Alleviation of line overloads and voltage violations

by corrective rescheduling

P.R. Bijwe
D.P. Kothari
L.D. Arya

Indexing terms: Overloads, Voltage violation, Corrective rescheduling, Optimisation

AQS = vector of reactive power compensation

Abstract: This paper presents simple and efficient changes
algorithms for the alleviation of line overloads At = vector
of tap changer position correc-
and voltage violations by corrective rescheduling. tions
The proposed approach utilises the decoupling of = weighting factor for control changes
real and reactive power and the decomposition WUP
NC = total number of controls for voltage
between optimisation without security constraints subproblem
and optimisation to satisfy security constraints. NB = total number of buses
Highlights of the proposed approach are: (i) a =weighting factors for limiting Q at
choice of performance index which ensures that WQP
PV-bus
alleviation of some of the existing violations does =control variable for voltage sub-
not create any fresh violations, thus avoiding the UP
problem
need for cycling in optimisation and, (ii) the use of QGP = reactive power generation at PV-buses
a classical optimisation technique for faster solu- wv, = weighting factor for correcting the nth
tions. Results for two sample test systems have load bus voltage
been presented to validate the proposed algo- v:, vi = base case and corrected voltages at nth
rithms. load bus
NS = number of switched capacitors/
reactors
AQL = change in the reactive power loss
List of principal symbols E, E =lower and upper limits for nth load
- bus voltage
ACT = change
in total operating cost of gen- AQGp and AQGp= lower and upper limits of reactive gen-
eration -
eration changes at pth bus
WF = weighting factor [SQkp] = sensitivity matrix relating [AQJ with
fLf! = ith line MW-flow with corrected and CAUl
base case generation schedule CSGl = sensitivity matrix relating [ A V ] with
ji = ith line MW-flow limit LAVI
NL = number of transmission lines t" = setting of nth OLTC
NG = number of generators h i = ith line charging susceptance
APGk = generation correction at kth bus bi = ith line series subsceptance
APL -= incremental transmission losses
-
APGk and APGk= lower and upper limits for kth bus gen-
eration correction 1 Introduction
Ck(PGd = kth generator operating cost
The operating point of a power system will undergo a

