and A and B are matrices. Equation (1) may be solved using
the 'IEEE Power Engineering Society for presentation many techniques. Trapezoidal integration [7] is employed here
at the IEEE/PES 1989 Winter Meeting, New York, New York,
.January 29 - February 3 , 1989. Manuscript submitted to transform the state equations to a set of linear difference
September 1, 1988; made available for printing equations [SI
November 14, 1988.

0885-8950/89/08OO-1293$01.OO 0 1989 IEEE

1294

where b'[kT] = f(x[kT], u[kT], u[(k+l)T]). A' is a sparse

.if;
mat,rix which is constant for a fixed time step T. The vector
x[(k+l)T] is solved for by either finding the inverse of A ' or
performing an LU factorization of A' to preserve the sparsity.
The latter is utilized in this paper because of the special
2 2
structure of the A' matrix. Since A' is constant for a fixed
time step, it is only factored once at tlie beginning of the
simulation. The right hand side of (2) is recalculated at each
time step. Figure 1. Pi-network section for transmission line
Nonlinear devices such as surge arresters are a part of The 11 pi-networks are connected in series as shown in
electromagnetic transient studies. In [4], an iterative Figure 2 to form the transmission line model. Note that the
procedure is utilized to find a solution when a nonlinear device shunt elements of adjacent pi-networks are combined in
is present in the system. In this paper the nonlinear device is parallel. The capacitor voltages and inductor currents are
treated as a piecewise linear resistaiice. The A matrix is designated as the state variables. Examination of Figure 2
formrilatetl in symbolic form and is evaluated for tlie different indicates that the state equations for the transmission line
values of the piecewise linear resisthiice. All A and A ' matrices have a special structure. The state for each of the shunt
are evaluated and factored or inverted before simulation capacitors is affected only by the two inductors connected to it
begins. and the capacitor itself. This will be demonstrated more
This state variable technique is easily implemented on clearly with an example system in a later section.
the personal computer. The student only needs a routine to
solve (2). Sparsity techniques can be ut,ilized in t,he LU
fact,orization of A' to reduce memory requircmeiits. Since A ' SOLUTION PROCEDURE
only iiceds to be factored once and iterative procedures are The model of the transmission line presented in the
unnecessary, the transient may be calculated quickly on a previous section is combined with other component models.
personal computer. This technique is illustrat,ed with a simple State equations are formulated for the system and can be
example. written in the form of (1). This set of linear ordinary
differential equations is transformed into a set of linear
TRANSMISSION LINE MOnEI, difference equations using trapezoidal integration [8]. If T is
The state variable technique discnssed here requires an the time step, equation (1) becomes
equivalent network from which state equations can be written.
The transmission line is modeled as an interconnection of n
pi-networks. Each pi-network contains a series resistance and
inductance and a shunt conductance and capacitance as seen in
Figure 1. RI, L', C', and G' are the resistance, inductance,
Ax[k+l] + Bu[k+l] + Ag[k] + Bu[k]
I
The indices k and k + l are utilized to indicate the values of
(3)

conductance, and capacitance per unit length, respectively, of or at time t=kT or t = ( k + l ) T , respectively. Rearranging (3)
the transmission line. The resistance and inductance for each yields
pi-network are determined by dividing the total resistance and
inductance for the line by the number of pi-networks n. The
shunt conductance and capacitance for each pi-network are
determined in the same manner. One half of the calculated
values is assigned to each end of the pi-network in Figure 1.

