Professional Documents
Culture Documents
ir
http://www.itrans24.com/landing1.html
a r t i c l e i n f o a b s t r a c t
Article history: To minimize line losses of power systems, it is crucially important to define the size and location of local
Received 6 June 2007 generation to be placed. On account of some inherent features of distribution systems, such as radial
Received in revised form structure, large number of nodes, a wide range of X/R ratios; the conventional techniques developed for
25 September 2008
the transmission systems generally fail on the determination of optimum size and location of distributed
Accepted 16 December 2008
generations.
In this study, a loss sensitivity factor, based on the equivalent current injection, is formulated for the
Keywords:
distribution systems. The formulated sensitivity factor is employed for the determination of the optimum
Distributed generation
Radial system
size and location of distributed generation so as to minimize total power losses by an analytical method
Power losses without use of admittance matrix, inverse of admittance matrix or Jacobian matrix. It is shown that, the
Optimal placement proposed method is in close agreement with the classical grid search algorithm based on successive load
Equivalent current injection flows.
Loss sensitivity factor © 2008 Elsevier B.V. All rights reserved.
0378-7796/$ – see front matter © 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.epsr.2008.12.007
Downloaded from http://iranpaper.ir
http://www.itrans24.com/landing1.html
T. Gözel, M.H. Hocaoglu / Electric Power Systems Research 79 (2009) 912–918 913
nb
Ploss = |Bi |2 · Ri = [R]T |[BIBC] · [I]|2 (5)
The proposed method is based on the equivalent current injec-
tion that uses the bus-injection to branch-current (BIBC) and i=1
branch-current to bus-voltage (BCBV) matrices which were devel- where Ri is the ith branch resistance and the branch resistance
oped based on the topological structure of the distribution systems vector is given in (6):
and is widely implemented for the load flow analysis of the distri- T
bution systems. The details of both matrices can be found in [16]. [R]nb×1 = R1 R2 · · · Rnb (6)
The method proposed here requires only one base case load flow to
determine the optimum size and location of DG. The total power losses can be written as a function of the real
and imaginary parts of the bus current injection:
2.1. Theoretical analysis Ploss=[R]T |[BIBC · [I]]|2 =[R]T |[BIBC] · [re(I)]+j[BIBC] · [im(I)]|2 (7)
In this section, the total power losses will be formulated as a where [re(I)] and [im(I)] are the vectors that consist of real and
function of the power injections based on the equivalent current imaginary parts of the bus current injection:
injection. The formulation of total power losses will be used for 2
determining the optimum size of DG and calculation of the system Ploss = [R]T (([BIBC] · [re(I)])2 + ([BIBC] · [im(I)]) ) (8)
losses. By substituting the equivalent bus injection expression (2) into
At each bus i, the corresponding equivalent current injection is (8), the total power losses can be rewritten as
specified by
P + jQ ∗
2
P cos() + Q sin()
Ii = i i
i = 1, 2, . . . , n (1) Ploss = [R]T [BIBC] ·
Vi |V |
where Vi is the node voltage, Pi + jQi is the complex power at each
2
T P sin() − Q cos()
bus i, n is the total number of buses, ‘*’ symbolizes the complex +[R] [BIBC] · (9)
|V |
conjugate of operator.
