Professional Documents
Culture Documents
Glamocanin Optimal Losses For Distribution Network 1990
Glamocanin Optimal Losses For Distribution Network 1990
3, AUGUST 1990
Abstract - A new algorithm for network recon- weakly meshed network to open the switch with
figuration of power distribution systems is lowest current for the optimal flow pattern.
presented. An optimal loss reduction is Planning of the feeders for a new substation
accomplished to maintain acceptable voltage and the relocation of s o m e loads /?/ is based
at customer loads as well a s to assure suffi- also on some heuristic rules for minimum-loss
cient conductor and substation current capac- design .
ity to handle load requirements. The suc-
cess of the algorithm depends directly upon The generalized approach to determine
the straightforward and highly-efficient the optimal network reconf iguration i 5 based
solution of quadratic cost transshipment on branch exchange procedure o r by moving
problem. The new'algorithm, described in sectionalizing positions. To evaluate each
this paper, completely eliminates the need loss-reduction move radial power flow method
for matrix operations and executes all opera- has to be carried out for feeder pair. The
tions directly on graph of the distribution existing algorithms, however,do not guarantee
system. global optimum of power losses. Namely,
open switch position can not occasionally be
moved due to:
Keywords: distribution automation ,distribu-
tion system planning,power loss reduction, - increase of power losses
power flow method,transshipment problem.
- voltage-drop constraints
J2w
t h e proposed programming technique.
t h e f i r s t step NaG)
Power l o s s e s as a f u n c t i o n o f t h e c u r - CM - minimal total transportation
rent J and conductor ( t r a n s f o r m e r ) s i z e s , c o s t o f s e c t i o n s , s t a r t i n g from
a r e shown on Fig.1: source node G t o s i n k node M.
current
- t o t a l t r a n s p o r t a t i o n c o s t i>f
f r o m t h e s o u r c e node t o any
the
unit
sink
aPLK =
c
Where:
APLK(M+)
APLK(~)
i f APL~.;(M*) < A P L K ( ~ )
if APLK(M*) 2. APLK(,)
(6)
For a g i v e n b a s i c f e a s i b l e s o l u t i o n l e &
CONCLUSIONS x be p a r t i t i o n e d t o b a s i c variables x
( m x l ) and nonbasic v a r i a b l e s x" ( (n-m)xl )
: x"=0 . Let matrix B ( mxm ) , det(B)fO
An a l g o r i t h m f o r o p t i m a l network recon- and N ( mu(n-m) ) be o b t a i n e d by a r r a n g i n g
f i g u r a t i o n o f d i s t r i b u t i o n systems i s pre- columns o f A t o match t h e elements o f
sented. The s i m p l i c i t y of t h e proposed ieeand xn :
methodology makes i t s u i t a b l e f o r use on
p e r s o n a l computer p r o v i d i n g t h e o p e r a t i o n and A = B N . (a4)
p l a n n i n g e n g i n e e r s w i t h u s e f u l and e f f i c i e n t
tool. The main a l g o r i t h m advantages are: By r e a r r a n g i n g m a t r i x c i n terms o f
b a s i c and nonbasic v a r i a b l e s , t h e above prob-
1 . Conceptually, i t i s s t r a i g h t f o r w a r d lem can be p u t i n t o t h e f o l l o w i n g form:
procedure o f " c r e a t i n g " t h e o p t i m a l
s o l u t i on. m i n i m i ze tpe objective function:
4. No s t a r t i n g s o l u t i o n i s needed. x b r O , (a-?)
x n = O . (a8) of 1 inear problem. This assumption suggests
that the fol 1 owing requirement must be met:
The basic variables xb can be found
from (ab) :
xb = B-l b - B-l N xn (a9) From (a3) follows that:
Y W M
(a13)
Fig.A Nonbasic branches M-N and N-M
I
-1 the flow of basic
branch j is oriented
opposite to loop i ,
Yj,i= 1 the flow of basic
branch j is oriented as Eq. (a221 shows that for an optimal
loop i is, solution the total transportation cost of the
(3 basic branch j and loop branches (for unit flow) from the source node
i are not incident, to sink node N is minimal if the path is
Y' - incident matrix ( nx(n-m) ) of formed of the basic branches of set TN.
all branches (basic and nonbasic) The same conclusion could be stated for sink
and loops, node M , as well as for any sink node.
I - unit matrix ( (n-m)x(n-m) ) .
