Professional Documents
Culture Documents
34 Bus Data With Line Length
34 Bus Data With Line Length
34 Bus Data With Line Length
M.Chis
M.M.A.Salama
S.Jayaram
~~~~~
1
n
m=l
n
- KeTI,21,Rfus(k,
k)
+
nz=l
-
RbA,,(k "-mFLD(m,
Kc(Qck)Qck
3)
) (4)
226 IEE Proc -Gene7 Transm Distrib , Vol 144, No 3, May 1997
T in eqn. 4 represents the total time period over which 3 Application of the method
the savings are calculated ( T = 8760 h for one year).
If the algorithm is used for online applications, eqn. The described method is applied to a 34-bus three-
4 will allow the calculation of the optimum compensa- phase radial feeder with lateral branches, shown in
tion level at each node during each time interval Fig. 3. Details of the feeder and the load characteris-
described by eqn. 3. This can also be used to determine tics are given in Table 1.
a near-optimal switching sequence.
-0
c
t Fig.3 Distribution network configuration
0
L
0
Y
- The total installed peak power is 5.4MVA, with an
U
0
a
0
average power factor of 0.85. The values the coeffi-
cients in eqn. 2 are Kp = lSO$IkW, K, = 0.07$/kWh,
and the variation K,(Q,) is shown by the curve a in
Fig. 2.
01 I I I
0 200 LOO 600 800 1000 The software implementation of the procedure has
capacitor size, 3-phase kVar been done using the Matlab'" environment for the ease
Fig.2 Capacitor annual cost/kVAr against capacitor size of operating with complex numbers; the loadflow
(i) = real cost results from our software have been compared with the
(ii) = approximated cost
ones obtained using two commercial packages,
EDSA'" and CYME'" , and proven to be consistent.
The capacitor size Qck will be chosen such that s k is All the calculations have been carried out in the per-
maximised at every step. The main problem in finding unit system with the three-phase base S, = 100MVA,
Qck that maximises S, resides in the nonlinearity of S, V B = 11kV.
with respect to Qck, due to the piecewise linear varia-
tion of K,(Q,), as shown in Fig. 2 [5]. To solve this an 4 Results
approximation was used, by interpolating through the
The initial power loss in the system is 221.72kW. The
midpoints of each interval of K,. The new variation application of the proposed method to the system
K,(Q,) is shown in Fig. 2 and is subsequently used in yields the following results:
eqn. 4 to solve for the estimated optimum value of Qck
from the condition (i) From eqn. 5 bus 26 needs an optimum compensa-
tion of 700kVAr (0.0070pu); however, to show the
effect of varying the compensation level on the savings,
(5) curve 1, Fig. 4, is shown. A value of 750kVAr (1 x 500
Once the Qck that maximises the objective function s k
+ 1 x 250kVAr for capacitor bank 1) is used to attain
11 272Wyear in savings and a peak power loss reduc-
is found, the nearest available kVAr rating multiple tion of 35.2kW. It is evident from curve (i), Fig. 4, that
units of capacitor is chosen and placed at node 'k'. using 750kVAr instead of 700kVAr does not offset the
Step 3: A load flow is performed in order to get the optimum savings significanfly, and yet it makes the
new values of the load currents and to check if the solution feasible.
voltage constraints are met. If any violation of the volt-
age constraints (*3% in urban areas, &6% in rural areas
[9]) occurs, the capacitor is removed and the next larg-
est loss node is selected as the next sensitive node, and
the procedure is repeated starting from Step 2.
Step 4: At every iteration the real dollar savings are
calculated using eqn. 4, where the actual &(e,)
is used
(Fig. 2, curve 1).
Step 5: The next largest loss node is selected as the next
sensitive node and Steps 2 4 are repeated.
Step 6: Total savings are calculated by adding s k for
each of the compensated nodes.