Downloaded from http://iranpaper.ir
http://www.itrans24.com/landing1.html
914 T. Gözel, M.H. Hocaoglu / Electric Power Systems Research 79 (2009) 912–918
n
at jth branch the power loss can be obtained as
P sin(k ) − Qk cos(k )
⎡ +2Rj · BIBC(j, k − 1) · k ·
2 |Vk |
n
Pk cos(k ) + Qk sin(k ) k=2
Plossj = Rj · ⎣ BIBC(j, k − 1)
|Vk | sin(i )
k=2 × BIBC(j, i − 1) · (14)
|Vi |
n 2 ⎤
Pk sin(k ) − Qk cos(k )
+ BIBC(j, k − 1) ⎦ (10) Sum of the above expression leads to the derivation of the total
|Vk | power losses per ith bus injected real power, ∂Ploss/∂Pi :
k=2
n
∂Ploss
nb
The total power losses are the sum of the each branch power =2 Rj · BIBC(j, k − 1) · re(Ik ) ·BIBC(j, i − 1)·
∂Pi
losses: j=1 k=2
⎡ 2 n
nb
n
Pk cos(k ) + Qk sin(k ) cos(i )
nb
n 2 ⎤ sin(i )
P sin(k ) − Qk cos(k ) × BIBC(j, i − 1) · (15)
+ BIBC(j, k − 1) k ⎦ (11) |Vi |
|Vk |
k=2
If the ith bus is not connected the jth branch then the elements
of BIBC matrix is zero (BIBC(j, i − 1) = 0) and the derivative of the
The voltage drop from each bus to the reference bus is obtained
corresponding element is equated to zero (∂Plossj /∂Pi = 0). Accord-
with BCBV and BIBC matrices as
ingly, the derivative of the total power losses per ith bus injected
[V ](n−1)×1 = [BCBV ] · [BIBC] · [I] (12) real power gives the sensitivity factor and can be expressed as
where BCBV matrix is responsible for the relations between branch ∂Ploss
nb n
currents and bus voltages. The elements of BCBV matrix consist =2 Rj dPBIBCi (j, k − 1) ·
∂Pi
of the branch impedances. Building algorithm of BCBV matrix can j=1 k=2
be found in [16]. In addition, building algorithm of BCBV matrix is
provided as follows for convenience: cos(i ) sin(i )
· re(Ik ) + · im(Ik ) (16)
|Vi | |Vi |
• Step 1. Read BIBC matrix, Zb branch impedance vector.
• Step 2. Convert Zb vector to a diagonal matrix Z by setting off- The sensitivity factor with the above relation can be shown in
diagonal elements to zero; (Z = diag(Zb)). matrix form as
• Step 3. Multiply transpose of BIBC matrix with Z matrix;
∂Ploss cos(i )
(BCBV = BIBCT Z). = 2[R]T [dPBIBCi ] · [re(I)]
∂Pi |Vi |
The voltage drop of a simple distribution system given in Fig. 1 sin(i )
is obtained as +[dPBIBCi ] · [im(I)] (17)
|Vi |
[V ] = [BCBV ] · [BCBV ] · [I]
where [dPBIBCi ] matrix is constructed by a simple algorithm given
⎡V ⎤ ⎡V ⎤
1 2 step by step as follows:
⎢ V1 ⎥ ⎢ V3 ⎥
⎢ V1 ⎥ − ⎢ V4 ⎥ • Step 1. Read BIBC matrix, i bus number for DG.
⎣ ⎦ ⎣ ⎦
V1 V5 • Step 2. Set dBIBCi matrix, dBIBCi = BIBC.
V1 V6 (13) • Step 3. Find the row with zero elements for the (i − 1)th column
⎡Z 0 0 0 0
⎤ ⎡
1 1 1 1 1 I2
⎤ ⎡ ⎤ of dBIBCi matrix; (zerorow = find(dBIBCi (:, i − 1) = 0)).
12
• Step 4. Convert all non zero elements of these rows to zero;
⎢ Z12 Z23 0 0 0 ⎥ ⎢ 0 1 1 1 1 ⎥ ⎢ I3 ⎥
= ⎢ Z12 Z23 Z34 0 0 ⎥ · ⎢ 0 0 1 1 0 ⎥ · ⎢ I4 ⎥ (dBIBCi (zerorow, :) = zeros(length(zerorow), n − 1)).