k} = cost coefficients of kth generator change due to various contingencies and disturbances on
the system. If the system survives the outage or dis-
turbance, it will operate in a new steady state in which
k] = base case and corrected real power
generation of kth generator
= incremental cost of kth generator
one or more transmission lines may be overloaded and
hence voltage constraints at some buses may be violated.
ICk System dispatchers will resort to corrective rescheduling
aik = generation shift factor
-40, = A-loss coefficient for removing constraint violations.
= inverse of penalty factor at kth gener- The problem of corrective rescheduling for the allevia-
bk
ator tion of overloads and voltage limit violations can be
= voltage corrections at PV-buses solved by decomposed or nondecomposed approaches. In
A VG the nondecomposed approach the unified single optim-
isation problem is solved with security constraints. This
0IEE, 1993 requires excessive computational storage and time. The
Paper 9284C (Pg),received 7th August 1992 decomposed approach is usually preferred keeping in
The authors are with the Department of Electrical Engineering and view any computational requirements. Such an approach
Centre for Energy Studies, Indian Institute of Technology, Delhi, India involves the solution of an optimisation problem without
IEE PROCEEDINGS-C, Vol. 140, NO.4, JULY 1993 249
 111
security constraints followed by solution of a corrective
rescheduling problem to remove constraint violations
with minimum deviation from the previously optimised
schedule.
The alleviation of overloads and voltage limit vio-
lations can be achieved through the solution of a single
large optimisation problem involving simultaneous real
and reactive power rescheduling. However, the two sub-
problems (overload and voltage limit violation
alleviation) are usually solved separately and sequenti-
ally, taking advantage of the famous (P-6) and (Q-V)
decoupling principle, thereby gaining substantial savings
in computational time and storage.
A large number of research papers are available on the
subject of base case aptimisation and corrective
rescheduling. Such methods [2, 5, 8-11] employ various
performance indices for optimisation. For active power
dispatch the minimisation of operating cost is the usual
criterion. For the reactive power subproblem it is the per-
formance index, based on loss or voltage deviations, that
provides the consideration.
Most methods of base case real and reactive power
optimisation utilise linear, nonlinear or quadratic pro-
gramming techniques [2-5,8-111. Medicherla et al. [6,7]
have criticised such optimisation techniques on the
grounds that these methods are computationally expen-
sive. They have developed an online overload alleviation
model using Newton-Raphson load flow. However, the
optimisation aspects cannot be completely ignored.
From the literature survey it appears that simple and
efficient algorithms, similar to the ‘classical economic dis-
patch‘ (ED), are not yet available for real time implemen-
tation. This is important because solution of this problem
provides the starting point for overall ‘security con-
strained optimisation’ (SCO). Hence, the main objective
here is to develop such algorithms. In this context it must
be emphasised that evaluation of an approximate sched-
ule, which is reasonably accurate and can be imple-
mented in real time to mitigate emergencies, is far more
important than an evaluation of a very accurate schedule
which cannot be implemented in real time. For achieving
this objective the following features are necessary and
therefore have been incorporated in the proposed
approach :
(i) Exploitation of weak P-6 and Q-V coupling to
solve the real and reactive subproblems separately or
sequentially depending on the requirements. This con-
siderably reduces the dimensionality of the problem.
(ii) Inexplicit representation of the network through
loss formulae and sensitivity coefficients. This eliminates
the need for computationally expensive explicit network
solutions during optimisation process.
(iii) Simpler and efficient optimisation strategies
similar to classical ED.
(iv) All available algorithms in the literature on
security constrained optimisation show that there is a
possibility of producing new violations while removing
existing violations. In the proposed approach all import-
ant monitored quantities are included in the performance
index, thereby eliminating such a possibility.
(v) To improve computational efficiency and to reduce
the dimensionality of the problem the following two step
approach is adopted: first the optimisation problem is
solved without security constraints, and secondly, with
this solution in hand, the optimisation problem is re-
solved such that violations, if any, are removed
with minimum deviation from the previously optimised
schedule.
250
, .
2 Alleviation of line overloads
Line overloads are eliminated by rescheduling real power
generators, phase shifters, network switching and, as a
last resort, by curtailment of interruptible loads.
2.1 Problem formulationand solution methodology
It is assumed that the problem of optimisation without
security constraints has already been solved. It will now
be shown how the security constraint is satisfied with
minimum deviation from the previously optimised sched-
ule. In the proposed formulation only real power gener-
ation scheduling is considered for enhancing security.
The objective then is to seek a generation correction
schedule which minimises the following performance
index (PI):
NL
J = ACT + 0.5WF 1u:fil2i=l
(1)
This is subject to the following incremental real power
balance equation and inequalities on generation correc-
tions :
NG
1 APGk - APL= 0
k= 1
(2)
(3)
The first term in eqn. 1 signifies a penalty on deviation in
the operating cost from the previously optimised sched-
ule. The second term signifies penalties on normalised
line flows, which is normally used to rank the severity of
line outage contingency. Stott et al. [l] have observed
that line flow limits and voltage limits can never be pre-
cisely quantified and are not to be rigidly enforced.
Hence, it is fully justified to treat the normalised line
loadings as soft constraints. Optimisation with the pro-
posed performance index leads to optimum redistribution
of power flows which, simultaneously, curbs the excessive
overloading tendency in all the lines as far as possible.
However, it must be noted that the overload alleviation is
