Line Limit Preserving Power System Equivalent

Wonhyeok Jang, Student Member, IEEE, Saurav Mohapatra, Student Member, IEEE,
Thomas J. Overbye, Fellow, IEEE, and Hao Zhu, Member, IEEE
University of Illinois at Urbana Champaign
Urbana, Illinois 61801
[email protected], [email protected], [email protected] and [email protected]

reduce the computational burden. Kron’s reduction is a widely

Abstract—A reduced order equivalent of a power system is a used method for power system network reduction, which is
simplified model of the original system with the ability to preserve based on successive node (or bus) elimination [9]. Although
some key characteristics and to provide adequate fidelity in Kron’s reduction is capable of calculating the line admittance
simulation studies with considerably reduced computational
values for equivalent lines, the information about line limits is
requirements. In this paper, an algorithm to construct an
equivalent system with the ability to retain thermal line limits is lost in the equivalent network (with fewer nodes).
presented. The application of Kron’s reduction and power It is inevitable to consider power injections at buses when
transfer distribution factor is used in the creation of the reduced determine line limits in power systems since bus injections
system. The criteria for this method and its step-by-step from generation or loads might have impact on them. Without
procedure with the implementation to a small power system are considering bus injections, total transfer capability (TTC) is
described. Also, the result of its application to the IEEE 118 bus
direction independent. With bus injections, however, TTC is
system is provided.
reduced to available transfer capability (ATC) to meet existing
Index Terms—Line limit, Kron’s reduction, power transfer transmission commitments and becomes direction dependent.
distribution factor (PTDF), power system equivalent This can be illustrated in Fig. 1 in which the bus 2 is to be
equivalenced. When bus 2 has no bus injection and TTC can be
easily obtained by choosing the lower value of two
I. INTRODUCTION transmission line limits for both directions between buses 1 and
3. This lower line limit would be assigned as the equivalent line
A N equivalent power system is a model system with
reduced number of nodes and branches than the
corresponding full system. The purpose of making an
limit when bus 2 is equivalenced. In this case, ATC is identical
to the TTC. When bus 2 has the bus injection of 50 MW load,
equivalent system is basically to have higher efficiency in the ATC becomes 90-50=40 MW in the right to left direction
computer simulation without excessively sacrificing the fidelity and 100-50=50 MW in the opposite direction.
of the simulation results of the original system. Traditional
methods such as Ward equivalent have been used to build 1
Z=j0.1
2
Z=j0.2
3
-90 MW 90 MW
power system equivalents [1]-[6]. The Ward equivalent fully Limit=100 MVA Limit=90 MVA

retains the internal system which is in proximity, but slack 90%

A

100%
A

MVA MVA
equivalences the external system which is in distant. Thus, this
does not work well in terms of retaining desired characteristics 0 MW
0 Mvar
across the entire system after significant reduction of elements
from the original system. Which characteristics of the original
(a) Without load at bus 2
system are preserved in the reduced model depends on the kind
1 2 3
of analysis being performed. This paper focuses on creation of Z=j0.1 Z=j0.2
100 MW Limit=100 MVA -50 MW
Limit=90 MVA
an equivalent system that preserve the thermal line limits of the
A A

original system. To have more accurate attributes of the slack 100% 56%
MVA
MVA

original system, the equivalent system has internal system that

50 MW
is spread throughout the system [7], [8]. Therefore, the limits of
0 Mvar
the reduced portion of the system become important. This is
contrast with the Ward equivalent in which the equivalent is
(b) With load at bus 2
completely external area to the retained system. Fig. 1. Example of the impact of bus injection
Line limits are a fundamental part of the formulation for
Power Flow problems, Optimal Power Flow (OPF) problems Even though bus injections play a role in finding line limits
and Security-Constrained Optimal Power Flow (SCOPF) in the reduced system, it is not easy to solve the complicated
problems for electric power systems. While solving such problem. It is known that for large systems it would not be
problems for large-scale networks, it is common practice to use possible to assign limits to equivalent lines that exactly match
equivalent networks derived from detailed models so as to those of the original system for all operating points. Therefore,