⎣ ⎦ ⎣ ⎦ ⎣ ⎦
Z12 Z23 Z34 Z45 0 0 0 0 1 0 I5
Z12 Z23 0 0 Z36 0 0 0 0 1 I6 To better explain [dPBIBCi ] matrix building algorithm, [dPBIBC4 ]
matrix which belongs to the sensitivity factor of the 4th bus,
∂Ploss/∂P4 , is given in (18) for the distribution system in Fig. 1:
2.2. The loss sensitivity factor ⎡1 1 1 1 1⎤
⎢0 1 1 1 1⎥
The derivation of the jth branch power loss per ith bus injected dBIBC4 = ⎢ 0 0 1 1 0 ⎥ (18)
real power ∂Plossj /∂Pi can be obtained from (10) as ⎣ ⎦
0 0 0 0 0
n 0 0 0 0 0
∂Plossj
P cos(k ) + Qk sin(k )
= 2Rj · BIBC(j, k − 1) · k ·
∂Pi |Vk | 2.3. Determination of optimal size
k=2
T. Gözel, M.H. Hocaoglu / Electric Power Systems Research 79 (2009) 912–918 915
size of DG, the derivative of the total power losses per each bus
injected real powers are equated to zero as
∂Ploss
=0 (19)
∂Pi
The expression of (16) can be shown in detail as
∂Ploss
nb
n
=2 Rj · dPBIBCi (j, k − 1) ·
∂Pi
j=1 k=2
k=
/ i
cos(i ) sin(i )
× · re(Ik ) + · im(Ik )
|Vi | |Vi |
nb
cos(i ) Pi cos(i ) + Qi sin(i )
+2 Rj ·dPBIBCi (j, i − 1)· ·
|Vi | |Vi |
j=1
sin(i ) Pi sin(i ) − Qi cos(i )
+ ·
|Vi | |Vi |
nb
n
=2 Rj · dPBIBCi (j, k − 1) ·
j=1 k=2
k=
/ i
cos(i ) sin(i )
× · re(Ik ) + · im(Ik )
|Vi | |Vi |
nb
+2 Rj dPBIBCi (j, i − 1) ·
j=1
Pi cos2 (i ) + Qi sin(i ) cos(i )+Pi sin2 (i )−Qi cos(i ) sin(i )
×
|Vi |2
(20)
∂Ploss
nb
=2 Rj · dPBIBCi (j, k − 1) ·
∂Pi
j=1 k=2
k=
/ i
cos(i ) sin(i )
× · re(Ik ) + · im(Ik )
|Vi | |Vi |
nb
Pi
+2 Rj · dPBIBCi (j, i − 1) · =0 (21)
|Vi |2
j=1
The real power injection at the bus i, Pi is extracted from (21) as Fig. 2. General flowchart of the proposed method.
nb n
|Vi | R dPBIBCi (j, k − 1)[cos(i ) · re(Ik ) + sin(i ) · im(Ik )]
j=1 j k=2
k=
/ i
Pi = − nb (22)
R
j=1 j
dPBIBCi (j, i − 1)
The minus sign in (22) indicates that Pi should be injected to the where two new terms, [re dIi ] and [im dIi ], are constructed by equat-
system. To facilitate a practical computation, (22) can be written in ing ith elements the real and imaginary part of the bus current
matrix format by omitting the minus sign as injection vector, [re(I)] and [im(I)] to zero. To illustrate the concept
[re dI4 ] and [im dI4 ] vectors that belong to the 4th bus real power
|Vi |[R]T [dPBIBCi ](cos(i )[re dIi ] + sin(i )[im dIi ])
Pi = (23) injection of the simple distribution system in Fig. 1, are provided in
[R]T dPBIBCi (:, i − 1) (24):
Downloaded from http://iranpaper.ir
http://www.itrans24.com/landing1.html
916 T. Gözel, M.H. Hocaoglu / Electric Power Systems Research 79 (2009) 912–918
Fig. 3. The optimal size for each buses DGs in 12 bus distribution test system.
The optimum size of added DG at bus i can be obtained by 3. The results of simulations and analysis
Pdgi = Pi + Ploadi (25)
In order to evaluate the proposed algorithm described in Section
2, three different test systems, taken from the literature, are used.
2.4. Determination of optimal size and placement for DG Accordingly, optimum size and place of DGs for the 12, 34 and 69
bus distribution test systems [17–19] are determined with the pro-
The objective is to minimize power losses, Ploss, in the system posed method, Acharya’s method [13] and the classical grid search
by injected power, Pdg. The main constraints are to restrain the algorithm. The classical grid search algorithm is employed with the
voltages along the radial system within 1 ± 0.05 pu. The proposed power flow program MATPOWER [20] to validate the results. The
method to determine the optimal size and placement of DG is given classical grid search algorithm is too costly because of computation
step by step as follows and also as a general flowchart in Fig. 2. time, that takes hours even days depending upon size of the system
and power steps. By using variable step size, successive load flows
• Step 1. Run the base case power flow. as also known sequential load flows could be employed instead of
• Step 2. Find the optimum size of added DG for each bus except the grid search algorithm. In this case, computation time will reduce
reference bus using (23) and (25). significantly.