subject to the availability of adequate corrective capabil-
ity in the system. If such a capability is absent overload
alleviation cannot be guaranteed.
If the operating cost of the kth generator is given by
ck(pGk) = dOk + dlk PGk + d2k
then the change in generation schedule from the base
case value pbG to PEG, the total change in operating cost, is
given by
NG
ACT = [ICk APGk + dzk APiJ (4)
k= 1
The line flows with new generation schedule can be
obtained using generation shift factors [131, as follows:
NG
f f =f! + aik ApGk (5)
k=l
where ail, is the ‘generation shift distribution factor’
(GSDF).
Substituting for ACT and ff from eqns. 4 and 5 in
eqn. 1 we have
Nr.
IEE PROCEEDINGS-C, Vol. 140, No. 4, JULY 1993
Using the real power transmission loss formula [12] Step 9: Check for convergence of PI in successive iter-
(below), the change APi due to generation changes can be ations. If there is no convergence increase W F and go to
obtained as follows: step 3
r NG 12 Step 10: Identlfy the lines still overloaded and for such
lines
Set J =kiJ
where 0 < ki < 1. This decides the per unit reduction in
effective limit. It implies higher weightages for these over-
loaded lines. Then go to step 3.
From the proposed algorithm it can be seen that it is
(9) very well suited for real time applications because of its
simplicity and computational efficiency. This stems from
Hence, substituting from eqn. 8 into the incremental the following algorithm features:
power balance equation, eqn. 2, we have (i) The linear set of optimality equations which may be
NG NG solved very efficiently for real power generation correc-
1 APGk - 1OAok APGk = 0
k= 1 k=l
tions. Only the NG number of such equations have to be
solved.
or (ii) The algorithm does not require explicit network
representation and computationally expensive power
flow solutions.
If the corrections are out of limits then network switching
where ak = 1 - aAOk. That is ak puts a penalty on corres- or, in extreme cases, load curtailment may have to be
ponding generation change. If losses are neglected, a, = taken to remove the overloads.
1. It is an inverse of the penalty factor (PF).
The optimality equations are obtained by differentiat-
ing the PI of eqn. 6 with respect to APG, and equating it 3 Alleviation of voltage violations
to zero. That is
A good voltage profile is important for three reasons: (i)
aJ better security, (ii) good quality of supply, and fiii) low
-- -0 for rn = 1, . ..,NG and # slack transmission loss. Hence, bus voltages must be main-
aApGrn
tained in a narrow band around specified value under all
It is easy to see that all NG generations cannot be operating conditions. This has been recognised as one of
rescheduled independently. One of these has to be the most important operational problems.
considered as a slack (dependent) variable represented as As explained earlier, a two step algorithm is adopted
S. in order to reduce the size of the SCO problem. The solu-
After some simplification of optimality equations we tion for the first step, i.e. optimisation without security
have constraints by any one of the well established methods, is
NG assumed to be available. It will now be shown how the
x B k m A P G k = e mf o r r n = l , ..., NG and # s (11) second step of satisfying the security constraints can be
k= 1 implemented efficiently. Real powers have been already
k+s
scheduled as explained in Section 2.
where Bk, and e, are constant terms. The objective in this case is to obtain a set of reactive
It may be noted that eqn. 10 (incremental power control corrections, AU (i.e. AVG , AQs, At), so as to bring
balance equation) and eqn. 11 are linear and hence can all the bus voltages within specified limits. This is
be solved very efficiently for evaluating the NG gener- achieved by minimising the following PI :
ation corrections. NC NG
J = 0.5
p= 1
W U , AU; + 0.5 1 W Q , AQ;,
P= 1
NB
2.2 Computationalalgorithm
The main steps in the computational algorithm are as
+ 0.5 1 n=NGil
WV,(Vi - cu,J2 (12)
follows : The minimisation is subject to the following incremental
Step I : Read input data. reactive power balance equation, inequalities on control
Step 2: Set low value of W F ; say 0.1. corrections and reactive power generation changes:
Step 3: Solve the set of eqns. 10 and 11 and obtain
NG NG iNS
generation corrections.
Step 4 : Check the limits on generation corrections. In
case of violation, set the correction to the limiting value
and go to step 3 for solving the reduced set of equations.
Step 5 : Calculate APGm using eqns. 10 and 11.
Step 6: Calculate the modified flows using eqn. 5.
Step 7: Calculate PI according to The PI consists of three terms. The first term signifies a
NL penalty on the total control effort. Individual control cor-
PI = c CfrKl’
i= 1
rections are kept within limits by adjusting the weighting
factors, W U , . Excursions of control in both directions
Step 8 : Check if all line flows are within limits. If ‘yes’, are penalised. Similarly, the second term limits the vio-
stop. lation of reactive generations at all PV buses. The weigh-
IEE PROCEEDINGS-C, Vol. 140, No. 4, JULY 1993 251
ting factor, W Q , , will be a nonzero positive value only for p = 1, ..., NG and p = NG + NS + 1, ..., N C . Also
when it is found that reactive generation violation has
taken place at the pth bus, otherwise W Q , is set to zero.
The last term puts a penalty on voltage deviations at PQ
buses. Here also any voltage deviations are considered as
soft constraints. The deviation at a bus is brought within
+
for p = (NG l), ...,(NG + NS).
It should be noted that
limits by adjusting the respective weighting factor, WV,.
The deviation has been measured from some average AU, = AVG, for p = 1, ...,NG
value of bus voltage, V , ,I , which is obtained as follows: A U , = A Q s p f o r p = N G + l , ...,( N G + N S )
-
AU, = At, forp = N G + N S + 1, ..., N C
The optimality equations are obtained by differentiating
Sensitivity coefficients, expressing changes in PV bus PI with respect to control variables and equating it to
reactive generations, and PQ bus voltage with respect to zero. It may be noted that all the control variables
changes in control variables, are evaluated using the per- cannot be rescheduled independently because of the
turbation technique. With these coefficients we have equality constraint. Hence, N C - 1 optimality equations
(Appendix 7.3): are obtained.
NC NC
for m = 1, . . . , N C
AQGk =
p= 1
SQkp Aup (16) 1
p= 1
Dpm Aup = em
m+s
NC Pfs

