Вы находитесь на странице: 1из 6

MULTI-OBJECTIVE FEEDER RECONFIGURATION BY DISTRIBUTION MANAGEMENT SYSTEM

I. Roytelman, Senior Member, IEEE V. Melnik, S.S.H. Lee, Senior Member, IEEE R. L. Lugtu, Member IEEE

Siemens Energy & Automation, Inc.


Empros Power Systems Control
Brooklyn Park, IvlN 55428

Abstract: Feeder reconfiguration for use by distribution man- ing switches is not as many as that for urban networks, but the num-
agement systems is discussed in this paper. Multiple objectives are ber has increased continuously over the past many years for the
proposed to reflect realistic operating environments while achiev- purpose of improving reliability.
ing all benefits from feeder reconfiguration. The multiple objectives The subject of distribution feeder reconfiguration has been investi-
considered are minimization of power losses, load balancing among gated by numerous researchers over the past several years. In the
supply transformers, minimization of the worst voltage drop, mini- early studies [1,2] involving urban distribution systems, a desirable
mization of service interruption frequency, and balanced service of radial distribution system solution is determined from a mesh net-
important customers for enhanced service reliability. The objective work in which all normally-open switches are simulated to be
function containing five different objectives are optimized subject closed. Based on a power flow solution to this distribution system
to capacity and protection device constraints. The overall solution of mesh network structure, switches to be opened are then identi-
approach is a two-stage process. In the first stage, a suboptimal fied. This idea based on a closed-loop power flow solution has been
solution is found by analyzing the mesh distribution system in extended and used in later work [6]. Such solution procedures find a
which all open switches are simulated to be closed. Applying spe- final feeder configuration which is independent of the initial status
cial power flow analyses to this mesh network, a radial distribution of switches.
system is determined as an intermediate solution. In the second
Simple and efficient branch exchange algorithms were proposed in
stage, this solution is continuously improved by the branch
[3,4]. In [4] a simple formula is presented to estimate the change in
exchange scheme. Special topology models are also developed to
feeder losses resulting from feeder reconfiguration. This formula
accelerate the search procedure. Use of the algorithm is illustrated
can be used as a criterion to determine how to improve intermediate
by numerical examples.
solutions. [5] shows a heuristic branch exchange scheme designed
Introduction for handling urban distribution networks.
Traditionally distribution planning requires feeder reconfiguration More papers on optimal feeder reconfiguration were published in
on a seasonal or annual basis. Real-time feeder reconfiguration, recent years [7-141. Some of them extended early ideas, or tried to
however, has the potential to provide substantial benefits since the find a compromise between fundamental approaches. For instance
magnitude of feeder load varies continuously. 'I'his potential can be in [8], a single loop is optimized in every step of the algorithm
attained by the distribution management system (DMS) where within its solution process. Expert systems and artificial intelli-
switches can be opened or closed on a real-time basis. gence approaches along with heuristic algorithms were employed
The problem of optimal feeder reconfiguration occurs on both in [9-141 for the feeder reconfiguration problem. Further develop-
urban and rural distribution systems. Initially the reconfiguration ments are needed in order for these algorithms to be practical for
problem concerned urban distribution systems whose infrastructure implementation.
is mostly underground cable systems designed for closed loop net- An interesting trend can be: observed from the above-mentioned
works but operated as radial feeders. In such systems each distribu- papers as to the objective functions and constraints used in their
tion transformer can be served from more than two circuits and the problem formulations. While the papers [1,2,4] used power loss
open/closed status of sectionalizing switches located at the trans- reduction as the only objective function, the goal of load balancing
former site determines how the transformer is served. The problem was noted in [3,5] and [9]. Either power loss reduction or load bal-
of feeder reconfiguration can be regarded as that of finding ancing was dealt with in [7].As far as constraints are concemed,
branches to be opened since each branch could be opened by two voltage drop constraints are accounted for in [13] and protective
switches located at both ends. Often the number of possible combi- device coordination is modelled in [lo].
nations of selecting open branches becomes very large. The Aim of the Paper
The feeder reconfiguration problems is also concemed with rural As noted above, solution approaches proposed in all previous
distribution systems with overhead networks. Such systems were
papers were designed to deal with feeder reconfiguration from the
initially designed as radial and equipped with very limited number standpoint of mostly feeder loss reduction and load balancing.
of tie and sectionalizing switches. Sectionalining switches can be However, individual utilities may pursue different goals because of
used to isolate faulted sections while tie switches can be used to differences in the characteristics of their distribution systems. For