Figure 2. Transmission line model

 1295

If the inputs 11 are known for all discrete points in time Since the surge arrester and transformer are connected to
t=kT for k = 0, 1, 2 . . ., x[k+l] can be determined using (4). one end of the transmission line, the shunt conductance and
Define capacitance at the end of the line in Figure 2 become (G/2 +
l / R A ) and (C/2 + CT), respectively. If the state equations
are formulated such that the odd-numbered state variables
(5)
correspond to capacitor voltages and the even-numbered state
variables correspond to inductor currents, the following A
matrix results

All matrices in (5) are constant for a fixed time step;

they are evaluated before simulation begins. An LU
G 2
factorization is also performed on A' before simulation begins.
.T -7
The vector y [ k ] must be evaluated at each discrete point in
1 R 1
time.
T --G -T
Because of the transmission line model employed here, A
1 G 1
is a tridiagonal matrix. Therefore, A' and A" are also T -T -T
tridiagonal. Only the nonzero elements of A" and B' are A =
stored to decrease memory requirements and execution time.
In order to retain the sparsity of A, Crout reduction [9] is
utilized to obtain an LU factorization of A' considering its
1 R 1
tridiagonal nature. Crout reduction may be applied to a T -77 -t
system of linear equations if the diagonal entries of L are
nonzero [9]. For a set of n equations in n unknowns, Crout
reduction requires only (5n - 4) multiplications/divisions and
(3n - 3) additions/subtractions to solve the set of equations.
This can result in considerable savings in execution time for
large n. Only the nonzero entries of A are shown in (7). Note that the
The simulation procedure will now be summarized: A matrix is tridiagonal for this system as was previously
stated. With the exception of the first and last rows, the
1. Calculate A', A", and B' from (5). elements of the superdiagonal alternate between (-1/L) and
2. Compute LU factorization of A'. (-1/C), and the elements on the subdiagonal alternate between
3. Set k = 0. (1/L) and ( l / C ) .
4. Calculate b'[k] from (6). For the n section line in this example, there are 2n+l
5. Solve for x[k+l] using (4). state variables. Since the square wave current source is the
6. k = k + 1. only source in this example, only one entry in B is nonzero.
7. Repeat 4, 5, and 6 until end of sirnulation. This current source is connected across one of the capacitors;
therefore, this entry must have an odd subscript because the
EXAMPLE SYSTEM capacitor voltages are the odd-numbered state variables. The
The state variable technique presented in this paper is nonzero entry in B corresponding to the current source will be
illustrated using an example from [ 5 ] . In this example, a designated bn+l and has a value of ( l / C ) .
lightning strike occurs at the center of a 20 kV, 10 km line Two simulations are performed to illustrate the state
which has R' = 0.05 Q/km, G' = 0.556 pS/km, L' = 1 mH/km, variable technique. The transmission line is divided into 50
and C' = 11.11 nF/km. A transformer is connected to one end pi-networks and a time step of 0.05 ps is employed. The
of the line which is modeled by a capacitance to ground of example system is first simulated without the surge arrester.
CT = 6 nF. The other end of the line is terminated in an open The simulation results are given in Figure 3. A second
circuit. The lightning strike is modeled by an ideal square simulation is performed with the surge arrester included. Two
wave current source connected to a node in the middle of the A' matrices, one for RA = 2 MQ and another for RA = 4.5 Q ,
line; the source has a magnitude of 20 kA for 20 ps. The are evaluated and factored for this simulation. When the
transformer is protected by a surge arrester whose v-i curve is voltage across the surge arrester exceeds 55 kV, its resistance is
approximated as: RA = 2 MQ for VA < 55 kV and RA = 4.5 Q changed from 2 MQ to 4.5 0. The A' matrix corresponding to
>
for VA 55 kV. the lower resistance is now used in the simulation until the
I296

,,,,,,,,,,,,,,,..,,1,,.1111111,,,illlll,,l,,l,l,l,,,,,,,,,,,,,,.,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,.,,,,,.,,~~ , . , , , 1

I
I I I
!
I I
! I
!