• Step 3. Calculate total power losses from (5) for each bus by placing The grid search algorithm is applied by placing a DG to each bus,
optimum size of power to the bus. and changing the size of DG from 0% to 100% of total load power
• Step 4. Choose the bus which has the minimum power losses after with the step size of 0.01 MW. The total power losses are calculated
adding DG as optimum location. by a complete power flow analysis for each case. The case with
Fig. 4. The total power losses per each DG placement in 12 bus distribution test system.
Downloaded from http://iranpaper.ir
http://www.itrans24.com/landing1.html
T. Gözel, M.H. Hocaoglu / Electric Power Systems Research 79 (2009) 912–918 917
Fig. 5. The optimal size for each bus and total power losses for corresponding DG in the 34 bus distribution test system.
Fig. 6. The optimal size for each bus and total power losses for corresponding DG in the 69 bus distribution test system.
minimum losses determines the optimum location and size of DG. of DG is bus 9 where the total power losses are minimal for this
After that, the bus voltages are controlled whether the bus voltages particular test system.
are within the acceptable range. The optimum size of DGs and the total power losses for the cor-
In 12 bus distribution test system, the optimum sizes of DG responding DG sizes, determined by three approaches, are shown
placed on each bus are determined by three different methods, in Figs. 5 and 6, for 34 bus and 69 bus distribution test systems,
shown in Fig. 3. In addition, the calculated total power losses for respectively. It is seen that the difference between the results are
each bus, where the optimum sized DG is added, are shown in Fig. 4. negligible. The optimum places of DGs are bus 21 and bus 61 for 34
As clearly be seen, the optimum size of DG at each bus is different and 69 bus distribution test systems where the total power losses
and the difference between the results, that is obtained with three attain the minimum value. Since both analytical methods are based
methods, are not significant. Particularly, both analytical methods, on equivalent current injection technique, their results are almost
proposed in this paper and in [13], almost give the same results in the same range. Absolute differences between the results are less
for optimum sizes and estimated power losses. The optimum place than 0.5 kW for the optimum sizes. The estimated total power losses
Table 1
Comparison of algorithms results for test systems.
Test system Total load (MVA) Ploss without DG (MW) Method Optimum place (bus no) Optimum size (MW) Ploss (MW) CPU time (s)
918 T. Gözel, M.H. Hocaoglu / Electric Power Systems Research 79 (2009) 912–918
are again very close and the differences between them are in the [2] G. Pepermans, J. Driesen, D. Haeseldonckx, R. Belmans, W. D’haeseleer, Dis-
ranges of 0.1 kW at most. tributed generation: definition, benefits and issues, Energy Policy 33 (2005)
787–798.
In Table 1, for three different test systems, the total system load, [3] P.P. Barker, R.W. de Mello, Determining the impact of distributed generation
the results of the optimization procedure, computation time, the on power systems. Part 1. Radial distribution systems, in: IEEE PES Summer
total power losses with and without DG for all three methods, are Meeting, vol. 3, 2000, pp. 1645–1656.
[4] N. Hadjsaid, J.F. Canard, F. Dumas, Dispersed generation impact on distribution
tabulated. It is seen that the total power losses are significantly networks, IEEE Comput. Appl. Power 12 (April) (1999) 22–28.
reduced for all test systems. In addition, the optimum placement [5] S. Rau, Y.H. Wan, Optimum location of resources in distributed planning, IEEE
and size of DG determined by all methods are in close agreement. Trans. Power Syst. 9 (November) (1994) 2014–2020.