provide alternative service paths. The number cif tie and sectionaliz- example, feeders in rural distribution systems are often reconfig-
ured for the purpose of enhancing power supply reliability. This is
particularly true for overhead networks which cover wide geo-
graphic areas and historically have higher failure rates than cable
systems. Because of longer feeder circuits, low voltage problems
could occur on rural distribution systems. The issue of power losses
on rural distribution systems may not be as critical as that on urban
distribution systems since typically rural circuits are not heavily
loaded.
$4.00 0 1995 IEEE
0-7803-2663-6195 517
Furthermore, a utility may wish to accomplish several goals simul- Stage 1 is intended to yield a reasonably good starting solution for
taneously or different goals seasonally. Therefore, in order for the Stage 2. Such a concept is important especially if the number of
DMS to be useful and practical, its solution procedure for feeder possible open switch combinations is very large. In the event that
reconiiguration should be flexible enough to accommodate various only a limited number of switchings is desired, this stage may be
circumstances. To handle the most general case, an attempt is made bypassed and the Stage 2 procedure will proceed with the existing
in this paper to solve the feeder reconfiguration problem having configuration as a starting solution. Furthermore, only those switch-
multiple objectives, in which practical constraints are also ings which yield improvements greater than a user-specified magni-
accounted for. In this regard, the following five objectives are mod- tude may be used in Stage 2 to limit the number of switchings.
eled within the problem formulation: Approach in Stage 2
- Minimization of feeder losses Using the solution obtained in Stage 1 as a starting solution, the
- Load balancing among supply transformers procedure iteratively iinds a better solution until stopping criteria
- Minimization of the worst voltage drop are met. At every intermediate iteration, the change in the objective
- Minimization of service interruption frequency function (containing all five objectives in this stage) is examined
- Balanced service of important customem when a switch adjacent to the position of a currently-open switch is
It is to be noted that additional objectives can also be included with- simulated to be opened while the currently-open switch is assumed
out altering the basic solution approach proposed here. The prob- to be closed. Such a method can be regarded as the branch
lem of feeder reconfiguration can be looked upon as an exchange scheme where small improvement is sought in every step.
optimization problem where the objective function reflects the In fact, because of the finite number of switches installed on the
above objectives. Strictly speaking, the attributes of the above five system, the procedure improves the current solution rather quickly.
objectives are not the same. Consequently it can be argued that The Objective Function and Constraints
simultaneous consideration of all five objectives cannot be possibly
done. However, it is inevitable that the objective function should The mathematical representation of the respective objectives fol-
contain all five objectives to handle the most general situation. In lows.
this paper it is proposed that the value of each of the five objectives (a) Minimization of Feeder Losses
be per-unitized and weighted by user-specified weighting factors. Feeder losses P, can be written as:
Our investigation of several numerical examples reveals that such
an approach is reasonable and doable. It should be noted that any
other approaches, including goal programming, will encounter the
same difficulty since human judgment must be involved in one way
or another in dealing with multiple objectives. where: Ii, Ri: Current and resistance in circuit branch i
The Overall Solution Approach Pj,APj: Load and power losses of distribution trans-
former j, respectively
The overall solution approach is a two-stage process as described nl, nt: The number of circuit branches and distribution
below. transformers, respectively
Stage 1: A suboptimal solution is first determined by a fast non- Typically only line losses may be taken into account. However,
iterative procedure. transformer losses can also be included to reflect load-voltage char-
Stage2: The suboptimal solution from Stage 1 is improved acteristics. Note that the denominator in Equation 1 is needed to
iteratively until no further improvement can be obtained. yield pa-unit values.
The concept behind the solution approach used in each stage is (b) Load Balancing among Supply Transformers
briefly discussed below while details are provided in later sections.
Unbalance in supply transformer loadings can be written as:
Approach in Stage 1
In this stage, all normally-open switches in the distribution system
B = l C Bs
(except for those that should not be closed for any reasons) are ss = 1
assumed to be closed initially. Therefore, the resulting distribution
system is of mesh network structure, and consequently the existing where: Bs = /P\Tp-NaVi
configuration is not given a favor. In this case, the problem is to
identify switches to be opened in order to yield a radial distribution hT s = 1sSIl/srat:
, I
s Load on supply transformer s
system. A solution is then found to optimize the most critical objec- in per unit
tive among five objectives. T h e use of a single objective can be jus- Nav = ()3/Ss/) /( zSLat) Average loading of all supply
:

tified since !he result will be used as a mere starting solution that . transformers in per unit
rat
will be improved in Stage 2. SS’SS : Real power load and rating of
supply transformer s
A “power-flow type” solution approach has been developed to find
n : Supply transformer number
a suboptimal solution for each of the five objectives. The selection
of the most critical objective will determine which method is to be Note that per-unit values are based on transformer MVA ratings.
used in Stage 1. Based on the flow pattem resulting from a single Although for convenience in explanation only transformer load bal-
power-flow solution of the mesh network, switches to be opened ancing is considered here, feeder load balancing can similarly be
a e determined by identifying the switches carrying the least flow. accounted for.
The idea behind this strategy is based on the observation that the (c) Minimization of the Worst Voltage Drop
pattem of flow on the mesh network will optimize each objective Suppose that the worst voltage drop occurs at node j. The amount of
and that any radial configuration which least disturbs such flow pat- voltage drop in per unit at node j is given by:
tern should be a good solution.
518
2. The trip setting of each protective device should be greater
than maximum load current but less than minimum fault
(3) current through the device. Such constraints can be written as:
load tr fault
where: Ii: Current in circuit branch i Is <Is <Is
Zi: Impedance of circuit branch i Any constraint violations can be handled by including them as pen-
The summation is performed over the path leading to node j from alty terms in the objective function during the implementation of
the supply transformer. Again per-unit values are obtained by divid- Stage 2. In Stage 1 branch overload violations are handled differ-
ing the circuit rated voltage. ently. Each overloaded branch may be opened and the line flow is
simulated by current injections at the both open ends with proper +/
(d) Minimization of Service Interruption Frequency
- signs to reflect the direction of flow. In this case, the magnitude of
Service interruption frequency index is defintrd as: injection is limited to the branch rating.
f"t \ f " t \ Note that feeder voltages are not included in the set of constraints,
(4) but they are reflected in the objective function in this paper. Of
course, treatment of voltage in other way will not alter the proposed
where: Pj: Load of distribution transformer j solution approach.
hj: Service interruption frequency at transformer j Solution Method: Stage 1
which is calculated to be the sum of failure frequen-
As noted above, in Stage 11 a suboptimal radial distribution system
cies of all branches in the series path from the sup-
solution is found from analysis of the mesh network. Note that it is
ply transformer to distribution transformer j
extremely difficult to find a solution while considering all five
The index (4) is a slightly modified version of the overall failure objectives. Therefore, only a single term in Equation 6 is optimized,
index for radial systems with perfect switching [15],where the load where the most critical objective may be selected. The justification
magnitude is used instead of the number of ciistomers at each node. of this simplification was mentioned earlier. The actual solution
Selection of a specific index is utility-dependent and would not algorithms will depend on which term in Equation 6 (i.e. which
alter the basic formulation presented in this paper. The above sim- objective) is to be used. Note that regardless of the method to be
ple form of reliability is selected just for illustration purposes. used, the starting point is always the mesh network. The difference
(e) Balanced Service of Important Customlers lies in how bus and line data are specified in each case. Solution
methods for five individual objectives are described below.
The objective is to not serve the majority O F important customers
from one or a limited number of supply transformers. Such strategy (a) Minimization of Feeder Losses
is desirable since failure of a substation transformer serving the Consider a mesh network in which the impedance of every circuit
majority of important customers could cause service interruption to branch is represented by its resistance portion only while its induc-
many important customers. The following index may be used: tive or capacitive portion is ignored. Then find a power flow solu-
tion to this resistive network where loads are represented as
constant current injections. It should be noted that the resulting pat-
(5) tem of power flow is determined such that power losses become
minimal. (See [6] for proof) Therefore, a suboptimal radial distribu-
where:

L,:
I
Es = L - L
s
Number of important customers served from sup-
tion feeder solution can be found by opening switches such that the
pattern of power flow is least disturbed. This can be accomplished
by successively identifying a switch that carries the least current
ply transformer s and opening it. In doing so, care should be exercised in order not to
La": Average number of important customers per sup- open certain switches so that no loads may be disconnected from
ply transformer the network. The following algorithm is designed to carry out such
The overall objective function to be optimized may be any combi- a task.
nation of the above five Equations (1) through (5) which can be Define the rank of a node as the number of branches connected to
written as: the node. Successively remove branches connected to nodes of rank
F = CL PL + C B B + C D D + C W W + C E E (6) one until there exist no nodes of rank one. In this case, the resulting
network will have no radial circuits. Then find a switch which car-
where user-specified weighting factors are used in Equation 6.
Many utilities have indeed tried to capture several objectives simul- ries the least current and open it. Repeat the above process until the
total number of open switches is equal to that on the initially exist-
taneously, but found no good way of doing it. We found that once
ing system.
individual objectives are per-unitized, the order of magnitude asso-
ciated with weighting factors can be fairly even. (b) Load Balancing among Supply Transformers
Two types of constraints are considered in this; paper although other To determine the pattem of power flow which balances loading on
constraints could also be taken into account withii the proposed supply transformers, find a power flow solution to the mesh net-
solution procedure: work in which equal amount of real power injection is specified at
1 . Each element (line or supply transformer) should be loaded
each supply transformer. Unlike Case (a), actual impedances o f cir-
cuit branches are used in line data. The suboptimal radial distribu-
within its capability. This can be written as:
tion system solution can be found by successively opening a switch
Ii < K i I r a x (7) which carries the least current. The method of such implementation
max
is the same as that for Case (a).
where Ii is current in element i, Ii is the rating of element i,
and Ki is used to reflect conditions assocl ated with element i.
519
(e) Minimization of the Worst Voltage Drop switch is closed, will yield the greatest decrease in the objective
Note that the pattern of power flow in a mesh network is generally function among all candidate switches. The best altemative switch
determined such that the worst voltage drop at a load node is mini- is then opened, and the adjacent currently-open switch is closed
mized. Reactive power controlling devices will, of course, consid- instead. This process of improving intermediate solutions continues
erably affect the actual feeder voltage proEle. To minimize the until no further improvement is possible. In general, there are two
effectsof reactive power support, their control settings are assumed altemative directions for each open switch. Since a strict hprove-
to be fixed. Then find a power flow solution to the mesh network ment is obtained in each iteration until a stopping criterion is met,
where power injections into the supply transformers are specified to the same switching combination will not be repeated and hence,
be those from the base case. The bus and line data are also those of there is no possibility of cycling.
the base case. The suboptimal radial distribution system solution The bulk of computational requirements in each iteration is associ-
can be found by successively opening a switch carrying the least ated with the calculation of the change in the objective function F in
current, as done in Cases (a) and (b). Equation 6 when simulating a move of an open-switch position to
(d) ~ i n i m i z a ~ i of
o nService Interruption Frequency adjacent switch locations. In other words, efficient computational
schemes are needed to determine the difference in the values of
Consider a resistive electrical circuit in which the failure rate of a
Equations 1 - 5 before and after the simulation of moving open-
branch is taken for branch resistance and real power load is taken
switch positions. Figure 1 is used to illustrate how to calculate
for dc load current. Note that the index in Equation 4 is equivalent
those values. Branch n-k is initially open, and node k is initially
to the sum of total voltage drops at all load points in the equivalent
served from Substation S(k). An altemative solution to be examined
radial distribution system having impedances of the described prop-
is to open branch k-m instead and close branch n-k, thereby serving
erty. Also note the similarity between the numerator in Equation 3
node k from Substation S(n). The change in the value of each term
and that in Equation 4.Therefore, to find a solution which mini-
in the objective function F can be computed as follows, where a
mizes Equation 4, again consider a mesh network whose branch
decrease in value is considered to be positive.
resistances are replaced by branch failure rates and find a current
flow solution. Then similar to Cases (a) to (c). open switches one by
one such that the flow pattem is least disturbed. Noting that a mesh
network offers higher reliability than individual radial feeders, such
procedure will yield a suboptimal solution which minimizes Equa-
tion 4. Note that in this case, we are dealing with a non-electrical A
network which mimics electrical circuits. Figure 1. Simulation of Opening Branch k-m
(e) Balanced Service of Important Customers and Closing Initially-Open Branch n-k
Strictly speaking, this can be looked upon as an assignment prob-
(a) Minimization of Feeder Losses
lem in which important customers are assigned to supply t r m -
formers. If this objective alone is used as an optimality criterion, Instead of using power flow solutions, we chose to use the simple
non-unique solutions will likely to result. In practice, this objective formula presented in [4],which can be written as:
may be used along with other objectives, and therefore, such prob-
lems can be avoided. To balance serving important customers from
several supply transformers, equal number of important customers
are assigned to each supply transformer and the number of custom-
where: Jk: The total complex load current served
ers is regarded as an injection. At load buses, the number of impor-
from node k
tant customers is taken for load current. The issue of branch
AVk,AVn: Voltage drop at node k and n, respec-
impedance is not important here, but failure rates used in Case (d)
tively, in a resistive network
can also be used to achieve higher service reliability. The solution
Rloop: Resistance of the loop containing branch
procedure is to find a current flow solution to this mimic electrical
n-k
mesh network, and open switches one by one such that the flow pat-
superscript c: Complex conjugate operator
tern is least disturbed. The procedure for finding a suboptimal radial
distribution system solution is then virtually the same as those for (b) Load Balancing among Supply Transformers
the above four cases. Applying Equation 2 to Figure 1 yields:
To sum up, a suboptimal radial distribution system solution can be
found by examining the pattern of flow in a mesh network. The
impedance and load characteristics of the mesh network depends
upon which objective is to be optimized in Stage 1. The solution where: B,e), Bs(nl: Unbalancing at supply transformers s(k)
obtained is then used as a starting solution for further improvement and s(n), respectively
in Stage 2. JSo(),Js(ny Total load current served from supply
Solution Method: Stage 2 transformers s(k) and s(n), respectively
In Stage 2, the starting solution passed from Stage 1 is successively V,O;),Vs(n): Voltage at supply transformers s(k) and
s(n), respectively
improved until stopping criteria are met. To do that, the change in
the objective function (Equation 6) is evaluated by simulating a ( e ) Minimization of the Worst Voltage Drop
move of an open-switch position. Such examination is made for It is easy to see that the move of the open-switch position as indi-
each of candidate switches adjacent to all currently-open switch cated in Figure 1 will increase voltage drop at any point within the
positions. In each iteration, the procedure identifies the best altema- section between S(n) and node n, and that it will decrease voltage
tive switch which, if it is opened and the adjacent currently-open drop over the section from s(k) to node m. In the event that the