The following optimality condition is a
The basic variables (flows in the basic direct consequence of the above property:
branches) and the objective function for the
basic feasible solution are: The t o t a l t r a n s p o r t a t i o n cost (far unit
fli.>w.) o f t h e branches f r o m t h e s o u r c e node t o
(xb)* = B-l b , (a14) any s i n k node must h e minimal.
z
* = c
b B-l
b . (alJ) The desired result can be achieved by
gradually adding branches one by one to the
Using (a15) and (a13), the objective network subtree, always satisfying the above
function can be written as : condition as well as the radiality condition.
min z = z* + c#x" . (al6) The derived optimality condition
serves
as the mathematical foundation of the trans-
shipment problem with quadratic costs to be
Interpretation o+ t h e optimal soluti on presented in the next section. It is also
noteworthy that transportation cost for unit
flow (established on Fig.1) R Jmax is a
Let the baijic feasible solution, similar base as the one obtained in /6/ for
described with ( x ) be the optimal solution optimal flow pattern.
Transshipment problem w i t h q u a d r a t i c costs Where:
Zm - i s t h e s e r i e s impedance o f branch
m .
1
781
Discussion node 1. Following the algorithm, either line section 1-2 or 1-3
can be added to the tree: adding 1-2 increases losses by 0.25 kW;
Ross Baldick, University of California, Berkeley, CA 94720. AS adding 1-3 increases losses by 2 kW, so line section 1-2 is added,
noted in [A], the search technique suggested in this paper is ba- increasing losses by 0.25 kW. Next, either sections 2 4 or 1-3
sically a ‘greedy algorithm.’ For the linearized loss model ini- can be added: adding 2 4 increases losses by 5 kW; adding 1-3
tially treated, the objective will be minimized by the greedy algo- increases losses by 2 kW, so line section 1-3 is added. Finally,
rithm [B]; however, there is no guarantee that the more accurate either 2-4 or 3 4 can be added: adding 2 4 will increase losses
quadratic approximation nor the actual losses will be minimized by 5 kW; adding 3 4 will increase losses by 4 kW, so line section
by a greedy procedure. Although greedy algorithms often perform 3 4 is added, for a total of 6.25 kW losses for the network created
well in practice, in general they are not optimal. The addition of by the algorithm. However, consider the configuration with line
voltage constraints exacerbates this problem. I therefore contest sections 1-2,24, and 3 4 closed and 1-3 open: the total losses are
point 1 made in the introduction to this paper. 5.25 kW, which is lower than obtained with the algorithm. The
To reinforce this theoretical observation, consider the exam- algorithm is clearly suboptimal in this case, although in practice
ple systems in the paper. The stated approximation error in power it might often give reasonably good results; the claims for the
loss in the first system is given as 7%, presumably as a proportion algorithm should be adjusted accordingly.
of the total losses; I assume that the author believes that this loss
estimate applies generally. In the second system, the improve- References
ment in losses is quoted as 7%, presumably as a proportion of [A] Hsiao-Dong Chiang and Rent5 Jean-Jumeau. Optimal Network
total initial losses: the error estimate is so large that we cannot Reconfigurations in Distribution Systems, Part 1: A New Formu-
be confident that the optimum has been achieved by the algo- lation and A Solution Methodology. Paper 90 WM 164-4 PWRD
rithm. Without checking the result against an algorithm, such as presented at the IEEE Power Engineering Society 1990 Summer
explicit search or branch-and-bound, that is guaranteed to find Meeting, Atlanta, Georgia, February 4-8, 1990.
the minimum, no claim of optimality is possible. [B] Eugene L. Lawler. Combinatorial Optimization: Networks
The following example demonstrates that the algorithm can and Matroids. Holt, Rinehart and Winston, 1976.
produce suboptimal results if there are compensating elements
such as shunt capacitors in the system. Consider the elementary
distribution network shown in figure 1, which has resistive lines,
a resistive and a resistive-inductive load, and a shunt capacitor.
For ease of calculation the loads are specified as current sinks,
although the same qualitative results could be obtained using
the equations in appendix B of the paper with loads specified
as power and reactive power. Furthermore, we assume that the
voltage constraints are not binding.
Substation
l 7
f?ptIlllal!ty C!OIldltlf?ll ( 6 ) ! i l i O l l l ( j bQ G a t i f l f l O d
i n o r d e r t o o b t a i n an o p t i i n a l : ~ o l u t l o i i .
VLRSTIMIR GLAMOCANIN. T t i e ilLlti)<.,r w o u l d 1 I k ( 7
For completeness, tlie propouod a l y u i l ~ I I I I I
to thank Mr. Raldick for his discussion of the will he applied to tho e l e i n n n t a r y distribution
paper. network wlth t h o cnnctraiiitt: and the d a t a
provided by the discussor.
