1

Relieving Overload and Improving Voltage by

the Network Contribution Factor (NCF) Method
H. Song, Student Member, IEEE, and M. Kezunovic, Fellow, IEEE

network switching for relieving overload condition [7,8]. The

Abstract--This paper introduces new Network Contribution power flow sensitivity matrices are used to rank the candidate
Factor (NCF) method for relieving overload and improving lines to be switched off to solve the voltage violation and
voltage by using the network contribution information and base overload problems [7]. The Z-matrix method is used to
load flow conditions. When line overload or low voltage occurs,
compare the changes in the Z elements due to one line being
irrespective if it is caused by disturbance, load increase, or
wheeling, we first find the following network contribution factors: switched off [8]. The Optimal Power Flow (OPF) may be a
Flow Network Contribution Factor (FNCF) and Voltage good method in the regulated environment for dispatching the
Network Contribution Factor (VNCF). Then we choose the most generation, adjusting the branch flow and relieving the
contributing elements and change their parameters, either by overload [9]. But the OPF method is not as suitable as before
line control (Thyristor Controlled Series Capacitor-TCSC, line in the competitive open access market.
switching, etc.) or bus control (shunt capacitor, Static Var
This paper presents a novel and comprehensive method of
Compensator-SVC, etc.). The variances of the line flow and bus
voltage are calculated. The results are verified by power flow using network contribution factors for the purpose of relieving
calculation. This method can quickly find the parameter overload, improving voltage, controlling emergency, etc. It
contributing to the largest variance based on the network gives useful control guidance for system operators and is easy
information and base load flow conditions. It can be used for re- to understand and implement. Based on network parameters
dispatching load flow, solving congestion, relieving overload, and base flow bus voltages, for the line flow change, the Flow
improving voltage, controlling emergency, etc.
Network Contribution Factor (FNCF) for each line parameter
variance (i.e., TCSC insertion or line switching) is estimated.
Index Terms--Network Contribution Factor, Congestion
Management, Overload, Voltage Control, Emergency Control, For the bus voltage change, the Voltage Network Contribution
FACTS Factor (VNCF) for each bus shunt element variance is
obtained. Given certain amount of the line flow decrease or
I. INTRODUCTION bus voltage increase, either FNCF or VNCF is used to get the

O ne of the most challenging problems for a competitive necessary and optimal line or shunt parameter variance.
power market is that congestion may occur frequently[1]. This paper introduces the mathematical formulation of the
Following one or several disturbances, the over stressed network contribution factor method. In Section II, both the
system may have one or several lines overloaded or security line parameter variance and bus (shunt) parameter variance
flow-gate limits violated. If the overload can not be relieved are considered. Numerical test results are presented in Section
quickly and appropriately, fault may occur (i.e., line sagged to III. Conclusion and references are given in Section IV and V
trees) and more elements will be tripped. Also, a backup relay respectively.
may operate due to the overload condition, leading to
cascading outages and large area blackout as the final result II. MATHEMATICAL FORMULATION
[2,3,4]. In the complex and competitive power market, it is From the fast decouple power flow, we know the
difficult for system operators to find effective means for approximate real power equation based on the simple fact that
relieving overload to make the maximum balance between the the line resistance is much smaller than the reactance, ri << xi
economy and security. There are examples of good and bad P
consequences of relieving overload during disturbances [5,6]. = B 'θ (1)
E
In general, compared with generator re-dispatching and
where, P, E, θ are the node real power injection,
load management, transmission network control is the fastest
magnitude and angle of the bus voltage respectively.
and cheapest control for relieving overload and controlling
stability. Some research results were reported on using ( B ' ) ij = −bij (2)
where bij is the series inductance of the line i-j
This work was supported by Pserc project, “Detection, Prevention and
Mitigation of Cascading Events”, and in part by Texas A&M University.
Given an n-bus-l-branch system, A is the node-branch
H. Song and Dr. M. Kezunovic are with the Department of Electrical incidence matrix, Yp is the primitive branch admittance matrix,
Engineering, Texas A&M University, College Station, TX 77843 USA (email: Ybs is the node shunt capacitance matrix,
songjefferson@neo.tamu.edu, kezunov@ee.tamu.edu ).
 2