520
worst voltage drop occurs in the circuit section served from node k, Step 4 To &-energize branch 7, first note that BEG(7) is 2. Then
the change in the objective value is given by: change DOWN(2) from 7 to 0.
The simplicity in changing topology is facilitated since information
AD = (Dk-Dn-/(Zs-n+Zn-k)Jk/)/Vrat
(11)
on the oriented network is stored through referring to node and
where: Zs-,: Series impedance from s(n) to node n branch tables rather than the traditional way of branch to branch
G-k: Impedance of branch 11-k reference as done in [15].
AVk, AV,: Voltage drop at nodes k and n, respectively, Example
prior to reconfiguation
The example system shown in Figure 3 reflects an urban distribu-
(d) Minimization of Service Interruption lrrequency
tion system. The system consists of 61 loads, 64 branches, 3 supply
The change in the value of the index is given. by: transformers, and 7 normally-open branches. Branches 11, 19. 25,
33,44,46, and 50 are initially open. The network is a mix of cables
Aw = ((hk-hn-hn-k)Pk)/
[c ) ps
(12)
where hk and k,, are service interruption frequency at nodes k and
(branches 1 - 28) and overhead lines (branches 29 - 64) with the R/
X ratio ranging from 0.3 to 2. The total load is 82.77 h4W + j46.70
MVAR. Forty-six important customers are served from this system.
n, respectively, and k,,.k is the failure rate of branch n-k. Branch failure rates are O.Ol/mile/year for all cables (branches 1 -
(e) Balanced Service of Important Customers 28). 0.02/mile/year for new overhead lines (branches 29 - 53) and
O.M/mile/year for old overhead lines (branches 54 - 64).
The amount of change for this index can be written as:
h Branch Table
i BEGEND
1 9 1
where Es(k) and Escn., are unbalancing at supply transformers s(k) 2 1 5
and s(n), respectively. 3 9 3
Tracing Tool 4 3 6
5 5 7
A feeder reconfiguration algorithm of production grade requires an
efficient tracing tool to identify upstream and/or downstream nodes
and to reflect changes in feeder configuration on the topology
model. The tool developed is explained using Figure 2, in which
branches and nodes are identified by indices i and j, respectively.
Each branch is identified by its two end nodes, as shown in BEG
and END arrays in Figure 2.b. Branches connected to a node are Node Table b)
referred to as upstream and downstream brcnches depending on the j up DOWN NEXT Open-BranchTable
direction of power flow from the branches into the node, as shown
in UP and DOWN arrays of Figure 2.c. Tci account for situations
where more than one downstream branch Ere connected to a node, 3 3
the additional array NEXT is used. If BEG and END nodes of a 8
branch defined in the branch table do not conform to the direction
of power flow, a negative sign precedes the branch number in the 7 5 0
node table in Figure 2.c. Open branches are listed in an array,
OPBR, as shown in Figure 2.d.
The topology model defined in Figure 2 can be used to efficiently
search the supply node for any specified downstream node. This 3
can be accomplished by repeating the procc:ss of reading UP array sten 3
in Fig. 2.c, switching to BEG array in Fig. 2.b and then going back [ step4
to UP array in Fig. 2.c. Similarly all downstream nodes can be iden- c> d)
tified by using the branch and node tables. Figure 2. Topology Model and its Modification to
Necessary modifications to be made to the contents of the node and Simulate Move of the Open-Branch Position from
open-branch tables in order to reflect a change in feeder configura- Branch 1-5 to Branch 5-2
tion are illustrated using an example where branch 7 is simulated to
be opened and branch 2 is closed instead. 'Io initiate the process, it Study results are summarized in Table 1 and Table 2. Shown in
Table I are the values of five individual objectives (shown in five
is observed that the branch to be closed is OPBR(1)=2 and the com-
mon node between the branches to be closed and opened is rows) for the initial system (column 3), those at the end of Stage 1
END(2)=5. when respective individual objectives are optimized (column 4).
