IEEE Transactions on Power Delivery, Vol. 11, No.
1, January 1996 165
AN ON LINE RELAY COORDINATION ALGORITHM FOR ADAPTIVE
PROTECTION USING LINEAR PROGRAMMING TECHNIQUE
Bijoy Chattopahyay M.S. Sachdev T.S. Sidhu
Member Fellow Senior Member
Mehta Tech, Inc. University of Saskatchewan
208N, 12 TH AVE Saskatoon, SK
Eldridge, IA 52748 Canada S7N OW0
Abstract: An adaptive system for protecting a distribution systematic approach for determining the relative sequence
network should determine and implement relay settings that setting of the relays in a multiloop network. They used a
are most appropriate for the prevailing state of the power linear graph theory approach which provided a directional
system. This paper presents a technique for determining loop matrix. A minimal set of break points spanning all
cbordinated relay settings. The technique uses the Simplex loops of the system graph were obtained from this matrix.
twophase method; Phase I determines whether the Damborg et al. [3] extended the graph theoretic concepts
constraints selected for illustrating the conditionality and proposed a systematic algorithm for determining a
between primary and back up relays are feasible, and relative sequence matrix corresponding to a set of sequential
Phase II finds the optimal relay settings. A looped pairs which reduced the number of iterations. Jenkins et al.
distribution system, protected by directional overcurrent [4] proposed a functional dependency concept for topological
relays, war; used for testing the technique. The tests were analysis of the protection scheme. They expressed the
conducted in a laboratory environment: some results from constraints on the relay settings through a set of functional
those tests are reported in the paper. dependencies. Relay coordination was carried out through
the identification of a break point set (BPS) and a relative
sequence matrix. The choice of the initial settings of the
I. INTRODUCTION BPS relays was used to select the settings of the remaining
relays. A parametric optimization approach was reported
The key issue in selecting the relay settings is to achieve the by Urdeanneta et al. [5] that optimized the time multiplier
minimum possible operating times while maintaining settings (TMS) using the Simplex method. Optimal values
coordination among all relays. Usually, finding the of the pickup currents for selected TMS were then
coordinated settings takes several iterations before a determined by using a generalized reduced gradient
satisfactory solution is achieved. Traditionally, a trial and technique.
error procedure is employed for setting relays in multiloop
networks. In the past few years, several mathematical Both the graph theoretic and functional dependency
techniques have been reported. Knable [l] proposed a approaches provide a solution which is the best of the
technique to break all the loops at the, so called, break alternative settings considered, but not necessarily an
points and locate the relays from which to start the optimal solution. In this paper, a relay coordination
coordination procedure. Since looped circuits are normally algorithm is described. It uses the Simplex twophase
protected by directional overcurrent relays located at both method; Phase I detects whether all the selected operating
ends, the loops formed in the clockwise and anti conditions between the primary and backup relays are valid,
clockwise directions are considered for determining the and Phase I1 finds the optimum relay settings. The
break points. Dwarakanath and Nowitz [2] suggested a operating conditions'that are detected in Phase I to be "not
valid are excluded at the beginning of the Phase 11.
95 WM 0356 PWRD A paper recommended and approved The optimization technique was implemented in the Power
by the IEEE Power System Relaying Committee of the
IEEE Power Engineering Society for presentation at System Research laboratory at the University of
the 1995 IEEE/PES Winter Meeting, January 29, to Saskatchewan [7]. Each relay was implemented on a
February 2, 1995, New York, Np. Manuscript submitted TMS320C25 DSP board placed in a host personal computer.
December 30, 1993; made available for printing One personal computer performed the role of a substation
November 30, 3994
computer and another acted as the central control computer.
08858977/96/!$05.000 1995 IEEE
166
All PCs were interconnected by communication links. properly coordinated. One possible approach to achieving
Relaying DSPs were interconnected by a local area network minimum shock to the system due to faults would be to
and the PC acting as a central computer was connected to minimize a weighted sum of the operating times of all
the PC acting as a substation computer by an RS 232 link. primary relays hoping that the operating times of individual
The software was tested in the laboratory to prove the primary relays would be close to the minimum individual
viability of the concepts. The objective of this paper is to operating times that might be possible. Since the lines are
present some of the optimization concepts and their use in short and approximately of equal length, equal weights (=1)
the project. were assigned for operating times of all the relays. For a
network consisting of m relays, the operating times of the
11. THE ALGORITHM primary relays for nearend faults can be expressed as:
The operating times of overcurrent relays protecting a
looped distribution network as a function of the location of m
fault is shown in Fig. 1. The fault currents are generally z=& 9
maximum when the system has maximum generating i=l
sources connected to it. The inverse time overcurrent relays
are set for this condition. When the generation is less, the where: ti,i is the operating time of the primary relay at
fault current levels are lower and the relays take longer to i for nearend fault at i.