978-1-4673-5602-2/13/$31.00 ©2013 IEEE 206

this paper focuses on calculating equivalent line limits without
considering bus injections by unloading systems as a precursor
of the comprehensive study in the future.
The Limit Preserving Equivalent (LPE) algorithm, presented yij yij
yik
in this paper, is based on Power Transfer Distribution Factor ~
yij
(PTDF) and capable of translating limits of physical lines in a
yjk
full-scale model to limits of equivalent lines in reduced-node
model. This will make equivalent networks an attractive option
Fig. 2. Network reduction by eliminating bus k
for carrying out studies about the full-scale network, such as
system capacity assessments, total transfer capability However, this equivalent line has no associated line limit,
calculations, operational reliability (security) analyses [10], which is exactly the challenge of our proposed line limit
economic analyses [11], and load-pocket identification [12], persevering equivalent approach. Hence, the proposed LPE
[13]. algorithm intends to provide the line limit for any pair of
This paper is organized as follows. Section II discusses neighboring buses of the eliminated bus.
features of network reduction and criteria of the proposed
algorithm. Special cases of obtaining line limits are introduced B. Criteria for Line Limit Preservation
in Section III. Section IV explains the proposed LPE algorithm To tackle this, it is desirable to match the total transfer
in detail with a small system example. In Section V, application capability (TTC) between the first neighbor buses in the
of the algorithm to a network example is presented. Finally, reduced system and those in the full system. Here, it should be
computational aspects of the algorithm and conclusion are noted that only the eliminated lines are considered in
drawn in Section VI and VII, respectively. calculating TTC of the original system. The TTC is defined as
the amount of power that can be transferred over the
II. PROBLEM STATEMENT transmission network in a reliable manner while meeting all of
A. Network Reduction a specific set of defined pre- and post-contingency system
conditions [15]. In this paper, however, since only bus-to-bus
Consider a power system with N buses indexed by
transfers are concerned, contingencies are not considered in the

1, … , . Let ℰ denote the set of all lines and Y its  ×  bus
calculation of the TTC. As will be detailed soon, the TTC
admittance matrix. To eliminate a given subset of buses, Kron’s
attributes are closely related to the line limits of a power
reduction has been adopted in this paper to build the equivalent
system, which speaks for their importance as the criterion of the
bus admittance matrix for the reduced power system. Although
LPE algorithm. Before that, notice that the TTC is independent
a closed-form solution is available through matrix computation,
of direction as bus injection is not considered in the paper. This
for efficient computations, practical implementations of partial may lead to different TTC values depending on the direction of
factorization of Y proceed by sequentially eliminating one bus
power flows. However, such differences are assumed
at a time. Specifically, to eliminate a given bus k, the
negligible in this paper.
admittance between any two buses i and j can be updated as To this end, suppose a bus k is to be eliminated, which
converts the original N-bus system to the reduced (N-1)-bus
Yik Ykj ~ one. As mentioned earlier, it is of interest to assign the
Yij ' = Yij − = Yij + Yij (1) equivalent line limits corresponding to k’s neighbor buses, as
Ykk
denoted by  =
 | ∈ 1,  of cardinality . By eliminating
bus  , all the  lines in  =
 | ∈ 1,  will also be
For the original system, the bus admittance
 takes value 0
eliminated, notice that this assumes possible parallel lines have
if there is no direct line between i and j. Otherwise, it is related been combined to a single line. Similarly, let =
to the corresponding line admittance quantity and is prescribed
ℎ" |# ∈ 1, $ of cardinality $ denote the set of lines in the
with a given line flow limit denoted by  . The essential goal original system directly connecting any two buses in  with
of the present paper is to update the line limit  for the admittance
 in (1). Also, let  % =
& |' ∈ 1, ( the set of
reduced system. equivalent lines added after the elimination. Clearly,  % has
Even though the update in (1) holds generally for any two ( = )*, number of lines whose limits are to be determined.
chosen buses i and j, the bus admittance only changes if and +