AV, =
p= 1
1 SV,, A U , (17) where D,, and e, are constant terms.
These N C - 1 equations are solved together with the
SQ,, and SV,, are elements of the sensitivity matrices incremental reactive power balance equation (eqn. 22). It
[SQ] and [SV] as given in Appendix 7.3. may be noted that these N C equations are also linear in
Hence, the voltage at any PQ bus with control correc- nature and can be solved very efficiently for N C control
tion is written as follows: corrections. Because of the computational efficiency and
NC simplicity, the proposed method is suitable for real time
V i = V: + P=l SV,, AU, applications.

Using eqns. 16-18 the PI can be expressed in terms of 3.1 Computational algorithm
control corrections as follows: Step 1 : Read input data: (i) line data, (ii) base case
voltage profile, (iii) limits on voltages, control corrections
and reactive generations, (iv) reactive loss formula coeffi-
cients, (v) sensitivity factors SV and SQ.
Step 2: Calculate I , as defined in eqn. 21.
Step 3: Calculate h, coefficients given in eqn. 22.
n=NG+l Step 4: Select a dependent control bus s.
where DV: = V: - V , ,,,. Step 5 : Select all weighting factors W V , = 1.0, W Q k =
Now the objective is to express the equality con- 0, wv, = 1.0.
straints as a linear combination of control corrections. Step 6: Solve N C - 1 linear optimality equations with
The total reactive power losses are incremental power balance equation and obtain control
corrections.
Step 7: Check for limit violations for control correc-
tions. If there are no limit violations, then go to step 8.
Otherwise increase weighting factors, W U , for violated
where bi = bi - bshi.Eqn. 20 can be linearised and AQL is corrections and repeat from step 6.
written as Step 8: Calculate AQGk using eqn. 16 for k = 1, ...,
NC NG and check for limit violations. If ‘yes’, increase corre-
AQL = 1 1, Aup
p= 1
(21) sponding weighting factors, W Q k and repeat from step 6.
Otherwise go to step 9.
where the 1,s are the constants obtained at base case by Step 9: Using eqn. 17 calculate all load bus voltage
differentiating Q i , given in eqn. 20, and by utilising the changes. If all bus voltages are within limits, go to step
sensitivity coefficients (eqn. 25, Appendix 7.3). 10, otherwise increase the corresponding weighting
Substituting for AQ?, from eqn. 16 and AQL from eqn. factors, W V,, and repeat from step 6.
21 in eqn. 13, the equality constraint is written as follows: Step 10: Stop.
NG NG+NS NC
Simultaneous overload and voltage deviation alleviation
k= 1
[%?rp Aup3 1
p=NG+l
Aup- 1 1, Aup= 0
p= 1 algorithm has also been developed. The results, however,
did not justify the increased complexity introduced by
Simplifying, we obtain combining the two subproblems.
NC
1 h, AU, = 0
p= 1
4 Results