Y p = diag [ y1 ... yl ] (3) yi' = yi + ∆yi (22)

Ybs = diag [ y s1 ... y sn ] (4) The Flow Network Contribution Factor (FNCF) N f can be
1 i is the sending node of branch j defined as follows,
 for line k, k ≠ i
Aij = − 1 i is the receiving node of branch j (5)
0 else N f ,k = −[ A1k ... Ank ] X 1 Ki (23)

for line k, k = i
Bus admittance matrix can be obtained from: n
Y = AY p AT + Ybs (6) N f ,k = ∑ A ji E jθ j / yi' − [ A1i ... Ani ] X 1 Ki (24)
j =1
Assign B ' as the negative value of the imaginary part of Y
In general, bus impedance matrix (imaginary part) X 1
matrix,
doesn’t change much from the original matrix X due to the
B ' = −imag (Y ) (7)
line parameter variance ∆y i . If we assume X1 ≅ X , the Flow
For the real power flow,
Network Contribution Factor N f will be constant except for
Pline = −imag (Y p ) AT ( Eθ ) (8)
line i.
The approximate node injection can be composed from the
We can easily get the line flow variance,
line flows associated with this node,
∆Pline− k = N f ,k y k ∆yi , k=1,…,l (25)
Pnode ≅ APline = A(−imag (Y p )) AT ( Eθ ) (9)
From (25) we can see that the line flow variances are
A. Line Parameter Variance related to three components: the Flow Network Contribution
Factor N f , this line’s series inductance and admittance
For the parameter variance of line i, ∆y i , assume that the
variance of the line i. For each line parameter change, we can
node injection Pnode and bus voltage magnitude E do not get all other line flow variances easily. Vice versa, we can
change much, and the bus voltage angle θ varies ∆θ . Then calculate the line parameter variance based on the exact line
assign flow change we want. This gives us a good guidance to issue
Y1 = −imag (Y p ) (10) network control to re-dispatch the line flows:
Since the node injection does not change, from (9) we get, Step 1, use the base network X matrix to get N f and
A(Y1 + ∆Y1 ) AT E (θ + ∆θ ) = AY1 AT Eθ (11) ∆Pline− k , so the positive or negative contribution of each line
thus parameter change to the line flow of interest is easily known.
∆θ = −( A(Y1 + ∆Y1 ) AT ) −1 ( A∆Y1 AT )θ (12) Step 2, choose the most contributable line, change its
parameter, get the actual N f and ∆Pline− k by using the real
where ∆Y1 = diag[0 .. ∆yi .. 0] (13)
X 1 and yi' , thus get the desired overload relieving.
From (8), we get the line flow change,
∆Pline = Pline
new
− Pline = ∆Y1 AT ( Eθ ) − Step 3, run power flow program to verify the result.
(14) For the line switching, just simply assign ∆yi = − yi ,
(Y1 + ∆Y1 ) AT ( A(Y1 + ∆Y1 ) AT ) −1 ( A∆Y1 AT )( Eθ )
∆Pline−i = − Pline −i . For other line flow changes, for a small size
since
T system, switching off an in-service line may make a big
 n  variance of X , so we use real X 1 and y i' to get the actual
∆Y1 A T ( Eθ ) = 0 .. ∑ A ji E jθ j .. 0  ∆y i (15)
 j =1  N f and ∆Pline− k . For a large size system, switching off one or
( A∆Y1 AT )( Eθ ) = Ki∆yi (16) several lines may not change X as much, so we still use the
where three steps above.
T The above method can give quick guidance for selection of
 n n n 
Ki= A1i ∑Aji Ejθ j .. Aii ∑Aji Ejθ j .. Ani ∑Aji Ejθ j  (17) the parameter to change and the exact parameter variance. The
 j=1 j=1 j=1  power flow calculation can verify it to get an accurate control.
X 1 = ( A(Y1 + ∆Y1 ) AT ) −1 (18) B. Bus Parameter Variance
−1
X = ( AY1 A ) T
(19) When shunt capacitor bank or SVC is being switched
we can get the following line flow variance equations, on/off at a bus, the capacitance variance at bus i is ∆ybs . For
for line k, k ≠ i , real power flow:
∆Pline − k = −[ A1k ... Ank ] X 1 Ki ( y k ∆y i ) (20) Pline = Y1 AT ( Eθ ) (26)
for line k, k = i , Pnode = B ' ( Eθ ) (27)
n
∆Pline−i = (∑ Aji E jθ j / y − [ A1i ... Ani ]X1Ki)(y ∆yi )
'
i
'
i (21) from (26) and (27) we can get
j =1
Eθ = ( B ' ) −1 Pnode (28)
where
 3