only if buses i and j are both neighbors of bus k, i.e., (, ) ∈ ℰ Finally, the set of power transaction bus-pairs is denoted as
and (, ) ∈ ℰ . Clearly, if either
 or
 is 0, then the - = ./0 = . ,  12,  ∈ 1, ,  ≠ 1 of cardinality 4 = )*+,,
fractional term
 in (1) becomes also 0, in which case the while each /0 is independent of direction.
corresponding line limit stays unchanged as  =  . To calculate the TTC, PTDF is introduced to linearly
approximate the impact of power flowing on any line with
Otherwise, if
 has indeed changed as given by (1), it is
respect to the power transfer of a transaction /0 . Specifically,
equivalent to adding another parallel line to connect buses i and
the PTDF of any line  ∈  for a transaction /0 ∈ - is
j, as illustrated in Fig. 2.
)89 ,
denoted by φ6 ∈ (0,1 . The power transfer for the
7

207
transaction /0 that obeys the limit of line  is upper-bounded Source
Source
)8 ,
by 67 /φ6 9 . Hence, the TTC of transaction /0 before the 2 80 MVA 2
7 90 MVA
0.08 pu 0.10 pu 90 MVA 80 MVA
0.10 pu
0.08 pu
reduction is determined by the minimum of those upper 68% 6% 6%
PTDF 68%
bounds, as follows: 3 PTDF
3 PTDF
PTDF

23% 2%
PTDF PTDF
Sink 100 MVA
26% 0.06 pu Sink 0.23 pu 0.27 pu

 F  PTDF

(w p )   9%
P = min  ( wl p ) 
i
(2) 4
PTDF
4
 ϕ l
l ∈L 32% 6% 0.54 pu
i

i
 PTDF
70 MVA
PTDF
60 MVA
0.12 pu 0.14 pu
1

)8 , Fig. 3. 4-bus system being equivalenced to 3-bus system

Similarly, once the post-reduction PTDF value φ < 6= 9 is
available, the corresponding TTC in the reduced system can be TABLE I
calculated as: TTCS BETWEEN FIRST NEIGHBOR BUSES IN THE ORIGINAL 4-BUS SYSTEM
/0 Binding line > (89) (MW)
 F ~  2-3 1-3 217.0
~ (wp ) l
P = min~  ( wτ p )  (3) 2-4 1-4 171.7
lτ ∈ L ~
 ϕ lτ  3-4 1-4 144.9

where 6= is the limit for the equivalent lines which is the
In each step of equivalencing, generation and load would be
moved to the retained buses. However, relocating eliminated
objective values to calculate in the paper. Even though line
generation/load to neighbor buses in the process of
limits use MVA for its unit, MW is used for power transfers as
equivalencing is not covered as the system is unloaded in this
well as TTCs since reactive power is operating point oriented,
paper.
hence it is not very significant to consider. To preserve the line
limits, the proposed LPE algorithm intends to match the TTC
III. SPECIAL CASES
quantities as given by (2) and (3) for any transaction /0 .
Interestingly, we only need to account for the eliminated lines This section will focus on how to combine series and parallel
in  and the newly added equivalent lines in  % since the lines in lines. These two cases not only give a simple illustration of the
general algorithm, but also become useful for preprocessing of
keep their line parameters after equivalencing.
the algorithm.
There are two common approaches to calculate PTDFs [15],
[16]; the linearized ac method that includes loss in the A. Series Calculation
calculation, and the lossless dc method that does not. In this Consider a set of m lines in series * =
? , +, … , @  and
paper, the lossless dc method is used to calculate PTDFs each line has its own admittance and line limit as shown in Fig.
because of its simplicity and computational advantages. One of 4. The series lines in * are being equivalenced to one single
the characteristics of PTDF to note is that PTDFs on the line with the admittance A6 and the new line limit 6 .
retained lines in the reduced system are not affected by
equivalencing. This is because PTDFs are based on the line
parameters, and the ratio of parameters of a retained line to the yl1 yl2 ylm
a b
rest of the system does not change by equivalencing according Fl1 Fl2 Flm
to the Kron’s reduction.
Fig. 3 shows a 4-bus system on the left and bus 1 is being
eliminated resulting in the 3-bus system on the right. Lines have y'l
a b
limits in MVA and reactances in p.u. The PTDFs are in % with ~
Fl
bus 2 as a source and bus 3 a sink for both systems in Fig. 3. By
Fig. 4. Series line equivalencing
removing bus 1, three equivalent lines are added between the
first neighbor buses, but their line limits are unknown. As
The limit of the equivalent line for multiple series lines must
mentioned above, the PTDFs on the retained lines from the
be the minimum of the series line limits. This is because the
original system remains unchanged in the equivalent system.
equivalent line limit should represent the maximum power that
However, the limits for the newly created equivalent lines are
can flow on the original series lines without violating the limits.
yet to be determined.
Therefore, the new equivalent line limit for series lines can be
TTCs on the lines being eliminated for the transaction
obtained as follows:
between the first neighbor buses and bus 1 in the original
system can be calculated with (2) and the results are shown in ~
Table I. The TTCs between the same buses in the reduced Fl = min {F l i } (4)
li ∈ LS
system should match the values in Table I and they will be
compared later in Section IV.

