S. C i v a n l a r J.J. G r a i n g e r H. Y i n S.S.H. L e e
Member, IEEE Senior Member, IEEE Non-Member Senior M e m b e r , IEEE
Abet ract
Feeder reconfiguration is defined as altering the topologi- Consequently, the peak loads on substation transformers, on
cal structures of distribution feeders by changing the individual feeders, or on feeder sections occur a t different times
open/closed states of the sectionalizing and tie switches. In this (noncoincidence of peaks). Feeder reconfiguration allows the
paper, a scheme is presented which utilizes feeder transfer of loads from heavily loaded feeders (or transformers)
reconfiguration as a planning and/or real-time control tool in to relatively less heavily loaded feeders (or transformers). Such
order to restructure the primary feeders for 1033 reduction. The transfers are effective not only in terms of altering the level of
mathematical foundation of the scheme is given; the solution loads on the feeders being switched, but a h in improving the
procedure is illustrated on simple examples. voltage profile along the feeders and effecting reductions in the
overall system power losses. Studies and experiments on feeder
INTRODUCTION reconfiguration are ongoing in many utilities and some of the
recent publications reflecting these efforts are listed in the refer-
The explosive growth in the areas of micro- and mini- ences [5-111.
computers, microprocessors and telecommunications technolo-
gies provides opportunities for advanced control of electric This paper discusses the problem of reducing power losses
power systems, particularly in the area of distribution system in distribution feeders via feeder reconfiguration. A scheme for
automation. Prototype distribution automation systems are determining the open/closed states of the tie and sectionalizing
currently being developed and tested on a small-scale basis switches t o achieve loss reduction is presented. While the scope
using the presently available data acquisition system [l-41. of the feeder reconfiguration problem discussed here is limited
Intensified research, development and demonstration efforts are to the discussion of losses, the results developed provide
now being directed towards the hardware and software to sup- significant insight into useful characteristics associated with the
port large-scale distribution automation schemes for system- modeling and properties of related feeder reconfiguration prob-
wide implementation by the electric utility industry. Three of lems.
the current distribution automation projects are a t
I WI
**
/----+-
".I..
The problem in question is now illustrated using the switching options for the example system. Actually, the total
three-feeder distribution system shown in Figure 2. The dotted number of switching options is much larger than fifteen; but
branches, 15, 21 and 26, represent ties connecting feeders, and some of those which cause isolation of part of a feeder are
normally open tie switches are assumed to be present on these directly eliminated. As noted earlier, the best switching option
branches. could be selected from the results of thirty load-flow studies
simulating all fifteen possible feeder configurations.
For notational convenience, these tie switches will be
identified by the corresponding tie numbers. Without loss of
generality, and mindful of the practical situation, let us assume
for ease of explanation that there are sectionalizing switches on DESIRABLE FEATURES OF SOLUTION METHOD
every branch of the system. All thirteen sectionalizing switches
will also be identified by the corresponding branch numbers. In light of the dimensionality concerns addressed above, it
appears that a desirable solution method should provide the
The load a t bus I 1 can be transferred to Feeder-I by ClOS- following two features: (i) capability to estimate with minimal
ing the tie switch 15 and opening the sectionalizing switch 19. computational efforts the change in losses resulting from feeder
Similarly, the loads at buses 9, 11 and 12 can be transferred to reconfiguration and (ii) criteria that may be used to eliminate
Feeder-I by closing the tie switch 15 and oDeninc the undesirable switching options in order to alleviate the dimen-
sectionalizing switch 18. Throughout this paper, we will focus sionality problem. I t will be seen that the formula developed in
our discussion on feeder reconfiguration by closing a single tie this paper for estimating the change in losses requires little
switch and opening a single sectionalizing switch to preserve additional information over the base case (i.e. prior to feeder
radiality of the two feeders that are under consideration. The reconfiguration) load-flow solution. Furthermore, the formula
combined pair consisting of a tie and a sectionalizing switch suggests a filtering mechanism for eliminating those switching
will be referred to as a switching option. While the successive options which would not yield loss reduction.
application of the proposed scheme could handle the case of
multiple switching operations in which several tie and sec- The primary objective in deriving the expression for power
tionalizing switches are simultaneously closed and/or opened, loss reduction via load transfer is to determine (i) whether a
the detailed implementation is beyond the scope of this paper, specified switching option would result in a 1033 increase or
and will not be further discussed here. decrease, and (ii) among the candidate switching options, which
option would yield the greatest reduction in IOSSeS. In other
It can be easily verified that there are fifteen feasible words, relative rather than absolute accuracy is sought here.