∆Pline = Y1 AT (( Bnew
'
) −1 − ( Bold
'
) −1 ) Pnode (29)
'
Since Y1 , A and Pnode are constant, B does not change TABLE I
much. The line real power flow will change, but not much, for BASE FLOW OF THE WSCC 9-BUS SYSTEM
(100MVA BASE, VALUE: P.U.)
the variance of bus capacitance.
However, based on reactive power equation of the fast
Line Line Flow Magnitude
decoupled method, the reactive power Q and bus voltage E
1-4 0.7164 + 0.2548i 0.7604
will change,
Q 2-7 1.6300 - 0.0126i 1.6300
= B '' E (30)
E 3-9 0.8500 - 0.1291i 0.8597
∆Q
= B '' ∆E (31) 4-5 0.4081 + 0.3106i 0.5128
E
4-6 0.3062 + 0.0889i 0.3189
 0   0 
 ..   ..  5-7 -0.8547 - 0.0193i 0.8549

∆Q     6-9 -0.6014+ 0.0195i 0.6017

∆E = (B '' ) −1 = (B '' ) −1 ∆ybs  E = X 2 ∆ybs  E (32)
E     7-8 0.7614 + 0.0503i 0.7631
 ..   .. 
 0   0  8-9 -0.2414 - 0.1388i 0.2785

For the n-bus network with m PQ buses, B '' is m × m Note: The assumed line flow direction is from the beginning node to the
ending node. Negative values of real power flow of line 5-7, 6-9 and 8-9
admittance matrix, X 2 is the inverse of B '' . indicate that the actual flow directions are opposite to the assumed directions.
If we assume B '' doesn’t vary much, we can get the bus
voltage magnitude variance A. Cases of the Line Parameter Variance
∆E j = ( X 2, ji E j )∆y bs j=1,…,m (33) Case 1 Insertion of TCSC at a line
The Voltage Network Contribution Factor (VNCF) N v can be If we assume that the line flow limit of line 5-7 is 0.8 p.u.,
defined as follows, we can see that it is overloaded at the base flow condition.
N v , j = X 2, ji E j , j=1,…,m (34) Consider insertion of Thyristor Controlled Series Capacitor
(TCSC) at each line (assume we have a TCSC at each line) to
adjust the line flow.

III. NUMERICAL RESULTS TABLE II

FLOW NETWORK CONTRIBUTION FACTOR (VALUE: 1E-3)
Take an example of the WSCC 9-bus system given in Fig.
(50% COMPENSATION OF A TCSC AT EACH LINE)
1.
Line 4-5 4-6 5-7 6-9 7-8 8-9

4-5 flow 0.0928 -0.1519 -1.5 1.2 0.2593 -0.1573

4-6 flow 0.1757 0.1142 1.7 -1.3 -0.2846 0.1701

5-7 flow 0.3145 -0.3026 -1.7 2.4 0.4759 -0.2951

6-9 flow -0.3433 0.3113 3.2 1.5 -0.5234 0.3058

7-8 flow 0.1444 -0.1371 -1.3 1.1 0.1236 0.1249

8-9 flow 0.2045 -0.1912 -1.9 1.4 0.2914 -0.0972

For line 5-7 flow, since its actual direction is opposite to

the assumed direction, positive value of the FNCF is needed
to decrease the flow. From Table II, we can see that the
biggest positive value of FNCF for line 5-7 flow is insertion
Fig. 1. WSCC 9-bus system of TCSC at line 6-9. Therefore, insertion of TCSC at line 6-9
will contribute the most to relieving the overload on line 5-7.
Table III gives the flow changes and errors by inserting
Table I gives the base flow condition of each line in p.u.. TCSC with 50% compensation capacity at line 6-9. Column 2
Table II presents the FNCF (Flow Network Contribution is the line flow change based on the Network Contribution
Factor) results if a TCSC with 50% compensation capacity is Factor from matrix X. Column 4 is the line flow change based
inserted at each of the lines 4-5, 4-6, 5-7, 6-9, 7-8 and 8-9 on the real Network Contribution Factor from matrix X1.
respectively. FNCF gives the guidance for the line flow Column 6 is the power flow result. Columns 3, 5 are errors
adjustment. compared with the power flow result.
 4