where The algorithms developed in Sections 2 and 3 have been

NG
implemented on 6 and 25-bus test systems (Appendices
7.1 and 7.2) to demonstrate their effectiveness in alleviat-
ing line overloads and voltage violations.
252 IEE PROCEEDINGS-C, Vol. 140, NO.4, JULY 1993
 The base case and corrected generation schedules for and voltage violations of the 25-bus test system. The
overload minimisation are depicted in Table 1 for the lower limits of reactive generations (five generation buses)
6-bus system. The corrected schedule has been obtained are -0.25, -0.45, -0.30, -0.50 and -0.50 P.u., respect-
with the help of the algorithm of Section 2. Table 2 ively, The upper limits are 1.5, 1.0, 1.0, 1.0 and 0.5 P.u.,
shows line flows with base case and corrected generation respectively. The desired voltage at each load bus is
schedules. It is observed from this Table that line over- between 0.98 and 1.02 P.u.. The lower and upper limits
load alleviation is achieved in the lines marked '*' with for PV-bus voltages are 0.98 and 1.15 P.u., respectively.
corresponding corrective rescheduling given in Table 1. The maximum real power generation limits at genera-
The algorithm of Section 3 has been employed for the tion buses are 3.00, 1.25, 1.75, 0.75 and 2.50p.u., respec-
alleviation of voltage violations. The system includes two tively. The minimum limits are 0.50, 0.20, 0.30, 0.10 and
OLTCs and two switchable reactors in addition to 0.40 P.u., respectively.
PV-bus voltage settings as control variable's. Table 3 Table 5 gives: (i) the base case generation schedule,
gives the base case and corrected values of control vari- and (ii) the corrected generation schedule are obtained by
ables. t4 and t , denote OLTC settings connected in lines the overload alleviation algorithm of Section 2.
4 and 7, respectively. Qc4 and Qe6 are reactive com-
pensations at buses 4 and 6, respectively. Table 4 depicts Table5: Base case and corrected real power generation
the voltage profiles with and without control corrections. schedules for alleviating line overloads in the 25-bus system
It is noted from Table 4 that initially all load bus volt-
Bus Base case Corrected
ages were violating the limits and with corrective real power real power
rescheduling all load bus voltages were brought within generations generations
limits (0.95 < V , < 1.05).
p.u. p.u.
The new algorithms developed in Sections 2 and 3 2.673 2.723
have also been employed for alleviating line overloads 0.993 0.936
1.479 1.513
Table 1 : Base case and corrected generation schedules for 0.391 0.480
alleviating line overloads in the 6-bus system 1.930 1.784