clear the faults. To reduce the shock to the system due to the
fault, it would be desirable to change the settings of relays so The operating times of the backup relays must be more than
that the relays take minimum possible operating time the sum of the operating times of the primary relays and the
without jeopardizing their coordination for all possible states coordination margin. This can be expressed as
of the system. This can be achieved by using adaptive
protection that implements optimal setting of the relays tbi,i> ti,i+At, fori =1 tom, (2)
during all operating states of the power system.
where: ti,i is the operating time for the primary relay
The relay coordination algorithm, which is based on a
for a nearend fault,
parameter optimization technique, optimizes an objective
function of operating times of the primary relays subject to t&i is the operating time for the backup relay
keeping the operation of the backup relays coordinated. for the same nearend fault and
Ideally, all primary relays should operate in minimum At is the coordination time interval (CTI).
possible time but this can not be achieved because it would
be impossib1e to keep the protection In the application reported in this paper, overcurrent relays
conformed to the following IEC characteristic [6]:
k xllMS
t= (3)
Inverse time Inpun 1
I characteristic I
t
0
E
I
I
I
I
I
I
where: k
n
is a constant,
is a characteristic index,
Imp* is the multiple of pickup current and
I
.d
r I TMS is the time multiplier setting,
Since the pickup currents of the relays are predetermined
from the system requirements, Equation 3 becomes
t = a x TMS , (4)
Fig. 1. Operating time of overcurrent relays as a function of
the location of fault.
167
procedure for minimizing the object function, z, is similar to
k the procedure used for minimizing the function, w.
where: a= n *
Imp 1
By making this substitution in Equation 1, the objective
function becomes III.CASE STUDIES
The developed algorithm was used to obtain in an online
mode the TMS values for protecting the distribution network
i=l of the City of Saskatoon using the approach described in the
this paper; more details are given in Reference 7. The
In this equation, ai's are known; values of TMSi are details of the network configuration are provided in
determined by minimizing z and satisfying the coordination Appendix A. Each line of the network is equipped with
between the primary and backup relays. This equation is overload, phase fault directional overcurrent and
optimized using the well known Simplex twophase method instantaneous overcurrent relays. Overload relays are
[9] subject to the condition that the operation of the back up connected to the ct's that see currents up to three times their
relays remains properly coordinated. Phase  I determines ratings. When the currents are more than those limits,
whether all the selected constraints are valid. If constraint is phase fault overcurrent ct's become effective [8]. The
not valid, it is excluded at the beginning of Phase  11. instantaneous units reduce the operating time for nearend
faults.
Phase I of the twophase Simplex method finds a feasible
solution and Phase I1 finds the optimal solution. In Phase I, While testing the algorithm for adaptive relaying functions,
the initial extreme point is moved from the origin to the the following four operating conditions were selected:
feasible region, In Phase 11, pivoting is done from the initial
extreme point to an optimal extreme point. Maximum system load and maximum system generation
with line 120 closed (MXLG1).
To begin with, the inequalities described by Equation 2 are
converted to equalities by introducing nonnegative surplus Minimum system load and maximum system generation
variables, .
x,,, xs2, ............ .,Xsm and nonnegative with line 120 closed (MNLG1).
artificial variables x,,, xt2,............, xtm . Since artificial Maximum system load and maximum system generation
variables are all nonnegative, they are all zero only when with line 120 open (MXLG2).
their sum is zero. The artificial variables are eliminated by
minimizing a function Minimum system load and maximum system generation
with line 120 open (MNLG2).
w = xt, + xtz+. ............+x, . (6)
For each operating condition, prefault currents and voltages
are obtained from the system measurements, and some are
During the minimization of w, the original objective
obtained from the state estimator. The adaptive protection
function z is also updated. The process by which w is
software simulates faults at the near and far ends of the
minimized is called Phase I.