208
 B. Parallel Calculation first neighbor buses can be greatly reduced by applying Tinney
Consider a set of m lines in parallel B =
? , +, … , @  and Scheme 2 [17]. This scheme calculates the number of first
each line has its own admittance and line limit as shown in Fig. neighbor buses for all the candidates at each iteration and
5. The parallel lines in B are being equivalenced to a single choose the next bus with the fewest first neighbors to be
line with the admittance A6 and the new line limit 6 . eliminated.
To explain the algorithm in detail, consider the N-bus system
from Section II with the same notation. In the process of
equivalencing bus k, assume f lines are eliminated from the
N-bus system and e equivalent lines are introduced in the
yl1 yl2 ylm y'l reduced (N-1)-bus system. The objective is to find the limits for
~ the e equivalent lines so that TTCs for every pair of the first
Fl1 Fl2 Flm Fl neighbor buses in the equivalent system equal those in the
original system, but only considering the eliminated f lines
Fig. 5. Parallel line equivalencing
when calculating the TTCs of the original system.
First of all, combine parallel lines between buses in C to a
The new limit of the combined parallel lines is calculated by single line with a combined line limit using (5). After that,
determining which line in the parallel bundle is binding. calculate PTDFs between the first neighbor buses in  on the
lines in . With the PTDFs and the given line limits of the
original system, TTCs for each transaction in the original
~  y' 
Fl = min  Fl i × l  (5) system can be obtained with (2). The TTC for each transaction
li ∈L P
 y l i  /0 in the original system should be less than or equal to the
power transfers on equivalent lines for the same transaction /0
The algorithm may create a lot of parallel equivalent lines as in the reduced system as illustrated in Table II. Note that there
a consequence. The combination of parallel lines and the are e transactions since ( = 4 = )*+, = D, hence Table II is
calculation of its parameter and limit is a crucial building block always square. By multiplying each PTDF term for both sides
in the series of node eliminations. During any stage of of the equation in each entry, inequality constraints for the
elimination, this will allow one to consider parallel limits of the equivalent lines can be obtained as a product of the
configurations of lines between nodes as one single line. TTC for a transaction from the original system and the PTDF
Therefore, for illustrating the algorithm for a general step of for the same transaction from the reduced system as shown in
single node elimination, it is sufficient to consider a single line Table III. Those entries give the minimum limit needed to allow
between any two nodes. This can be thought of as a for the original TTCs. All the entries in each row have to be
pre-processing procedure, and results in a neater algorithm. smaller than or equal to the corresponding equivalent line limit.
One entry for each column has to have an equality constraint
IV. LINE LIMIT PRESERVING EQUIVALENT ALGORITHM so that the TTC of that transaction can be determined. Also, one
The goal to achieve in preserving line limits is to match the entry for each row has to have an equality constraint so that the
TTCs between any of first neighbor buses in the reduced limit of that equivalent line can be determined. Often, the
system match the TTCs between the same buses in the original solutions are trivial such that just choosing the largest value in
system. In the process, buses are eliminated one at a time for the each row could be the exact solutions. If each row and each
computational advantage. The overview of the algorithm is column has one solution, then those set of solutions are the
described below. As each bus is being equivalenced, exact equivalent line limits since they satisfy all the inequality
and equality constraints.
1) Combine limits of parallel lines to be eliminated
TABLE II
2) Calculate PTDFs between the first neighbor buses of INEQUALITY CONSTRAINTS FOR POWER TRANSFER CAPACITIES
the bus that is being eliminated
3) Calculate TTCs between the first neighbor buses, only 6E 6E 6E
(8 )
≥ >(8E) (8 )
≥ >(8G ) … (8 )
≥ >(8H)
considering the lines that are being eliminated φ
< 6E E φ
< 6E G φ
< 6E H
4) Determine limits for equivalent lines so the TTCs in the
reduced system match those in the original system 6G 6G 6G
5) In case of non-exact limits for equivalent lines, (8 )
≥ >(8E) (8 )
≥ >(8G ) … (8 )
≥ >(8H)
φ
< 6G E φ
< 6G G φ
< 6G H
determine lower and upper limits
⋮ ⋮ ⋱ ⋮
Selection of a set of buses to be eliminated could vary
according to the application of the algorithm. However, the
6H 6H 6H
order of bus elimination might have a great impact on the
(8 )
≥ >(8E) (8 )
≥ >(8G ) … (8 )
≥ >(8H)
results of the algorithm since the computational expense φ
< 6H E φ
< 6H G φ
< 6H H
depends on the number of first neighbor buses. The number of