This process is repeated iteratively until the total sav-
ings reach a peak value; that will be the optimum com- capacitor bank size,pu
Fig. 4 Dollar savings against capacitor bank sizes
pensation scheme for the network and also the (i) capacitor bank 1, node 26
optimum number of capacitors. The flowchart in Fig. 1 (ii) capacitor bank 2, node 11
(iii) capacitor bank 3, node 21
illustrates the procedure described above. (iv) capacitor bank 4, node 20
IEE Psoc.-Gener. Transm. Distsib., Vol. 144, No. 3, May 1997 221
Table 1: Distribution system data
Load Line impedance Load factors
Node Length TI T2 T3
no. P (kW) Q (kVar) r (Sllkm) x (Q/km) (km) 0O:OO 08:OO 16:OO
08:OO 16:OO 24:OO
1 0 0
2 230 142.5 0.195 0.080 0.60 0.55 0.70 0.65
3 0 0 0.195 0.080 0.55 0 0 0
4 230 142.5 0.299 0.083 0.55 0.55 0.70 0.65
5 230 142.5 0.299 0.083 0.50 0.55 0.70 0.65
6 0 0 0.299 0.083 0.50 0 0 0
7 0 0 0.524 0.090 0.60 0 0 0
8 230 142.5 0.524 0.090 0.40 0.55 0.70 0.65
9 230 142.5 0.524 0.090 0.60 0.55 0.70 0.65
10 0 0 0.524 0.090 0.40 0 0 0
11 230 142.5 0.524 0.090 0.25 0.55 0.70 0.65
12 137 84 0.524 0.090 0.20 0.50 0.60 0.55
13 72 45 0.524 0.090 0.30 0.45 0.65 0.60
14 72 45 0.524 0.090 0.40 0.45 0.65 0.60
15 72 45 0.524 0.090 0.20 0.45 0.65 0.60
16 13.5 7.5 0.524 0.090 0.10 0.60 0.70 0.65
17 230 142.5 0.299 0.083 0.60 0.55 0.70 0.65
18 230 142.5 0.299 0.083 0.55 0.55 0.70 0.65
19 230 142.5 0.378 0.086 0.55 0.55 0.70 0.65
20 230 142.5 0.378 0.086 0.50 0.55 0.70 0.65
21 230 142.5 0.378 0.086 0.50 0.55 0.70 0.65
22 230 142.5 0.524 0.090 0.50 0.55 0.70 0.65
23 230 142.5 0.524 0.090 0.50 0.55 0.70 0.65
24 230 142.5 0.524 0.090 0.60 0.55 0.70 0.65
25 230 142.5 0.524 0.090 0.40 0.55 0.70 0.65
26 230 142.5 0.524 0.090 0.25 0.55 0.70 0.65
27 137 85 0.524 0.090 0.20 0.50 0.60 0.55
28 75 48 0.524 0.090 0.30 0.55 0.75 0.70
29 75 48 0.524 0.090 0.30 0.55 0.75 0.70
30 75 48 0.524 0.090 0.30 0.55 0.75 0.70
31 57 34.5 0.524 0.090 0.30 0.57 0.63 0.58
32 57 34.5 0.524 0.090 0.40 0.57 0.63 0.58
33 57 34.5 0.524 0.090 0.30 0.57 0.63 0.58
34 57 34.5 0.524 0.090 0.20 0.57 0.63 0.58
(ii) The second node to be compensated is bus 11, point will offset the benefits. The variation of the total
which requires an optimum capacitor of 400kVAr annual savings is represented in Fig. 5.
(0.0040pu; curve (ii), Fig. 4); a rated value of 400kVAr
(1 x 150 + 1 x 250kVAr for capacitor bank 2) is used
to obtain additional savings of 2634$/year and to
reduce the peak power losses by another 11.24kW. 15 000
c 5000
additional savings of 1130$/year and an additional g
peak power loss reduction of 5.64kW.