1219
The amount of loss change resulting from transferring a Bua t o Section Section End Bus End Bun End Bus End Bum
group of loads from Feeder-I1 to Feeder-I can be estimated Bus Resistance Reactance Load Load Capacitor Voltage
(P U) (P.U) (MW) (MVAR) (MVAR) (P.W)
from the following simple equation: 1-4 0 075 01 20 16 I 0 9911-0 370
4-5 0 08 0 11 3 0 15 1 11 0 988/-0 544
4-e 0 09 0 18 20 08 I 12 0 986/-0 697
6-7 0 04 0 04 15 I2 0 9861-0 704
2-8 0 I1 0 11 4 0 27 o 9791-0 783
where 8-9 0 08 0 11 50 30 I2 0 9711-1 461
8-10 0 11 0 11 10 09 0 9771.0 770
D set of buses which are disconnected from Feeder-I1 and 9-11 011 0 41 08 01 06 0 971/-l625
9 12 008 0 11 45 20 37 0 9891-1 636
connected to Feeder- I
3 13 011 0 11 10 0 9 0 994/-0 332
m tie bus of Feeder-I to which loads from Feeder-I1 will be 13-14 009 0 12 10 07 18 0 9861-0 459
connected 13-16 008 0 11 10 09 0 992/-0 627
15-16 I 004 I 0 04 21 I 10 18 I 0 991/-0.596
n tie bus of Feeder-I1 that will be connected to bus m via 5-11 1 004 I 0 04 I I
a tie switch 10-14 I 004 I 0 04
17-18 I 0.09 I 0.12 I I I U
Ii complex bus current a t bus i
Rloop series resistance of the path connecting the two substa-
tion buses of Feeder- I and Feeder-I1 via closure of the (I) Suppose the load a t bus 11 is transferred from Feeder-I1 to
specified tie switch (see the example given below) Feeder-I by closing the tie line switch 15 and opening the
component of E = RBusIBUs corresponding to bus m. sectionalizing switch 19. In this case, D={ll}, m = 5 ,
Em
REUS is the "bus resistance matrix" of Feeder-I before n = 11, and
the load transfer which is found using the substation AP = Re[2Ill(E5- Ell)'] + Rloop!!ll12
bus as reference. ISUS is the vector of bus currents for
where RIoop is the total resistance of the path along the
Feeder-I
branches 11,12,15,19,18 and 16.
E, similar to E, but defined for bus n of Feeder-11.
(11) Suppose the loads a t buses 9, 11 and 12 are transferred
Re{.), *, 11 from Feeder-I1 to Feeder-I by closing tie switch 15 and
real part, complex conjugate, and magnitude operators, opening sectionalizing switch 18. In this case,
respectively. D={9,11,12}, m =5, n = 11 and
It is to be noted that E, and E, are computed using base-case AP = Re[2(19 + I , , + Ilz)(E5-Ell)*]
bus currents I, before the load transfer. It is suggested to
incorporate the effects of capacitors into bus currents to facili-
tate computational efficiency. The derivation of Eq.(l) is pro-
where R I o o pis the same as above.
vided in the Appendix from which it is apparent that AP
represents a kw loss reduction (increase) when it is negative
(positive). The second term on the right-hand side of Eq.(l) is always
positive. Therefore, a reduction in losses cannot be achieved
unless the first term becomes significantly negative. Since com-
FEEDER-I FEEDER-I1 FEEDER-I11 plex values are dealt within the first term, it may not be simple
::
U VJJJ U to draw any definite conclusions. However, we note that voltage
-7 22
phase-angle differences are small on most distribution systems,
and that complex bus currents I, may be mostly in phase with
voltage phasors due to capacitor VAR compensation on well
designed systems. Under these circumstances, loosely speaking,
'0 the first term becomes negative if bml< bn\
It follows from the above observation that loss reduction
can be attained only if there is a significant voltage difference
across the normally open tie switch and if the loads on the
23 higher voltage drop side of the tie switch are transferred to the
other side. It will be seen that the above observation can be
d 14
n
6-26- 25 1 used as a most attractive criterion to eliminate undesirable
switching options during the elimination process.
It is also noteworthy that in Eq.(l), information regarding
E is required only a t the terminal buses where the tie switch is
located, and that the configuration of the group of loads to be
Figure 2 Three-feeder ezainple system
transferred or the geographic extent of the overall distribution
system does not matter to the result.
Example: STRATEGY
The use of Eq.(l) is now illustrated using the three-feeder sys-
tem of Figure 2. The simplicity of the proposed methodology makes it s u i t
1220
able for an on-line control strategy for feeder loss reduction. A R
The strategy for selecting the best switching option is further
explained via the example system of Figure 2. Shown in Table 1
are the bus data, line data and the base-case load-flow solution.