Generator Base case real Corrected real

power generations power generations
p.u. p.u. Table 6 presents line flows with three real power gen-
1 0.3362 0.4472 eration reschedules presented in Table 1. The overload
2 0.3801 0.2601 alleviation was achieved by the corrected schedules in '*'
marked lines.
Table2: Line flows with base case and corrected gener-
ation schedules for the 6-bus system Table 6: Line flows with base case and corrected real power
Line Line flows with Line flows with Line flow generation schedules for the 25-bus system
base case corrected limits Line Base case Line flows with Line flow
schedule schedule line flows corrected schedule limits
1 0.1594 0.2119 0.27 p.u. p.u.
2 0.1769 0.2353 0.30 1 -0.366 -0.358 0.46
3 0.0412 0.0558 0.08 2 0.421 0.41 1 0.49
4 -0.0559; 0.0124 0.04 3 -0.1 03 -0.045 0.17
5 -0.2030; -0.1 383 0.15 4 0.061 0.069 0.08
6 0.1618; 0.1 127 0.15 5 0.299 0.287 0.36
7 0.1573 0.2071 0.25 6 0.360 0.358 0.42
* Line overload alleviationachieved 7 0.214 0.175 0.30
8 0.340 0.331 0.39
9 0.338 0.329 0.38
Table3: Magnitudes of base case and corrected control 10 0.192 0.231 0.26
variables for alleviation of voltage violations in 6-bus 11 0.409 0.415 0.49
system 12 -0.4864+ -0.408; 0.43
Sr. Control Base case Corrected 13 0.363 0.368 0.44
variable value value 14 0.21 4 0.219 0.26
15 0.306 0.293 0.37
p.u. p.u. 16 0.796 0.734 1.03
0.98 1.04873 17 0.578; 0.506; 0.53
0.97 1.0502 18 0.061 0.023 0.10
0.00 0.007 19 0.060 0.059 0.075
0.00 0.01 50 20 0.124 0.116 0.150
1.o 1.0110 21 0.071 0.074 0.075
1.o 1.0203 22 0.070 0.058 0.08
23 -0.079 -0.075 0.11
24 0.071 0.063 0.08
Table 4: Base case and corrected voltage profile for 6-bus 25 0.021 0.013 0.025
system 26 0.023 0.016 0.026
27 0.1 89 0.196 0.23
Bus Base case voltages Corrected voltages 28 -0.1 12 -0.1 04 0.13
p.u. p.u. 29 0.1 91 0.165 0.20
1 0.9800 1.04874 30 0.038 0.013 0.05
2 0.9700 1.05025 31 0.1 08 0.113 0.15
3 0.9082 0.9756 32 0.1 16 0.127 0.20
4 0.9182 1.0034 33 -0.1 35 -0.1 25 0.16
5 0.8920 0.9731 34 0.050 0.052 0.095
6 0.9085 0.9956 35 -0.100 -0.098 0.105

I E E PROCEEDINGS-C, Vol. 140, N O . 4, JULY I993 253

Table 7: Base case and modified PV-bus voltage schedules 2 KALTENBACH, J.C., and HAJDU, L.P.: ‘Optimal corrective
for alleviating voltage violations in the 26-bus system rescheduling for system security’, IEEE Trans., 1971, PAS-90, (2),
Bus Base case PV-bus Modified PV-bus pp. 843-851
3 THANIKACHALAM, A., and TUDOR, J.R.: ‘Optimal resched-
voltages voltages uling of power for system reliability’, IEEE Trans., 1971, PAS-90,
p.u. p.u. (9, pp. 2186-2192
1 1.020 1.03 4 DAVIES, H., and CHEN, M.: ‘An optimization technique and
2 0.969 1.002 security calculations for dispatching computers’%IEEE Trans., 1972,
3 0.995 1.050 PAS-91, (3), pp. 883-889
4 0.960 1.015 5 MAMANDUR, K.R.C., and BERG, G.J.: ‘Economic shift in electric
5. 0.981 1.007 power generation with line flow constraints’. Paper F77617-6, IEEE
PES summer meeting, 1977
6 MEDICHERLA, T.K.P., BILLINTON, R., and SACHDEV, M.S.:
Table 8: Bus voltage profiles with base case and modified ‘Generation rescheduling and load shedding to alleviate line
PV-bus voltage schedule$ for the 25-bus system overloads-analysis’,IEEE Trans., 1979, PAS-%, (12),pp. 1876-1885
~ ~~
7 MEDICHERLA, T.K.P., BILLINTON, R., and SACHDEV, M.S.:
Bus Voltage with Voltage With ‘Generation rescheduling and load shedding to alleviate line
base case PV-bus corrected PV-bus overloads-system studies’, IEEE Trans., 1981, PAS-100, (I), pp.
, settings settings 36-41
8 SULLIVAN, R.L.: ‘Power system planning’ (McGraw-Hill, New
p.u. p.u. York, 1977)
i 1.020 1.034 9 NARITA, S., and HAMAAM, M.S.: ‘A computational algorithm for
i 0.969 1.002 real time voltage control of system voltage and reactive power’,
3 0.995 1.050 IEEE Trans., 1971, PAS-90, (6), pp. 2495-2508
4 0.960 1.010 10 HORTON, J.S., and GRIGSBY, L.L.: ‘Voltage optimization using
5 0.981 1.007 combined linear programming and gradient techniques’, IEEE
6 0.961 1,009 Trans., 1984, PAS-103, (7),pp. 1637-1643
7 0.965 0.998 11 BIJWE, P.R., KOTHARI, D.P., NANDA, J., and LIN-
8 0.968 1.ooo GAMURTHY, K.S.:‘Optimal voltage control using a constant
9 0.962 0.994
sensitivity matrix’, J. Electr. Power Syst. Res., 1986,11, pp. 195-203
10 0.974 1.004 12 ARYA, L.D.: ‘Security constrained power system optimization’.
11 0.977 1.006 PhD Thesis, IIT, Delhi, India, 1991
12 0.970 1.002 13 WOOD, A.J., and WOLLENBERG, B.F.: ‘Power generation, oper-
13 0.964 1.015 ation and control’ (John Wiley & Sons, New York, 1984)
14 0.953 0.989
15 0.955 0.987
16 0.972 0.997
17 0.978 1.007
18 0.961 0.999 7 Appendix
19 0.964 1.010
20 0.940 1.002 7.1 Data for the 6-bus system (at 100 MVA base)
21 0.951 1.002
22 0.957 1.000
23 0.989 1.015
24 0.970 1.007 Table 9: Load data
25 0.980 1.012
Bus Loads

