www.elsevier.com/locate/epsr
Received 29 January 2002; received in revised form 28 March 2003; accepted 28 May 2003
Abstract
A method comprising of heuristic search technique and simulated annealing (SA) has been proposed for solving the problem of
optimal capacitor placement in radial distribution system and the effects of network and load unbalances, supply harmonics and
load non-linearities have been studied. The effect of network unbalances is most severe and can be incorporated very easily. But a
new formulation of the problem is needed to include the effects of load unbalances, supply harmonics and load non-linearities as
these are uncertain in nature.
# 2003 Elsevier B.V. All rights reserved.
Keywords: Distribution network; Capacitor placement; Harmonics; Unbalances; Simulated annealing; Heuristic search
* Corresponding author.
(i) Vi;min 5Vi 5Vi;max
E-mail addresses: tirtha_ghose@yahoo.com (T. Ghose), qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
hm
skgoswami_ju@yahoo.co.in (S.K. Goswami). where, Vi ah1 ½Vi;h ½2
0378-7796/03/$ - see front matter # 2003 Elsevier B.V. All rights reserved.
doi:10.1016/S0378-7796(03)00149-4
168 T. Ghose, S.K. Goswami / Electric Power Systems Research 68 (2004) 167 /173
Table 1
Bus data
1 0.0, 0.0
2 1840.0, 460.0
3 980.0, 340.0
4 1790.0, 446.0
5 1598.0, 1840.0
6 1610.0, 600.0
7 780.0, 110.0
8 1150.0, 60.0
9 980.0, 130.0
10 1640.0, 200.0
Fig. 2. Flowchart of the perturbation mechanism.
170 T. Ghose, S.K. Goswami / Electric Power Systems Research 68 (2004) 167 /173
Table 2 Table 5
Branch data Operational constraints and variable cost data
Table 4
Impedance matrix including mutual coupling
Z (V/mile)
Table 6
Results showing the effects of supply harmonics, mutual coupling, loading unbalances and load non-linearity
A
495 780 1.03261 0.9002
1 (a) 540 934 1.0399 0.925 6.24
1 (b) 546 689 1.042 0.9118 6.81
2 (a) 498 952 1.032 0.9009 4.75
2 (b) 504 585 1.034 0.8684 4.845
3 (a) 532 572 1.039 0.9240
3 (b) 538 287 1.0417 0.910
4 (a) 533 957 1.039 0.9241 1.05
4 (b) 539 681 1.0418 0.9105 1.813
5 (a) 536 697 1.039 0.924 3.892
5 (b) 542 426 1.042 0.9108 3.763
Case Supply harmonics Mutual coupling Loading unbalance Load non-linearity
citor placement problem was solved for balanced . The system cost increases in all the cases due to
sinusoidal system assuming load to be perfectly linear. increased energy losses. Cost increase is minimum in
The effects of network and load unbalances, supply case where the network unbalance is ignored and
harmonics and load non-linearity has then been inves- maximum in case where the network unbalance is
tigated on this compensated system and the results thus considered.
obtained are reported in Table 6. The required capacitor . The difference in the THD values of cases shows that
value changes with the changing of the non-linearity and the network unbalances has great effect on the THD
degree of unbalance in loading as shown in bar graph of values also.
Fig. 5. The following observations can be made: . With unbalanced load, variation of voltage from
peak to light load condition is more than the
variation of voltage with balanced load (Fig. 5).
Fig. 3. Capacitor kvar needed for different loading unbalances Fig. 4. Capacitor kvar needed for different loading unbalances
considering mutual coupling effects of the network. ignoring mutual coupling effects of the network.
172 T. Ghose, S.K. Goswami / Electric Power Systems Research 68 (2004) 167 /173
Appendix A
This paper investigated on the necessity of considering Qdown is the capacitor kvar in the midway between
both unbalance and the presence of harmonics in Qlow and Qold, obtained as,
distribution system while solving the capacitor place- Qold Qlow
ment problem and has proposed a method for solving Qdown Qold (2)
2
the problem using heuristic search along with SA
technique. Test results indicate that the capacitor kvar At the beginning of the search, Qlow is set equal to
obtained by ignoring network unbalance and supply Qmin and Qhigh is set equal to Qmax.
harmonics gives overcompensation and thus is not Let Sold, Sup and Sdown are the solutions correspond-
optimal. A novel heuristic search technique has also ing to Qold, Qup and Qdown.
been developed to quickly identify the feasible capacitor The search algorithm continues as follows:
setting. This search technique has been suitably used to
1) If either Sold, Sup or Sdown is feasible, stop, other-
enhance the solution speed of the SA technique.
wise.
Though the presence of network unbalances may be
2) (i) If Sold is superior of the three solutions: move
incorporated very easily, it is very difficult to incorpo-
Qlow and Qhigh toward Qold as Qlow /Qdown,
rate the presence of supply harmonics, load non-
Qhigh /Qup, Qold /Qold and calculate Qup and
linearity and loading unbalances as these data are Qdown using Eqs. (1) and (2). (ii) If, Qdown is superior
generally not known to the utilities and the great of the three solutions, move Qlow and Qhigh towards
uncertainty involved in these data. It is, however, not Qdown as Qlow /Qlow, Qhigh /Qold, Qold /Qdown.
impossible to estimate a possible range of variation of (iii) If, Qup is superior of the three solutions, move
these data in a utility system from sample test measure- Qdown and Qhigh towards Qup as Qlow /Qold,
ments. The authors feel that a new formulation of the Qhigh /Qhigh, Qold /Qup.
capacitor placement problem is possible where the
supply harmonics, load non-linearity and loading un- The search is continued until a feasible solution is
balances may be incorporated as fuzzy data and the obtained, or the best infeasible solution is found.
capacitor placement problem may be solved to have the
most acceptable solution instead of the optimum solu-
Appendix B