[6] K.H. Kim, Y.J. Lee, S.B. Rhee, S.K. Lee, S.-K. You, Dispersed generator placement
Again, due to fact that both formulations are based on the equivalent using fuzzy-GA in distribution systems, in: IEEE PES Summer Meeting, vol. 3,
current injection technique, the proposed and Acharya’s method July, 2002, pp. 1148–1153.
concur on the same optimum size and placement for all test sys- [7] J.O. Kim, S.W. Nam, S.K. Park, C. Singh, Dispersed generation planning using
improved Hereford Ranch algorithm, Electr. Power Syst. Res. 47 (October(1))
tems. On the other hand, although approximate power losses differ
(1998) 47–55.
from each other, the differences are in acceptable range. It is clear [8] K. Nara, Y. Hayashi, K. Ikeda, T. Ashizawa, Application of tabu search to optimal
that the worst method according to computation time is the grid placement of distributed generators, in: IEEE PES Winter Meeting, vol. 2, 2001,
search algorithm. There are significant differences for the compu- pp. 918–923.
[9] H.L. Willis, Analytical methods and rules of thumb for modelling DG-
tation time between other methods, too. The difference becomes distribution interaction, in: IEEE PES Summer Meeting, vol. 3, Seattle, WA, July,
more significant when the system size increased. For the small- 2000, pp. 1643–1644.
est test system the proposed method is 1.5 times faster than the [10] T. Griffin, K. Tomsovic, D. Secrest, A. Law, Placement of dispersed generation
systems for reduced losses, in: 33rd Annual Hawaii International Conference
Acharya’s method. The difference is mainly due to the fact that on Systems Sciences, Maui, HI, 2000, pp. 1–9.
Acharya’s method needs to invert large matrices and also requires [11] C. Wang, M.H. Nehrir, Analytical approaches for optimal placement of DG
a second power flow. sources in power systems, IEEE Trans. Power Syst. 19 (November(4)) (2004)
2068–2076.
[12] T. Gözel, M.H. Hocaoglu, U. Eminoglu, A. Balikci, Optimal placement and sizing
4. Conclusion of distributed generation on radial feeder with different static load models, in:
Future Power Systems, 2005 International Conference, Amsterdam, November
16–18, 2005.
This study presents and evaluates an analytical method which [13] N. Acharya, P. Mahat, N. Mithulananthan, An analytical approach for DG allo-
can be used to determine the optimal placement and sizing of DG cation in primary distribution network, Int. J. Electr. Power Energy Syst. 28
without use of admittance, impedance or Jacobian matrix with only (December(10)) (2006) 669–678.
[14] J.-H. Teng, C.-Y. Chang, A network-topology-based capacitor control algo-
one power flow for radial systems. The method is easy to be imple- rithm for distribution systems, Transm. Distrib. Conf. 2 (2002) 1158–
mented and faster for given accuracy. The derived sensitivity factor 1163.
is more suitable for distribution systems and could be utilized by [15] H.N. Ng, M.M.A. Salama, A.Y. Chikhani, Classification of capacitor allocation
techniques, IEEE Trans. Power Deliv. 15 (January) (2000) 387–392.
means of simple matrix algebra.
[16] J.-H. Teng, A network-topology-based three-phase load flow for distribution
The optimal size and location of the DG, which is determined systems, Proc Natl. Sci. Counc. ROC(A) 24 (4) (2000) 259–264.
by the method, is also evaluated against Acharya’s method and the [17] D. Das, H.S. Nagi, D.P. Kothari, Novel method for solving radial distribu-
classical grid search algorithm. It is found that the proposed method tion networks, IEE Proc. -Gener. Transm. Dist. 141 (July(4)) (1994) 291–
298.
is in close agreement with Acharya’s method and the grid search [18] M. Chis, M.M.A. Salama, S. Jayaram, Capacitor placement in distribution systems
algorithm. It is appeared that the proposed method is faster than using heuristic search strategies, IEE Proc. -Gener. Transm. Distrib. 144 (May(3))
other methods in the computation time and it is appropriate for the (1997) 225–230.
[19] M. Baran, F.F. Wu, Optimal capacitor placement on radial distributions systems,
distribution systems. IEEE Trans. Power Deliv. 4 (January(1)) (1989) 725–734.
[20] R.D. Zimmerman, C.E. Murillo-Sánchez, D. Gan, MATPOWER A MATLAB Power
References System Simulation Package, www.pserc.cornell.edu/matpower/matpower.
html (access date: April 20, 2007).