0
(iv) Bus 20 requires an optimum compensation of 2 3
70kVAr (0.00007pu; curve (iv), Fig. 4); an 150kVAr
bank (bank 4) is used instead to bring additional sav- -5 OOOL number of capacitor banks
ings of 133$/year and to reduce peak power losses by Fig.5 Total annual savings against number of capacitor banks
1.29kW. 0 compensation for maximum additional savings.
0 compensation for maximum peak power loss reduction
At this point the overall savings, represented in Fig. 5,
reach the maximum value of 15 169$/year; in other The concept of maximum power loss reduction at
words, the cost of installing new capacitors from this each step has also been applied to the system shown in
228 IEE Proc -Gener Transm D u m b , Val 144, No 3, May 1997
Fig. 3. Comparative results (total annual savings and 1.5 169$/year as opposed to a maximum of 3347$/year
power losses) with our method are shown graphically when the idea of maximum loss reduction is employed.
in Figs. 5 and 6. When the compensation level is cho- In our method we used more efficient compensators
sen such as to maximise the loss reduction at each step, that are based on the savings.
the compensation scheme for the system is as follows: Also, our method needed to calculate the system
(i) 1400kVAr (0.0140pu = 1 x 1000 + 1 x 2.50 + 1 x power flow only four times for the four sensitive nodes.
1SOkVAr for capacitor bank 1) at node 26 for a peak By comparison, a significantly larger number of itera-
power loss reduction of 41.07kW and savings of 3347$1 tions is needed when the algorithm proposed in [lo] is
year used.
(ii) 750kVAr ( 0 . 0 0 7 5 ~=~1 x 500 + 1 x 250kVAr for
capacitor bank 2) at node 11 for an additional loss 5 Complexity of the problem
reduction of 10.64kW and total annual savings of In this Section the effectiveness of using the proposed
2028 Slyear technique to reduce the problem complexity and hence
(iii) 300kVAr ( 0 . 0 0 3 ~=~2 x 15OkVAr for capacitor the computation time is demonstrated. The simplified
bank 3 at node 17 for a peak power loss reduction of formulas presented in the Appendix, along with the
another 1.17kW and total savings of 93 $/year concept of the sensitive nodes, considerably reduce the
(iv) 250kVAr (0.0025 = 1 x 250kVAr for capacitor number of power flow runs required during the execu-
bank 4) at node 4 for a peak power loss reduction of tion of the algorithm. Power flow by the Newton-
0.81kW and money losses of -1409$/year. This is the Raphson method requires o(n3) arithmetic operations
point of maximum loss reduction in the system and any per iteration for the inversion of the system Jacobian
addition of capacitors will increase the losses due to matrix by triangular factorisation. If this is applied to
overcompensation. It is important to notice here that the n nodes, the total computational cost will be o(n4)
although the power loss reduction is possible up to the for the selection of each capacitor. When eqn. 14 is
installation of capacitor bank 4, the savings drop con- used to calculate the power loss reduction, n summa-
tinuously from a peak of 3347$/year to 1409$/year in tions, each requiring o(n) operations, need to be calcu-
losses. lated. Therefore, the computational cost is reduced by
I two orders of magnitude to o(n2)operations.
220
6 Conclusions
Heuristic search strategies are applied in this paper to
200 determine the optimal capacitor placement and rating
3, for distribution systems. The solution to the capacitor
problem using heuristic rules involves searching
180 through a set of possible solutions. This is achieved in
this paper by examining the solution at a set of critical
nodes named sensitive nodes. The sensitive nodes are
160 selected based on the losses caused in the system by the
0 1 2 3
number of capacitor banks
reactive components of the load currents. The pro-
Fig.6 Peak power loss in system against number of capacitor hunks posed approach determines the compensation level
G compensation for maximum additional savings. from the condition of maximising the net savings, as
W compensation Cor maximum peak power loss reduction
opposed to only minimising the losses as in other pub-
lished methods [7, 101.