lines. The software determines the secondary pickup currents
for overload, phase fault overcurrent and instantaneous
If the value of w is zero at the end of phase I, the values of
overcurrent relays from the criteria mentioned in the next
all artificial variables are zero and a feasible solution to the
section.
original problem has been achieved. At this stage all
artificial variables become nonbasic. If w cannot be reduced
IV.RELAY SETTING CRITERIA
to zero, it is concluded that no feasible solution exists. One
or more artificial variables, which are still in the basis, have
The relay settings for adaptive protection are based on the
positive values. In such a case, the infeasible constraints
selected power system operating conditions. For
corresponding to the remaining positive artificial variables comparison, a nonadaptive approach has also been
are withdrawn from the original problem. Phase I is
considered and the settings are selected based on the worst
repeated to obtain the feasible solution. Phase 11 of the
case scenario to prevent unnecessary trippings.
Simplex method then optimizes the objective function, The
168
A. Overload Pickuu compared to settings provided by the traditional non
adaptive approach..
In the nonadaptive approach, the relays are set at 2.0 times
the maximum load current. When the adaptive approach is
used, the overload settings equal to two times the load being Table 1: Load, fault currents and settings of the relays.
catered could be used. The load will be usually changing
with the passage of time. The objective is that if the
communication between the relays and the central computer
fails, the protection system should continue to operate
properly. To ensure this, the overload settings are not
allowed to become less than two times the present load and
also not less than 1.5 times the maximum load anticipated
over the course of the next twenty four hours. 11
B. Overcurrent Pickup
I 4
The pickup settings of the phase fault overcurrent relays are
determined considering both the load and fault currents.

16
With the decrease in the load current, the pickup settings
are reduced but are kept at more than the maximum line
current experienced over a 24 hr period.

21
C. Instantaneous Overcurrent Settings 
22
The tap settings of the instantaneous relays are selected to be
1.3 times the farend fault currents. Since the fault currents
change with the change in system generation and circuit 31
topology, the settings for instantaneous relays are also
changed.
In the case of the nonadaptive approach, the instantaneous
tap settings are determined considering the maximum far
end fault currents and are kept fixed.
V. RESULTS
After selecting the pickup settings and calculating the fault
currents, the coordination software minimized the operating
time for the near end faults to determine the optimum time
multiplier settings. A coordination time interval of 0.2 sec
was used. Table 1 shows the calculated pickup and time
multiplier settings of overload, overcurrent and
instantaneous overcurrent relays. The table also includes the
load currents, fault currents and ct ratios. Figs. 2 and 3
show the operating characteristics for relays 11 (at Ave C
S/S) and 51 (at PI. Hill S/S) which protect the line between VI. CONCLUSIONS
Aye C and P1. Hill substations. Fig. 4 shows the operating
characteristics for relay 14 (at Ave C S/S) which protects This paper has described an online relay setting and
line 120. Fig. 5 shows similar characteristics for relay 41 coordination technique that can be used to obtain settings of
(at Cowley S/S) which protects the line between Cowley and relays provided in a distribution network. The algorithm,
PI. Hill substatios. All characteristics are shown for adaptive based on a linear programming technique using the Simplex
and nonadaptive approaches using the developed algorithm. twophase approach, optimizes the operating times of the
It is clear from these figures that the technique provides relays. The developed technique has the ability to identify
better solution to the distribution network protection the infeasible constraints (conditions) and to isolate them
169
w 2 ' :I 1'
02  I
t
I
L
I '
03 j 4' f :3' [l] Knable, A.H., 'A Standardised Approach to Relay
i... i i Coordination', IEEE Winter Power Meeting, 1969.
0.0 I
170
[31 Damborg et al., ‘Computer Aided Transmission IEC Standard, Single Input Energising Quantity
Protection System Design, Part I: Algorithm,’ IEEE Measuring Relays With Dependent Specified Time,
Transactions on PAS, Vol. PAS103, 1984, pp. 51 Publication No. 2554, 1976.
59.
Chattopadhyay, B., Sachdev, M.S., Sidhu, T.S.,
[41 Jenkins, L., Khincha, H., Shivakumar, S., Dash, P., ‘Protection of a Distribution Network  An Adaptive
‘An Application of Functional Dependencies to the Approach’ Canadian Journal of Electrical and
Topological Analysis of Protection Schemes’, IEEE Computer Engineering, Vol. 19, No. 3, July 1994,
Transactions on Power Delivery, Vo1.7, No.1, pp. 99108.