Table 7 depicts (i) the base PV-bus voltages, and (ii) PD QD

the corrected PV-bus voltage by the voltage violation p.u. p.u.

alleviation algorithm of Section 3. Table 8 shows the bus 1 0.000 0.000
2 0.000 0.000
voltages with the PV-bus voltage schedules as given in 3 0.275 0.065
Table 7. It is observed that by both the corrected sched- 4 0.000 0.000
ules the load bus voltages are brought within bounds 5 0.150 0.090
(0.98 < 6 < 1.02). 6 0.25 0.025

5 Conclusion
Table 10: Line data
Two new simple and efficient algorithms for the allevia- Line From To Line parameter
tion of line overloads and voltage violations by corrective bus bus
rescheduling have been proposed. They utilise the number number ri xi
decoupling of active and reactive optimisation problems. p.u. p.u.
In order to reduce the size and complexity of the problem 1 1 6 0.123 0.518
further, a decomposition between optimisation without 2 1 4 0.080 0.370
security constraints and optimisation to satisfy security 3 4 6 0.097 0.407
constraints was employed. The algorithms entail per- 4 6 5 0.000 0.300
5 5 2 0.282 0.640
formance indices which ensure that alleviation of the vio- 6 2 3 0.723 1.050
lations do not result in new violations. A classical 7 4 3 0.000 0.133
optimisation technique was applied in the algorithms.
The results obtained for the two sample test systems with
the proposed algorithms clearly demonstrate their poten- Table 11 : Generator data
tial for power system control.
Generator Real generation Reactive generation
6 References Maximum Minimum Maximum Minimum

1 STOTT, B., ALSAC, O., and MONTICELLI, A.J.: ‘Security p.u. p.u. p.u. p.u.
analysis and optimization’, Proc. IEEE, 1987, 75, (12), pp. 1623- 1 1,000 0.100 1.000 -0.2
2 1,000 0.1000 1.000 -0.2
1644

254 IEE PROCEEDINGS-C, Vol. 140, NO.4, JULY 1993

 I

The operating cost characteristics of the generators are Table 13: Load data
given below : Bus Real load Reactive load
Cl(PG1 = PG1 + O.O5P61 PLI QD