Fig. 6 illustrates the peak power losses in the system This method has the following advantages;
against the number of capacitors (as described above). (i) Since the number of sensitive nodes is relatively
For the sake of comparison, it should be pointed out small compared to the total number of nodes in the
that the use of the analytical method presented in [7] distribution system, the size of the problem is consider-
reported a reduction in the peak power losses of 40kW ably reduced. This makes this method very attractive
using a total compensation 1650kVAr. This was based when dealing with large distribution systems.
on load factors varying in different parts of the feeders (ii) Realistic sizes and locations for shunt capacitors are
in the range (0.45-0.63). The method presented in this considered in this algorithm. Unlike the analytical
paper gives a peak power loss reduction of 53.37kW approach, the heuristic rule-based approach accounts
using 1550kVAr for the optimum case (first four for the difference between the theoretical-optimal and
capacitors). The analytical method presented in [7] actual sizes and locations of the shunt capacitors.
decoupled the system into subsystems and the optimal Because of the iterative nature of the heuristic
capacitor sizes and ratings were calculated for each approach, the solution is updated (from the knowledge
subsystem. This decoupling technique produces a near- base) and improved (based on the rules) at each itera-
optimal solution; the heuristic technique proposed in tion.
this paper handles the whole system and therefore pro-
(iii) No voltage violation due to the addition of capaci-
duced better results.
tors is allowed by the algorithm.
It can be observed that the peak power loss reduction
using the objective function that focuses on financial (iv) This method requires less computation time com-
savings is less than that obtained when applying the pared to other analytical methods. This feature makes
concept of maximum power loss reduction, as seen it appropriate for online voltiVAr compensation.
from Fig. 6. However, the main difference between the (v) When this method was applied to a test system it
two concepts is in the dollar savings, shown in Fig. 5; provided better savings compared to two other pub-
for the optimum case our method provides a total of lished methods (one heuristic and one analytical).
IEE ProccGeneu. Trunsm. Distuih., Vol. 144, No. 3, M a y 1997 229
~
8 Appendix
Let I, = I,, + jIrk be the complex load current at bus
In this Appendix, we will derive a new formula for the ‘k’ before the addition of the capacitor. If in eqn. 11 we
loss reduction in the feeder due to capacitor compensa- substitute I, = -jIrk we get:
tion. We use the same outline proposed by Civanlar
et al. [l 11 for calculating the loss change resulting from
the transfer of a group of loads from one feeder to
another by reconfiguration. Let us assume that a radial
distribution feeder A, shown in Fig. 7, has the bus cur-
aGtss= 2Re
{ -.I&k
+ I:h&s ( k k )
m=l
n
RfUs( km)(Iam- j I T m )
I
rents 11,I,, ..., I,, represented by the vector:
I,T = [Il,I2,. . . ,I,,. . . ,In]
To determine the reduction in losses in the feeder A
(6)
Rfus(k,m)(-IThITm-3IrkIam)
k)
I
the vector of bus currents has to be modified as fol-
lows: apk, = -2 RkS(k”dTm + k)
: f +
1 = [ I l , I 2 , .. ,I, I,,. * ., I n ] (7)
m=l
(12)
where I, is the vector of bus currents after the addition where I, = I,, + jITmis the load current at bus ‘m’.
of the capacitor C of current I, at node ‘k’. The value expressed by eqn. 12 actually represents
The change in the power losses on feeder A due to the change in the system power loss when eliminating
the capacitor addition is: the reactive component of the load current at node ‘k’
0 by adding a capacitor of equal value. In this case, the
ap;ss = P&S - Pkxs (8) capacitor causes the loss reduction:
where do,,is the power loss in the distribution system n
before the addition of the capacitor and is given by
[ll]:
Qp! = -Apk,, = 2 C RfUS( k ,m)IrhIrm-I:&& ( k ,k )
m=l
230 IEE Proc -Gener Transm Distrib , Vol 144, No 3, May I997