January 1992, pp. 7783.
IEEE PES Power System Relaying Committee,
PI Urdaneta, A.J., Nadira, R., Jimerez, L.G.P., ‘Relay Performance Considerations with Low Ratio
‘Optimal Coordination of Directional Overcurrent CTs and High Fault Currents’, IEEE Trans. on
Relays in Interconnected Power Systems,’ IEEE Power Delivery, Vol. 8, No. 3, 1993, pp. 884897.
Transactions on Power Delively, Vol. 3, NO. 3, July
1988, pp. 903911. Strayer, J.K., Linear Programming and Its
Applications, SpringerVerlag Inc., New York,
1989.
AppendixA
The selected distribution system, shown in Fig. A1, is a reduced version of the ‘City of Saskatoon’ distribution network. It
consists of five switching stations. The substations have 72kV/14.4kV step down transformers which are connected to 14.4
kV bus. These busses are interconnected by lines to form the distribution network.
Fig. A. 1. Single line diagram of the distribution network.
171
BIOGRAPHIES Tarlochan S. Sidhu (M'90, SM'94) received the B.E.
(Hons.) degree from the Punjabi University, Patiala, India in
Bijoy Chattopadhyay (M'93) received his B.E. from the
1979 and the M.Sc. and Ph.D. degrees from the University
Regional Engineering College, Durgapur, India in 1978. He
of Saskatchewan, Saskatoon, Canada in 1985 and 1989
obtained his M.Sc and Ph.D degrees from the University of
respectively. He worked for the Regional Computer Center,
Saskatchewan, Saskatoon, Canada in 1986 and 1993
Chandigarh, India from 1979 to 1980 and developed
respectively. He worked as a design and project engineers in
software for computerbased systems. He also worked for
a power consultancy company and as a distribution planning
the Punjab State Electricity Board, India from 1980 to 1983
engineer in a power utility company.
in distribution system operation and thermal generating
Dr. Chattopadhyay is currently manager of software
station design. After obtaining the Ph.D. degree, he joined
engineering at Mehta Tech, Inc. His areas of research
BellNorthern Research Ltd., Ottawa, Canada and worked
interest are power system analysis, application of
on a software development project for about one year. He
microprocessors for power system monitoring, protection
joined, in 1990, the University of Saskatchewan where he is
and control, and substation automation.
presently Associate Professor of Electrical Engineering.
His areas of research interest are power system protection
Mohindar S. Sachdev (M'67, SM'73, F'83) was born in and control and applications of microprocessors and neural
Amritsar, India, in 1928. He received the B.Sc. degree from networks for power system monitoring, protection and
the Banaras Hindu University, India, the M.Sc. degrees from control.
the Punjab University, Chandigarh, India and the University
of Saskatchewan, Saskatoon, Canada and, the Ph.D. and the
D.Sc. degrees from the University of Saskatchewan,
Saskatoon, Canada.
He worked for the Punjab P.W.D. Electricity Branch and the
Punjab State Electricity Board, India from 1950 to 1968 in
System Operation, Design and Planning. In 1968, he joined
the University of Saskatchewan where he is currently
Professor of Electrical Engineering. His areas of interest are
power system analysis and power system protection.
Dr. Sachdev is a Fellow of the Institution of Engineeh
(India) and a Fellow of the Institution of Electrical
Engineers, London (England). He is a Registered
Professional Engineer in the Province of Saskatchewan and
a Chartered Engineer in the UK.
172
Discussion Transient Configurations", Proceedings of the IEEECSEE
(Chinese Society of Electrical Engineers), International
A. J. Urdaneta, L.G. Perez (Universidad Sim6n Conference on Power System Technology (ICPST'94),
Bolivar), H. Restrepo (GERS Ltda. Consultores) : Beijing, China, October, 1821, 1994.
The authors are to be commended for a very interesting
article. Manuscriptreceived February 24, 1995.