209
 TABLE III
INEQUALITY CONSTRAINTS FOR EQUIVALENT LINE LIMITS
/? /+ ⋯ /L B. Example of Non-Exact Solution Case
Exact solutions may not exist in some cases since it would
(8 ) (8 ) (8 )
6E ≥ >(8E ) × φ
< 6E E >(8G) × φ
< 6E G ⋯ > (8H ) × φ
< 6E H not be always possible to assign limits to equivalent lines that
fit all operating points. In this case, solutions can be bound with
(8 ) (8 ) (8 )
6G ≥ >(8E ) × φ
< 6G E >(8G) × φ
< 6G G ⋯ > (8H ) × φ
< 6G H lower and upper limits. Consider that the limit on line 1-4 in the
original 4-bus system is reduced to 20 MVA in order to have
⋮ ⋮ ⋮ ⋱ ⋮ non-exact equivalent line limits. The reduced line limit results
in reduced TTCs as shown in Table VI and inequality
(8 ) (8 ) (8 )
6H ≥ >(8E ) × φ
< 6H E >(8G) × φ
< 6H G ⋯ > (8H ) × φ
< 6H H
constraints for equivalent lines are calculated accordingly in
Table VII.

TABLE VI
A. Example of Exact Solution Case TTCS OF 4-BUS SYSTEM WITH REDUCED LINE LIMIT
Consider the 4-bus system in Fig. 3 for example. The /0 Binding line > (89) (MW)
inequality constraints for equivalent line limits in the reduced
2-3 1-3 217.0
3–bus system are shown in Table IV. In this case, selecting the
largest value in each row satisfies all of the inequality 2-4 1-4 57.2
constraints for each line limit and also enforces all the 3-4 1-4 48.3
transactions, hence producing the exact solutions. Fig. 6 shows
the reduced 3-bus system with the new limits for the equivalent TABLE VII
INEQUALITY CONSTRAINTS FOR EQUIVALENT LINE LIMITS FOR NON-EXACT
lines and the TTC of 217 MW for the transaction between bus 2 SOLUTION CASE
and 3. With the new limits, the TTCs between the first neighbor /+M /+N /MN
buses in the equivalent system can be calculated using (3) and
6GO ≥ 50.8 MW 1.6 MW 9.9 MW
they are identical with the original system as shown in Table V.
6GP ≥ 5.2 MW 13.8 MW 10.5 MW
TABLE IV 6OP ≥ 19.1 MW 6.2 MW 9.5 MW
INEQUALITY CONSTRAINTS FOR EQUIVALENT LINE LIMITS IN THE 4-BUS
SYSTEM EXAMPLE
/+M /+N /MN In this case, taking the largest value in each row does not
6GO ≥ 50.8 MW 4.8 MW 29.9 MW satisfy equality constraints for /MN . The allowable power flow
in transaction /MN is overestimated since none of the entries in
6GP ≥ 5.2 MW 41.4 MW 31.5 MW
its column is enforced. This sets upper bound for the equivalent
6OP ≥ 19.2 MW 18.6 MW 28.5 MW line limits.
Furthermore, it is possible to have lower bound by insuring
217 MW
that the flow in every transaction is equal or less than the TTCs
slack
26 Mvar in the original system. However, some of the inequality
2
constraints would be in violation, these limits underestimate the
90 MVA 80 MVA
TTC in some transactions. In this paper, lower limits are
0.08 pu 0.10 pu
motivated by defining a “limit violation cost” (LVC) for each
A
3
A
entry in the matrix, which is the sum of violations for all entries
100% 13%
MW MW in the row as follows:
51 MVA 41 MVA
0.23 pu 0.27 pu