1. Mr. Baldick points out the question I step: Tritlally the s u b t r e e ES consists of
of approximation error of 7 % a s a portion of node 1 a s a souree node: ES = [ 1-11. Using
the total losses in the first system presented ( 1 ) and (2) the transportation cost of unit
in the paper. That is incorrect because, the current of line sections can be estimated:
use of (5) in estimation of the change in c =R J =R =O. 1 . Because the
power losses, acc,ording to the paper, produces 2 1-2 max( 1-2) 1-2
less than 7% error,in comparison with the current capacity of line sections J is not
max
change in power losses obtained by exact provided by the discussor i t is assumed that
radial powpr flow method. the same conductor is used for each line
However, one may question whether or not segment: c =R =O. 2 ; c = x i .
an approximated load flow algorithm, in this 3 1-3 4
case ( 5 ) , should be used. To answer, let us I1 step: To satisfy first optimality
consider a sink node which does not belong to condition the selection of the node which will
the current network subtree: enter the subtree is performed using (3) :
- If there is not any branch which could c = min [ c = 0 . 1 ; c =0.2 i c4='? i
2 2 3
connect the sink node to the subtree - that is I11 step:Network subtree ES has to be updated
a trivial case. by section 1-2 : ES = ES (-1 1 1 - 2 1 = [ 1 - 1 ; 1 - 2 1 .
- If there is only one branch which connects Because there is only o n e section which
the sink node to the subtree - that is also a connects node 4 with s u b t r e e ES , second
trivial case: second optimality condition ( 6 ) optimality condition ( 6 ) is not applied.
does not need to be fulfilled. Therefore, the minimal total cost of node 4
- If there are two or more branches which could be calculated using ( 7 ) :c =c t R
connect the sink node with subtree - the 4 2 2-4
change in power losses should be estimated =O.ltO.l =0.2 . The value of minimal total
preliminary by ( 5 ) . If the estimates have the cost of node 3 remains as before.
values which are not separable (with respect I1 step: c = min I c = 0 . 2 ; c = 0 . 2 1
4 3 4
of the expected error approximation), the 111 step: ES = ES 'U (2-41 = 1 1 - 1 ; 5 - 2 ; 2 - 4 1 .
exact estimation of the change in power losses The obtained network is the optimal
should be performed by radial flow method configuration of the network with removed
given in Appendix B . The purpose in pointing shunt capacitor (node 3 is treated as a
out the value of error tolerance is to provide transshipment node).
an information - when ( 5 ) could be used. However, the author does not see any
problems in applying the proposed algorithm on
2. Question has been raised regarding the network with shunt capacitor at node 3 :
the validity of the proposed algorithm applied IV step: Node 3 could be connected to the
to the networks with compensating elements. current subtree with two sections 1-3 and 4-3.
The discussor's observations contain two Applying second optimality condition to node
improper premises: 3 the following approximated change in power
The discussor's sugestion to solve the
losses is obtained: L P L 3 ( 1 ) = 2 k W ; dPL3(4)=OkW.
problem of optimal network reconfiguration and
the problem of capacitor placement is 111 step: Because APL , node 3
3(4) < "L 3 ( 1 )
understandable, but i t Is the author's opinion
a s the last node of the network should be
that he would n o t recomend to solve them
connected by line section 4-3:
simultaneously. An earlier work t 1 A l has
ES =ES (-1 ( 4 - 3 1 = [ 1 - 1 ; 1 - 2 ; 2 - 4 ; 4 - 3 1
shown that to avoid biasing the solution, the
The obtained solution is exactly the
capacitors should be removed during the
same as the one advocated by the discussor.
process of optimal network reconfiguration.
In fact, i t is realized fully by the proposed
The additional loss and voltage profile
algorithm ( hPL =OkW 1 that the additional
improvement could be obtained by applying 3(4)
capacitors (determining their number and savings could not be obtained by applying the
locations) to the reconfigured network . shunt capacitor at node 3 .
Mr. Baldick in his discussion shows References
that i f only second optimality condition ClAl R.E.Lee and C.L.Brooks, "A Method and its
(6) is used, the suboptimal result would be Application to Evaluate Automated Distribution
obtained. In this regard the discussor is Control", I A E E Trans. on P o w e r Delivery, Vol.
reffered to the basic model of the paper which 3 , No. 3 , July 1 9 8 8 , pp. 1 2 3 2 - 1 2 3 0 .
explicitly states that b.~
- oth first
optimality condition (3) and second