those at the end of Stage 2, when the same individual objectives are
The following four steps determine necessary modifications to be optimized (column 5). the moves number of open-branch positions
made to the tables: during Stage 2 (column 6), open-branch positions for the Stage 1
Step 1: To close branch 2, first note that BEG(2) is 1. Then (column 7), and open-branch positions for the Stage 2 (column 8).
change DOWN( 1) from 0 to 2. It is seen that optimum solutions have been found upon completion
Step 2: To change the supply branch 601 node 5, change UP(5) of Stage 1 when the objective function F was optimized to achieve
from 7 to 2. either the objective of minimization of feeder losses or that of mini-
Step 3: To open branch 7,change OPBR( 1) from 2 to 7.
52 1
while ignoring other objectives. It is observed that the optimum of a
single objective is achieved at the expense of other objectives. The
29 bottom row in the table shows the results of optimizing all five
30 objectives simultaneously. The weighting factors used in this case
31 are CL=0.4, CB3.2, CDd.2, Cw=O.l, and CE=o.l. It is seen that
such an optimum solution can achieve all five objectives in a coor-
dinated manner.
Conclusions
The problem of feeder reconfiguration is revisited. Instead of
restricting the scope of the problem so that a specific solution
scheme may be applicable, in this paper an attempt is made to deal
Figure 3. Example Distribution System Served from with realistic situations. Since feeder reconfiguration can offer vari-
Three Supply Transformers ous benefits, it is reasonable to try to obtain all such benefits. It is
recognized that not all such benefits are conforming since their
mization of the worst voltage drop. Changes in open-branch posi-
attributes are intrinsically different. In this paper this problem of
tions from the initial system configuration to the stage-I suboptimal
multi-objective nature is handled by applying weighting factors to
solution are seen to be significant. However, in transition from
multiple objectives while the values of individual objectives are
Stage 1 to Stage 2, rather minor changes are noted as far as open-
per-unitized. Our investigation of test systems reveals that such an
branch positions are concemed. It is observed that optimum config-
approach is proper and the results look promising.
urations differ significantly depending upon which objective is
sought, This suggests that to obtain the overall best configuration Practically, the entire problem cannot be solved within a single
from the standpoint of multiple objectives, those objectives must be solution framework. Therefore, the two-stage solution approach is
considered simultaneously. Such a conclusion can be drawn from followed. In the first stage, a suboptimal solution is found by exam-
Table 2 in which several optimum solutions obtained at the end of ining the mesh network which is obtained by closing all normally-
Stage 2 are tabulated in a different format. open switches. The obtained suboptimal solution is used as a start-
Table 1. Stage 1 a n d Stage 2 Solutions for Each Obiective ing solution in the second stage where improved solutions are itera-
tively found. A tracing tool which is needed for efficient
implementation of the proposed procedure is also discussed in the
paper.
References
1. A. Merlin and H. Back, “Search for a Minimal-Loss Operating Spanning
Tree Configuration in Urban Power Distribution Systems”, Roc. of 5th
Power Systems Com. Con., Cambridge, U. K., Sept. 1 - 5,1975.
2. V.G. Holmsky “Calculation and Optimization in Power Systems”, Hight
School Press, Moscow, 1975,280 p.
3. C.H. Castro, J.B. Bunch, and T.M. Topka, “Generalized Algorithms for
Distribution Feeder Deployment and Sectionalizing”, IEEE Trans. on
Power Apparatus and Systems, PAS 99, No. 2, 1980, pp. 549 - 557.
4. S . Civanlar, J.J. Grainger, H. Yin, and S.S.H. Lee, “Distribution Feeder
Reconfiguration for Loss Reduction”, IEEE Trans. on Power Delivery, July
1988, pp. 1217 - 1223.
5. K Aoki, H. Kuwabara, T. Satoh, and M. Kanezashi “An Efficient
Algorithm for Load Balancing of Transformers and Feeders by Swith