∑W maxW {0,(P )}

217 MW A

× ϕ~li × ϕ~li
68% (w p ) (wq ) (wq ) (w p ) (w p )
0 Mvar
MW
Vli = −P (6)
29 MVA 4 wp∈
wq ∈
0.53 pu
Fig. 6. Reduced 3-bus system with equivalent line limits for the transaction
between bus 2 and 3 The LVC matrix for the inequality constraints can be
obtained using (6) as shown in Table VIII. To illustrate this in
TABLE V
COMPARISON OF TTCS BETWEEN THE ORIGINAL AND THE EQUIVALENT
detail, for the first row of the table, the entry for /+M is 0
SYSTEM FOR EXACT SOLUTION CASE because it involves no limit violations as it is the largest in the
Original 4-bus system Equivalent 3-bus system row. However, the entry for /+N is related with the other two
/0 (89 ) transactions in the row, hence its LVC is calculated as
Binding line > (MW) Binding line >(89) (MW)
(50.7 − 1.6) + (9.9 − 1.6) = 57.4 and the third entry
2-3 1-3 217.0 2-3 217.0 becomes 50.7 − 9.9 = 40.8. Therefore, each entry is the sum
2-4 1-4 171.7 2-4 171.9 of the differences between itself and larger entries in the row.
3-4 1-4 144.9 2-4 144.7