The bus voltage magnitudes and angles (in degrees) are shown
in the right-most column.
When closing the tie switch 15, five options for opening
t
sectionalizing switches 11, 12, 19, 18, and 16 are available.
>
Since bIl1 b5!, transferring loads on Feeder-I to Feeder-I1 is
expected to increase losses. Consequently, opening the sec-
tionalizing switch 11 or 12 is regarded as undesirable and need
not be considered. Therefore, associated with closing the tie
switch 15 are three candidate options, viz., opening the sec-
tionalizing switches 19, 18, and 16, respectively. Similarly,
since b1d> bl.,!,opening switch 22 or 24 is considered to be
undesirable when the tie switch 21 is closed. For a similar rea-
son, transferring loads on Feeder-I11 to Feeder-I when the tie
n........
- Results from Load-Flw __c
.......
4
2.5
r;
(x'
One can easily note that the term iR i a (k,j)I j is the vol-
I=,
The resistances of the sections along the shortest paths as indi-
cated in Figure B.1 are represented by R I , j=1,2, ... and
tage drop between bus k and the substation assuming a purely RI', = 1,2,.... The voltage drop E , can be computed from
resistive equivalent feeder. Denoting this voltagedrop as fl
E, = c RbA,&J)I,
]=2
P.6)
Ek = ]=1 Rd,, (k,j) (A.9)
Here R& corresponds to Rhus of Feeder-A of Figure B.1.b
Eq.(A.B) can be rewritten as whereas RE, corresponds to that of Feeder-B of Figure B.1.a.
The change in power loss for the transferred feeder section
A P k , = 2 R e [ I , E:] + I I, I R k 3 (k,k) (A.10) which is shown with dashed lines in Figure B.l is
Here Rd,, (k,k)corresponds to the total resistance of the shor-
test path between bus k and the substation. APk,, = I Il I R , - 1 Il 1 R l (B.7)
F.Derivation of E a .fl)r. where R , is the resistance of the normally-open branch (1,l').
The proof is made by induction in three steps as follows : The total power loss will then be
Step - (i) : Prove that Eq.(l) is valid for transferring the A P l = AP,& i AP,& APkSa
i- (B.8)
lateral that carries a total current of I, connected to the tie bus Using the following identities
1 shown in Figure B.1.a from Feeder-A to Feeder-B, i.e.
E, = E, + I1 RI (B.9)
A P , = 2 R e [ I ,( E,, - E, )'] + I I , 1' Rloop (B.l)
where E,., E l are the voltages of the tie buses, computed using E2 = E2 - Wkm) (B.lO)
Eq.(A.9), and Rloopis already defined in the main text. Using and substituting Eqs. (B.4), (€3.9) and (B.10) into Eq. (B.2), it
Eq.(A.10) the changes in power losses on Feeder-A and Feeder- can be shown that
B (See Figure B.l) are f l
AP;,,= 2 R e [ - 1 1 El ']+lI1lz
AP;,, = - [Z Re ( I , E l ') + I I, I R& (2,2)] (B.2) ] =2
Rl+211112R1 ( ~ . i i )
f 2
A P l = 2 R e [ I l ( E,, - E , )' 1+ 11, I Rloop (B.13)
RbBu6 (lY1) = Rf where
]=1
1223
Substituting Eqs.(B.18) and (B.19) into Eq.(B.l7), 6P can be
found as
6 P = 2 Re I Z O + ~ )( E l f ~ El ) * J
Step - (ii) : Assume that Eq.(l) is valid for opening the switch
o n section 1 of Figure B . l t o transfer the parts of the Feeder-A
t o Feeder-B, i.e. Upon substituting Eqs. (B.20) and (B.15) into Eq.(B.lG), it can
A P f = 2 Re [( Zl +Iz + . . . + I f )(El, ~ El) ' ] be readily verified t h a t the resulting A P ( / is that of Eq.(l)
(B.15) applied t o Figure B.1 when the ( 1 + 1 ) '"switch is opened.
+ I (I1 +I, +... + I f) l2 RIoop This completes the proof by induction.
where Zj represents the total current a t the j t h lateral along
the shortest path between bus 1 and the substation of Feeder-
A.
1 Discussion
+RI( i Z J ) +Rz(
1=2 1 =3
+ '.. +R(i-1)4
and
f l I
E(i+i)= E1 - ( 2 RI, ( C 1, (B.19)
1=(I + 1) ]=1
R l I1