TABLE V
TABLE III LINE FLOWS WITH LINES DISCONNECTED (VALUE: P.U.)
LINE FLOW CHANGE FOR THE LINE 6-9 PARAMETER VARIANCE
(CHANGE, FLOW VALUE: P.U., ERROR: %)
Flow Original 8-9 off 5-7 off
4-5 0.4081 0.8856 1.2599
Line Change Error Change Error Power flow
4-6 0.3062 -0.1638 -0.4705
4-5 0.0784 5.21 0.0836 -0.99 0.0828 5-7 -0.8547 -0.3765 0

4-6 -0.0733 19.64 -0.0851 6.69 -0.0912 6-9 -0.6014 -1.0697 -1.4131
7-8 0.7614 1.2305 1.6190
5-7 0.0761 10.08 0.0853 -8.3 0.0846
8-9 -0.2414 0 0.6055
6-9 -0.0884 0.6 -0.0883 -0.113 -0.0884

7-8 0.0786 8.86 0.0838 2.88 0.0863 B. Bus Parameter Variance

8-9 0.0766 10.85 0.0815 5.16 0.0859 We consider the bus parameter variance, caused by
switching on 0.5 p.u. (in base MVA capacity) shunt
From Table III, we can see that the line flow change capacitance at bus 5.
obtained by the Network Contribution Factor method is very
close to the power flow result. By this method, we can quickly TABLE VI
BUS VOLTAGE MAGNITUDE VARIANCE (VALUE: P.U.)
find the necessary parameter change to relieve the overload,
and avoid the many trials of power flow calculation.
Voltage Method 1 Method 2 Method 3

Case 2 Line Switching Bus 4 0.0156 0.0163 0.0176

In reality, TCSC is not installed everywhere in the network. Bus 5 0.0433 0.0453 0.0469
To get an easy control of the line parameter, due to the
Bus 6 0.0121 0.0126 0.0140
economic and technological considerations, line switching
may be more applicable for the system operators. Bus 7 0.0101 0.0106 0.0117
Table IV and Table V are given to show the line flow Bus 8 0.0079 0.0082 0.0093
results of switching the line 8-9 and line 5-7 respectively.
Bus 9 0.0044 0.0046 0.0053
TABLE IV. LINE FLOW CHANGES FOR LINE 8-9, 5-7 DISCONNECTED
(VALUE: P.U.)
Method 1 uses the original matrix B '' to get the voltage
Flow Line 8-9 off Line 5-7 off
variance. Method 2 uses the new matrix B '' after switching on
4-5 0.2339 0.2414 0.2361 0.8009 0.8326 0.8518 the shunt capacitor. Method 3 is the power flow method. We
4-6 -0.2263 -0.2404 -0.2296 -0.7974 -0.8596 -0.7767 can see that their bus voltage variances are very close. The
Voltage Network Contribution Factor (VNCF) method can get
5-7 0.2225 0.2379 0.2403 0.8547 0.8547 0.8547
voltage improvement with good accuracy.
6-9 -0.2145 -0.2329 -0.2354 -0.7927 -0.8801 -0.8117 For the real line flow, switching the shunt capacitor at the
7-8 0.2193 0.2257 0.2434 0.7975 0.8289 0.8576 bus does not cause a major change. We can see this from
Table VII.
8-9 0.2414 0.2414 0.2414 0.8242 0.8584 0.8469