i1 Objective
1 I 1 1
Table 2. Comparison with Multi-Objective Solution
Optimization P;g Load
Balance
V;ltt~ Service
Interruption
1 Important

Balance
1 6.
Operation in Large Scale Distribution Systems”, IEEE Trans. on Power
Delivery, October 1988, pp. 1865 - 1872.
D. Shirmohmadi, H.W. Hong, “Reconfiguration of Electric Distribution
Networks for Resistive Line. Losses Reduction”, IEEE Trans. on Power
Delivery, April 1989, pp. 1492 - 1498.
7. M. Baran, F.F. Wu, “Network Reconfiguration in Distribution Systems for

1 1 1 1 1 :I. 1 1
1 0 (Initialcondition I 5.82 1 12.73 1 14.45 1 6.42 1 21.74 I Loss Reduction and Load Balancing”, IEEE Trans. on Power Delivery,
April 1989, pp.1401 - 1407.
1 Powerloss 5.42 6.97 12.75 7.48 17.39 8. S.K. Goswami, S.K. Basu, “A new Algorithm for the Reconfiguration of
Distribution Feeders for Loss Minimization”, IEEE Trans. on Power
Delivery, July 1992, pp. 1484 - 1491.
3 IVoltagedrop 6.27 21.07 8.17 39.13 9. Y.Y. Hsu, Y. JweHwu, S.S. Liu, Y.W. Chen, H.C. Feng, Y.M. Lee,
“Transfomer and Feeder Load Balancing Using a heuristic Approach”, 92
4 Serviceintenup- 13.65 52.10 22.37 4.09 65.22 WM 167-7 PWRS.
tion 10. Y.Y. Hsu, Y. Jwo-Hw, “Planning of Distribution Feeder Reconfiguration
5 Important cus- 7.78 25.12 25.17 0.00 with Protective Device Coordination”, 92 SM 501 - 7 PWRD.
torner balance 11. C.S. Chen, M.Y. Cho “Energy Loss Reduction by Critical Switches”, 92
SM 502 - 5 PWRD.
M Multi-objective 5.44 6.39 12.75 7.03 17.39 12. T. Taylor, D. Lubkeman, “Implementation of Heuristic Search Strategies
for Distribution Feeder Reconfiguration”, IEEE Trans. on Power Delivery,
Vol. 5, 1990, pp. 239 - 246.
13. K. Nara, A. Shiose, M. Kitgawa, T. Ishihara, “Implementation of Genetic
Algorithm for Distribution Systems Loss Minimum Re-Configuration”, 91
SM 467 - 1, P W .
14. K.H. Jung, H. Kim, Y. KO, “Network Reconfiguration Algorithm for
Automated Distribution Systems Based on Artificial Intelligence
Approach”, IEEE Trans. on Power Delivery, Vol. 8, No. 4, 1993, pp. 1993 -
1941.
15. J. Endrenyi, “Reliability Modeling in Electric Power Systems”, John
WileyWons Ltd., 1978,338 p.

522

Вам также может понравиться