210
 TABLE VIII
LIMIT VIOLATION COSTS
Exact solution Number of fill-ins
/+M /+N /MN 1.2
High/low solution
12

Normalized average TTC

6GO 1.1 10

Number of fill-ins
0.0 57.4 40.8
1 8
6GP 13.9 0.0 3.3 0.9 6
6OP 0.0 16.2 9.6 0.8 4
0.7 2
0.6 0
Now, this is a resource allocation problem that one entry
0.5 -2
from each row and each column has to be chosen so that it 39 87 111 13 22 36 52 71 83 102 114 19
minimizes the sum of the limit violation costs. The Hungarian Bus numbers in the order of elimination
method (also known as Munkres assignment algorithm) is one Fig. 7. Simulation results from IEEE 118-bus system
of the algorithms that solve this kind of minimum matching
problem [18]. According to the Hungarian method, if 0.0 were The left axis shows the normalized average TTC which is the
chosen for both the first and second row, and 9.6 instead of 0.0 post-elimination average TTC on the eliminated lines divided
in the third row, it would provide minimum sum of limit by the pre-elimination average TTC on the equivalent lines.
violation costs in Table VIII. For the lower bound solution, Therefore, 54 steps with exact line limits produce 1 and the rest
therefore, the new limits would be 50.7 MW for line 2-3, 13.8 two steps with non-exact line limits show their upper and lower
MW for line 2-4 and 9.5 MW for line 3-4. boundaries in the figure. The number of fill-ins introduced
Table IX compares the TTCs on the eliminated lines in the during the equivalencing process gradually increase according
original system and those on the equivalent lines in the to Tinney Scheme 2. Generally, lines with high impedance after
equivalent system. When the largest value in each row is a series of equivalencing steps can be ignored, but in this case,
chosen for the equivalent lines as the upper limits, the TTC in just one equivalent line has more than 3 p.u. impedance with its
transaction /MN is overestimated about 132%. On the contrary, exact limit of 11.03 MW. Simulation results with a random
when the power flow for transaction /MN is enforced by order of bus elimination show that many more lines have upper
applying the Hungarian method to the limit violation cost and lower boundaries and also it takes much more time to run
matrix, the TTC in transaction /+M is underestimated about the simulation as the efficiency of the algorithm significantly
50%. depends on the number of the first neighbor buses.
If a line with two limits is involved in the next equivalencing,
this upper limit and lower limit becomes input for the next VI. COMPUTATIONAL ASPECTS
process of limit calculation. Both of them will generate another Consider an N-bus system, in which M buses are being
upper and lower limit and the bigger value of the upper limit equivalenced in a sequential manner. Let  be the number of
and the smaller value of the lower limit should be chosen for the first neighbor buses for bus i. For each elimination step, )Z+7 , of
equivalent line limits.
equivalent line admittances are calculated with Kron’s
TABLE IX
reduction. Also, the number of calculations of PTDFs in pre-
COMPARISON OF TTCS BETWEEN THE ORIGINAL AND THE EQUIVALENT and post-elimination stages is )Z+7,[)Z+7, +  \. For non-exact
SYSTEM FOR NON-EXACT SOLUTION CASE
solution cases, minimum matching algorithm is also needed.
Original Equivalent
The brute-force approach for assignment problems has to
4-bus system 3-bus system
/0
consider n! combinations where n is the dimension of the
>(89) with >(89 ) with
Binding >(89 ) Binding
upper limit lower limit matrix, which is the number of equivalent lines, )Z+7 , in the
line (MW) line
(MW) (MW) paper. However, the Hungarian algorithm reduces the
M
2-3 1-3 217.0 2-3 216.1 107.5 complexity to O ^)Z+7, _ . Therefore, the complexity of the
2-4 1-4 57.2 2-4 57.3 57.3 proposed algorithm can be rapidly increased with the number of
3-4 1-4 48.3 2-4 63.6 48.2 first neighbor buses. However, Tinney Scheme 2 reduces the
number of fill-ins by ordering the buses during the partial
factorization and consequently the computational burden. In
V. NETWORK EXAMPLE addition, the impedances of certain equivalent lines become
very high after a series of equivalencing steps and hence they
The IEEE 118-bus system is used to apply the proposed LPE
can be ignored in the process to reduce the computation further.
algorithm. A set of buses to be eliminated are selected based on
the amount of power flow going through the bus less than 60
VII. CONCLUSION
MW. Thus, total 56 buses are chosen to be eliminated and the
resultant system has 62 buses with 42 retained lines. As In this paper, the procedure to calculate equivalent line limits
aforementioned, Tinney Scheme 2 is applied to reduce the when building an equivalent power system is presented. The
number of neighbor buses as it reduces the number of fill-ins. proposed algorithm assumes that the system is unloaded so the
Fig. 7 shows the results of the simulation. bus injections from generation or loads do not affect on line