Figure 3. Transformer voltage without surge arrester

Figure 4. Transformer voltage with surge arrester

surge arrester voltage drops below 55 kV. Figure 4 shows the can be determined by repetitively solving this set of linear
results of this simulation. Figures 3 and 4 compare favorably equations. This technique can be easily implemented by the
with the results given in [5] and with results obtained from student on a personal computer. A subroutine to solve a set of
EM T I’ . linear equations is all the student needs. The technique can
also be extended to include nonlinear elements such as surge
CONCLUSION arresters.
This paper has presented a state variable technique for
teaching power system transients. Transmission lines are ACKNOWLEDGEMENT
inodeled as a cascade connection of pi-networks. Using this This work was supported by NASA through the Commercial
line model and other component models, the student can Center for the Development of Space at Auburn University.
formulate a state model for the power system using elementary
network theory. The state equations are converted to a set of
linear difference equations using trapezoidal integration. If the
inputs are known for all time, the response of the power system
 1297

REFERENCES R M. Nelms (S 78, M 82) received the B.E.E. and M.S.

-
The Electromagnetic Transients Program fEMTP) degrees in Electrical Engineering from Auburn University in
Version 1.0, Electric Power Research Institute, Palo lto, 1980 and 1982, respectivley. In 1987 he received the Ph.D.
Alto, Calfornia, 1985. degree in Electrical Engineering from Virginia Polytechnic
Hermann W. Dommel, "Digital Computer Solution of Institute and State University. He is presently an Assistant
Electromagnetic Transients in Single- and Multiphase Professor in the Department of Electrical Engineering at
Networks," Transactions On Power ADDaratuS Auburn University. His research interests are in power
-
And Svstems, vol. PAS-88, no. 4, pp. 388-399, April electronics, power systems, and electric machinery.
1969.
Hermann W. Dommel and W. Scott Meyer, G . B. Sheble' (M 71, SM 85) is an Associate Professor
"Computation of Electromagnetic Transients," of Electrical Engineering, Auburn University, Auburn,
Proceedings of the IEEE, vol. 62, no. 7, pp. 983-993, Alabama. Dr. Sheble' received his B.S. and M.S. degrees in
July 1974. Electrical Engineering from Purdue University and his Ph.D.
[41 Hermann W. Dommel, "Nonlinear and Time-Varying in Electrical Engineering from Virginia Tech. His industrial
Elements in Digital Simulation of Electromagnetic experience includes over fifteen years with a public utility,
Transients," Transactions on Power ADDaratUS with a research and development firm, with a computer vendor
-
and Svstems, vol. PAS-90, pp. 2561-2567, Nov/Dec. and with a consulting firm. His interests include automatic
1971. generation control, computer-directed arming of relay systems,
J. Duncan Glover and M. Sarma, Power Svstem network analysis, load forecasting, and thermal and hydro
Analysis and Design, PWS, Boston, 1987. scheduling.
C. A. Gross, Power Svstem Analvsis, John Wiley and
Sons, New York, 1986. Steven R. Newton was born in Florence, Alabama in
W. A. Blackwell and L. L. Grigsby, Introductorv 1961. He obtained the B.E.E. degree from Auburn University
Network Theory, PWS, Boston, 1985. in 1987 and is currently completing the requirements for the
Fernando L. Alvarado, "Parallel Solution of Transient M.S. degree at Auburn University. His interests include power
Problems by Trapezoidal Integration," IEEE system analysis and control theory.
Transactions on Power ARDaratus and Svstems, vol.
PAS-98, no. 3, pp. 1080-1090, May/June 1979. L. L. Grigsby (M 57, SM 70) was born in Floydada,
Richard L. Burden, J. Douglas Faires, and Albert C. Texas in 1929. He received the B.S.E. E. and M.S.E.E.
Reynolds, Numerical Analysis, PWS, Boston, 1980. degrees from Texas Tech University in 1957 and 1962,
L. V. Bewley, Traveling Waves on Transmission respectively, and the Ph.D. from Oklahoma State University in
Svstems, Dover, New York, 1963. 1964. He has taught electrical engineering at Texas Tech
University, Oklahoma State University, Virginia Polytechnic
Institute and State University and is currently Georgia Power
Distinguished Professor of Electrical Engineering at Auburn
University. His research interests include modeling,
simulation, optimization, and control of electrical power
systems.