CZ(PG2) = PG, + 0.1PiZ 1

p.u.
2.00
p.u.
0.65
In the 6-bus system the shunt compensation is provided 2 0.10 0.03
3 0.50 0.17
at buses 4 and 5. The limits on reactive power injections 4 0.30 0.10
are : 5 0.25 0.08
6 0.15 0.05
0 < Qc4 p.u. < 0.05 7 0.15 0.05
8 0.25 0.00
0 < Qca P.U. < 0.055 9 0.15 0.05
10 0.15 0.05
OLTC are provided in lines 4 and 7 at buses 6 and 4, 11 0.05 0.00
respectively. The limits on tap settings are: 12 0.10 0.00
13 0.25 0.08
0.9 < t, < 1.1 14 0.20 0.07
15 0.30 0.10
0.9 < t , < 1.1 16 0.30 0.10
17 0.60 0.20
Series resistance and shunt susceptance of both 18 0.15 0.05
OLTC = 0.0 P.U. 19 0.15 0.05
20 0.25 0.08
21 0.20 0.07
7.2 Data for the 25-bus system ( 100 MVA base) 22 0.20 0.07
23 0.15 0.05
Table 12: Line data 24 0.15 0.05
25 0.25 0.08
Line From bus To bus r, Xi bshi
number number
p.u. p.u. p.u.
1 1 3 0.0720 0.2876 0.0179 Table 14: Generator data
2 1 16 0.0290 0.1379 0.0337
3 1 17 0.1012 0.2794 0.0148 Generator Real generation Reactive generation
4 1 19 0.1487 0.3897 0.0224
5 1 23 0.1085 0.2245 0.0573 Maximum Minimum Maximum Minimum
6 1 25 0.0753 0.3593 0.0873 p.u. p.u. p.u. p.u.
7 2 6 0.0617 0.2935 0.0186 1 0.0 4.0 0.0 3.0
8 2 7 0.0511 0.2442 0.0155 2 0.0 2.0 -0.5 1.o
9 2 8 0.0579 0.2763 0.0175 3 0.0 2.0 -0.5 1.o
10 3 13 0.0564 0.1487 0.0085 4 0.0 2.0 -0.5 1.o
11 3 14 0.1183 0.3573 0.01 85 5 0.0 2.0 -0.5 1.o
12 4 19 0.0196 0.0514 0.01 13
13 4 20 0.0382 0.1 007 0.0220
14 4 21 0.0970 0.2547 0.0558
15 5 10 0.0497 0.2372 0.0577
16 5 17 0.0144 0.1 269 0.1 335 7.3 Newton-Raphson method
17 5 19 0.0929 0.2442 0.0140
18 6 13 0.0263 0.0691 0.0040 In the Newton-Raphson method of power flows given in
19 7 8 0.0529 0.1465 0.0078 polar form, the linearised equations are expressed as
20 7 12 0.0364 0.1736 0.0110 follows :
21 8 9 0.0387 0.1847 0.0118
22 8 17 0.0497 0.2372 0.0572
23 9 10 0.0973 0.2691 0.0085
24 10 11 0.0898 0.2359 0.0135
25 11 17 0.1068 0.2807 0.0161
26 12 17 0.0460 0.2196 0.0135
27 14 15 0.0281 0.0764 0.0044 This equation can be partitioned in the form :
28 15 16 0.0256 0.0673 0.0148
29 17 18 0.0806 0.2119 0.0122
30 18 19 0.0872 0.2294 0.0132
31 20 21 0.0615 0.161 3 0.0354
32 21 22 0.0414 0.1087 0.0238
33 22 23 0.2250 0.3559 0.0169
34 22 24 0.0970 0.2595 0.0567
35 24 25 0.0472 0.1458 0.0317

State changes are expressed in terms of reactive control

changes. [SI- &] are sensitivity submatrices, which are
obtained from the elements of [J3.
Also from eqn. 24, [A&] - [AU] sensitivity relations
can be written as

where [SQ] is submatrix obtained from the elements of

[A, and where, [AV] = [AV, AQs AtlT.
255