211
limits. For computational advantage, The Kron’s reduction is [11] T. J. Overbye, D. R. Hale, T. Leckey, and J. D. Weber, “Assessment
of Transmission Constraint Costs: Northeast U.S. Case Study,”
applied with partial factorization in order to obtain the line Proceeding of IEEE Winter Meeting, Singapore, January 2000.
parameters of the equivalent system as buses are sequentially [12] K. M. Rogers, and T. J. Overbye, "Clustering of Power System Data
equivalenced. The next bus to be eliminated is chosen based on and Its Use in Load Pocket Identification," 2011 44th Hawaii
the bus valence to reduce the computation by reordering buses International Conference on System Sciences (HICSS), pp.1-10, 4-7
Jan. 2011.
during the partial factorization. TTCs are calculated using line [13] B. C. Lesieutre, K. M. Rogers, T. J. Overbye, and A. R. Borden, "A
limits and PTDFs of pre- and post-elimination stages. The Sensitivity Approach to Detection of Local Market Power
preserved attribute by assigning limits to the equivalent lines Potential," IEEE Transactions on Power Systems, vol. 26, no. 4, pp.
1980-1988, Nov. 2011.
was that TTCs between the first neighbor buses of in the [14] Available Transfer Capability Definitions and Determination,
original system match with those for the same buses in the North American Electric Reliability Council, Jun. 1996.
equivalent system. The algorithm can determine exact limits of [15] J. Wood and B. F. Wollenberg, Power Generation, Operation, and
Control. New York: Wiley, 1996, pp. 108–111.
equivalent lines in the reduced system when it is possible. In [16] C. Duthaler, M. Emery, G. Andersson, and M. Kurzidem, “Analysis
case where exact limits do not exit, the solutions could be of the Use of PTDF in the UCTE Transmission Grid,” 16th Power
bounded with upper and lower limits. There are a couple of Systems Computation Conference (PSCC), Glasgow, 14th−18th Jul.
2008.
areas that this algorithm has to deal with in the future research. [17] W. F. Tinney and J. W. Walker, "Direct solutions of sparse network
The impact of power injections on line limits needs further equations by optimally ordered triangular factorization,"
investigation. In addition, how to handle the range of line limits Proceedings of the IEEE, vol. 55, no. 11, pp. 1801-1809, 1967.
[18] H.W. Kuhn, “The Hungarian Method for the Assignment Problem,”
in case of non-exact solution cases will be important issue as Naval Research Logistics Quarterly, vol. 2, no. 1-2, pp. 83-97, Mar.
those ranges of limits have to be single values for the reduced 1955.
system to be used in any applications.

ACKNOWLEDGMENT
This work was supported by the U.S. Department of Energy
through the Consortium for Electric Reliability Technology
Solutions (CERTS) and the Illinois Department of Commerce
& Economic Opportunity (DCEO) through the Illinois Center
for a Smarter Electric Grid (ICSEG).

REFERENCES
[1] J. B. Ward, “Equivalent circuits for power-flow studies,” AEEE
Transactions on Power Apparatus and Systems, vol. 68, pp. 373–
382, 1949.
[2] W. F. Tinney and J. M. Bright, “Adaptive reductions for power flow
equivalents,” IEEE Transactions on Power Systems, vol. PWRS-2,
no. 2, pp. 351–360, May 1987.
[3] S. Deckmann, A. C. Pizzolante, A. J. Monticelli, B. Stott, O. Alsac,
“Studies on Power System Load Flow Equivalencing,” IEEE
Transactions on Power Apparatus and Systems, vol. PAS-99, no. 6,
pp. 2301-2310, Nov./Dec. 1980.
[4] S. Deckmann, A. C. Pizzolante, A. J. Monticelli, B. Stott, O. Alsac,
“Numerical Testing of Power System Load Flow Equivalents,”
IEEE Transactions on Power Apparatus and Systems, vol. PAS-99,
no. 6, pp. 2292-2300, Nov./Dec. 1980.
[5] T. Baldwin, L. Mili, and A. G. Phadke, “Dynamic ward equivalents
for transient stability analysis,” IEEE Transactions on Power
Systems, vol. 9, pp. 59–67, 1994.
[6] F. C. Aschmoneit, J. F. Verstege, “An external system equivalent
for on-line steady-state generator outage simulator,” IEEE
Transactions on Power Apparatus and Systems, vol. PAS-98, no. 3,
pp. 770–779, 1979.
[7] D. Shi, D.L. Shawhan, N.Li, D.J. Tylavsky, J.T. Taber, R.D.
Zimmerman and W.D. Schulze, “Optimal Generation Investment
Planning: Pt. 1: Network Equivalents,” in Proc. 2012 North
American Power Symposium (NAPS), Champaign, IL.
[8] D. Shi, D.L. Shawhan, N.Li, D.J. Tylavsky, J.T. Taber, R.D.
Zimmerman and W.D. Schulze, “Optimal generation investment
planning: Pt. 2: Application to the ERCOT system,” in Proc. 2012
North American Power Symposium (NAPS), Champaign, IL.
[9] G. Kron, Tensor Analysis of Networks. Wiley, 1939.
[10] T. J Bertam, K. D. Demaree, and L. C. Dangelmaier, "An Integrated
Package for Real-Time Security Enhancement", IEEE Transactions
on Power Systems, vol. PWRS-5, no. 11, pp. 592-600, May 1990.

212