TABLE VII
Columns 2,3,4 are line flow changes when line 8-9 is LINE FLOW VARIANCE (VALUE: P.U.)
switched off by Flow Network Contribution Factors (obtained
from matrix X and X1) method and power flow method
Line Method 1 Method 2
respectively. Columns 5,6,7 are similar except for line 5-7
being switched off. 4-5 0.0106 -0.0040
We can see from Table IV that for the line switching 4-6 -0.0077 0.0032
control, the results of the Flow Network Contribution Factor 5-7 -0.0095 -0.0035
method are very accurate since they are very close to those of
6-9 -0.0031 0.0032
the power flow method. Since the line 5-7 carries a big flow,
when this line is off, there is a big flow transfer to other lines, 7-8 -0.0054 -0.0035

so other lines may be heavily loaded, as described in Table V. 8-9 -0.0022 -0.0034
Thus, the line switching control to relieve overload has to be
carefully used because it may result in overloading of other
lines. Method 1 is the Network Contribution Factor method.
Method 2 is the power flow method. We can see that the shunt
capacitance variance does not contribute much to real flow
 5

change of the line. It contributes to the variances of the bus

voltage and reactive power injection.

IV. CONCLUSION
This paper introduces a novel and comprehensive method
by using Network Contribution Factors. With the aid of the
FNCF and VNCF, necessary line parameter variance and bus
parameter variance can be calculated to adjust the line flow
and bus voltage without many trials of running the power flow
program. It can be used for re-dispatching load flow,
managing congestion, relieving overload, improving voltage,
controlling emergency, etc. The method is simple, fast and
accurate.

V. REFERENCES
[1] G.D. Irisarri, J.R. Latimer, S. Mokhtari, N. Muller, I.W. Slutsker, “The
future of electronic scheduling and congestion management in North
America”, IEEE Trans. Power Systems, vol. 18(2), pp. 444 – 451 May
2003.
[2] C.W. Taylor, D.C. Erickson, “Recording and analyzing the July 2
cascading outage [Western USA power system]”, IEEE Computer
Applications in Power, vol. 10 (1), pp. 26-30, Jan. 1997.
[3] C.W. Taylor, “Improving grid behavior”, IEEE Spectrum, vol. 36 (6), pp.
40-45, June 1999.
[4] U.S.-Canada Power System Outage Task Force, “U.S.-Canada Power
System Outage Task Force Interim Report: Causes of the August 14th
Blackout in the United States and Canada”, Nov 19, 2003, Available:
http://www.nerc.com
[5] NERC Disturbance Analysis Working Group, “Peninsula Florida
Disturbance – March 12, 1996”, NERC 1996 System Disturbances
Report, Aug 2002, Available: http:// www.nerc.com
[6] NERC Disturbance Analysis Working Group, “New York Power Pool
Disturbance – Aug 26 and Oct 30, 1996”, NERC 1996 System
Disturbances Report, Aug 2002, Available: http:// www.nerc.com
[7] N. Muller, V. H. Quintana, “Line and shunt switching to alleviate
overloads and voltage violations in power networks”, in Generation,
Transmission and Distribution, IEE Proceedings C, vol. 136(4), July,
1989, pp. 246 –253
[8] E.B. Makram, K.P. Thorton, H.E. Brown, “Selection of lines to be
switched to eliminate overloaded lines using a Z-matrix method”, IEEE
Trans Power Systems, vol. 4 (2), pp. 653 – 661, May 1989.
[9] S.K. Joshi, K.N. Patel, “Real time economic dispatch”, in Proc. 2000
Power System Technology, International Conf. PowerCon, vol. 3, pp.
1263-1268, 2000

VI. BIOGRAPHIES
Hongbiao Song (S'04) received his B.S. and M.S. degrees in electrical
engineering from North China Electric Power University, China in 1999 and
2002, respectively, and currently is a Ph.D. candidate in electrical engineering
at Texas A&M University. His research interests are power system analysis,
simulation, protection, stability and control.

Mladen Kezunovic (S’77–M’80–SM’85–F’99) received his Dipl. Ing. degree

from the University of Sarajevo, the M.S. and Ph.D. degrees from the
University of Kansas, all in electrical engineering, in 1974, 1977 and 1980,
respectively. He is the Eugene E. Webb Professor and Director of Electric
Power and Power Electronics Institute at Texas A&M University, College
Station, where he has been since 1987. His main research interests are digital
simulators and simulation methods for relay testing as well as application of
intelligent methods to power system monitoring, control, and protection.
Dr. Kezunovic is a Fellow of the IEEE and member of CIGRE-Paris.