The application of the Simplex method for linear
programming to calculate the time dial settings for a
coordinated operation of directional overcurrent relays,
was proposed in ref. [7] of the paper. The authors are
responsible for the first online application of this ALEXANDER P. APOSTOLOV, (Rochester  Integrated Systems
methodology, in a real time adaptive protection scheme. Division), Rochester, NY :
We would appreciate the authors comments on the
following points . The authors have presented a very interesting paper on online
I. The transient configurations that take place when only relay coordination using linear programming technique. The
one of the main relays has operated (first main relay) and algorithm described is based on the assumption that the lines are
the second one (second main relay) has not operated yet, short and approximatelyof equal length, i.e. equal weights (=1) are
must be considered in the mathematical formulation of assigned for the operating times of all relays. However, in many
the linear programming problem, as a key issue for cases it is necessary to coordinate relays protecting long lines with
achieving a feasible relay coordination scheme,[A] in relays protecting short lines or vice versa. How are such system
order to obtain real operation times that match those configurations going to affect the performance of the algorithm
predicted by the algorithm, and a completely selective described ?
protective scheme. This point is specially critical when The instantaneous overcurrent setting is described as based on the
determining the settings of the instantaneous units. farend fault current. However, for mutually coupled transmission
In the online application of the paper, it seems that the lines it is possible to have maximum external fault current for a
transient configurations were not included in the short circuit on the parallel line with a breaker opened by the
formulation. instantaneous protection of the faulted line. How is such sequential
What were the differences between the predicted operation considered by the algorithm ?
operation times and the real operation times of the relays In systems which use breaker failure protection it is necessary to
for the second main relays and its correspondent backups? coordinate the remote backup relays with the breaker failure relay.
11. The second issue that we would like to raise, is related Is this considered in the algorithm ?
to the selection of the weighting factors in the objective
function. Manuscript received March 1, 1995.
Due to the particular characteristics of the mathematical
formulation, it can be shown [ref. [A of the paper] that the
solutioh to the optimization problem is not dependent
upon the choice of the weights, as long as they are
positive real numbers. This particular characteristic of the B. Chattopadhyay (Mehta Tech. Inc.), M.S. Sachdev and
optimization problem is perhaps more clearly explained if T.S. Sidhu, Power System Research Group, University of
the relay operation times are assumed as independent Saskatchewan, Canada. We thank the discpssers for their
objective functions in the optimization formulation, due to interest in the paper. Our response to their questions and
the fact that this independent objectives do not compete comments is as follows:
one with each other, and therefore, the reduction of one
necessarily leads to the reduction of the others. The ______
A.J. Urdaneta et. al.
constraints that relate (or couple) the operation times of
different units, are the coordination constraints. These 1. Directional overcurrent relays remain coordinated during
constraints, are such that always dTijk/ dTlmn > 0 if they transient configurations arising from the tripping of one of
are active (=0 if they are not active), for all i,j,k,l,n,m, the circuit breakers controlling a line. The instantaneous
Also, it is not complicated to show that the minimization relays also remain coordinated for such transients. If a
of the operation times of the main relays only, leads to situation is encountered when the relays loose their
exactly the same setting solution than the minimization of coordination during a transient, the requirements of the
all the operation times, associated to the operation as condition can be incorporated in the process.
main and as backup units. Therefore, the choice of the The purpose of the study reported in this paper was to
value of w=l for all the weights is very appropriate. determine optimized coordinated relay settings for the
Finally, we would like to congratulate the authors for the distribution network of the City of Saskatoon. This was
application of linear programming techniques to the
achieved and the characteristics of some of the relays are
solution of the adaptive coordination problem of
directional overcurrent relays, in an online environment. shown in the paper. The operating times of the relays were
neither predicted nor checked from simulations. This can,
however, be done by applying, to the relays, currents
References calculated by the EMTDC via D/A converters. This
[A] Restrepo,H., Urdaneta,A.J., Mhrquez, S , Fajardo, J., procedure is, however, not essential to prove that the
"Optimal Coordination of Directional Overcurrent Relays: operation of the relays will remain coordinated.
173
2. We thank the discussers for confirming our observation can not be protected by directional overcurrent relays only; it
that the selecting equal weights in the objective function is a becomes necessary to use other relaying schemes, such as
valid approach. distance relays. The sequential operation described by the
discusser was, therefore, not studied.
A. Apostolov
 3. It is possible to include breakerfailure relays while
1. We do not see any problems in applying the proposed determining the optimized coordinated settings of the relays.
algorithm to systems that contain long as well as short lines. This can be accomplished by including appropriate
2. The proposed algorithm was tested using the looped constraints while minimizing the objective function.
distribution network of the City of Saskatoon. Overcurrent
and directional overcurrent relays are used to protect the Manuscript received June 5, 1995.
radial and looped lines of this network. The loop networks
that contain parallel lines with sources at intermediate buses