Professional Documents
Culture Documents
This work was partially suppolted by CEEE, Companhia Estadual de (cutting load) at nodes.
Energia Elttrica, and by FAPERGS, Funda@o de Amparo i Pesquisa do
g - Column vector of current injections at nodes.
Estado do F h Grande do SUI. -
S.Haffner, L. A. Pereira. L. F. A. Pereira, F. A. B. Lemoe, A. L. de Lima, - Column vector o f minimum and maximum
-
g and g values for injections of current at nodes.
C. A. Brixius and L. R. Giron are members of GSEE, Electrical Energy
Systems Group of PUCRS, Pontifical Catholic University of Rio Grande do d - Column vector of demands for current at nodes
SUI. (e-mail: haffner@ieee.org, Ipereira@eepucrs.br, pereira@ee.pucrs.br, (Ioads).
lemos@ee.pucrs.br, schelos@ee.pucrs.br,brixius@ee.pucrs.br, Irgiron@pucrs.br). V - Column vector of voltage magnitudes at nodes.
L. F. A. Pereira is a member of GACS. Automation and Systems Control - Column vector o f minimum and maximum
Group of PUCRS, Pontifical Catholic University of Rio Grande do SUI.(e- y and
- voltages at nodes.
mail: pereira@ee.pucrs.br).
S F - Node-branch incidence matrix of the fixed divided into various steps [8]-[lo]. Irrespective of the model
network. used to represent the problem and of the technique used to
SR - Node-branch incidence matrix of the pre-existing
network associated with removal altematives. solve it, a strong link is found between the model and the
Node-branch incidence matrix associated with method of solution. In some cases, additional simplifications
ss -
substitution alternatives in the pre-existing are introduced so as to adapt the model to the limitations
network, imposed by the method of solution.
Node-branch incidence matrix associated with
S“ - An analysis of published work on the subject shows that
alternatives for removal or substitution in the
preexisting network. very many models are used to help distribution companies in
SA - Node-branch incidence matrix associated with the complex tasks of allocating equipment, satisfying a series
altematives for addition to the preexisting of constraints and seeking to minimize costs of expansion
network.
Impedance matrix of branches of the fixed
(investment) and of operation (losses, penalties for failure of
ZF - supply, etc.).
network (diagonal matrix whose elements
correspond to the branch impedance magnitude). For the optimization model to meet the desired needs, it is
ZR - Impedance matrix of branches that are candidates necessary to take full account of the individual characteristics
for removal (diagonal matrix whose elements of the particular distribution company involved, and also the
correspond to the branch impedance magnitude).
Impedance matrix for branches that are background in which this company operates. The company’s
Z,” and Zf -
characteristics define the standards adopted in its installations,
candidates for substitution. Diagonal matrices
whose elements correspond to the branch and determine the main technical characteristics of the
impedance magnitude in the initial (pre-existing) distribution network; the operational background defines a set
and final (substitution alternative) configurations.
Impedance matrix for branches that are of service constraints (voltage drop, energy quality, reliability,
candidates for removal or substitution. Diagonal etc.) to be considered in the analysis. Thus the model to be
matrices whose elements correspond to the used must portray as faithfully as possible the distribution
branch impedance magnitude in the initial (pre- network to be analyzed, in such a way that the results obtained
existing) and final (substitution alternative)
configurations.
can be useful in practice. Unlike the planning of expansion of
M - Diagonal matrix whose elements are large transmission systems [l I], the planning of distribution
enough to keep restrictions always inactive. systems must consider radial networks, the possibility of
- Total number of isolated (island) nodes in the union or subdivision of feeders, and various alternative cables
%land
initial network configuration.
for the same stretch. In many cases, it is also necessary to take
into account maximum and minimum limits for the network
11. INTRODUCTION
voltage.
458
~
Thus the general problem of planning the expansion of a fJ'+[z~~'[ssp' 'v2-M(1-xS) (6.2)
distribution network- becomes a problem of mixed integer
f' + [z$[rP .v 5 M Y
programming, defined as follows.
,r+[2$[SrP .V2-MX'
A. Objective function
The obiective function is the minimization of the cost of fS +[z;.r'[sTp . Y 5 M(I - 2 )
installation and alteration to the distribution network, with a (8.2)
penalty for failure to meet demand (shedding load):
c ~ ~ x ~ + c ~ ~ +c' x ~ex'++cA*xA
c ~ + +xa .~r (1) JA +[Zq'[pP . v M ( l - x q (9.1)
(9.2)
459
~
From the operational limits, we have: Although the results obtained with this simplified model
are not the same as those given by the solution of non-linear
load flow, only small differences were found in tests
undertaken in radial distribution networks with three types of
load (constant power, constant impedance, and a mixture of
50% of each) and a typical power factor of 0.85. Moreover,
such differences showed a definite trend, with errors of up to
(13) 2% in the value of nodal voltage. Considering that the
permitted values for voltage drops are in a range fiom 5 to
lo%, this error is unacceptable, but it is possible to reduce it
considerably by introducing a correction factor K to the
branch impedance values, of the following form:
f'=O, vx;=l,lE1;Zrs zEmced= K .zbn (26)
The factor K must be determined for each type of
If+?", vx;=1, Z€R" conductor, and can be obtained using the values of voltage
J" =o, vx; =o, [Enrs drop calculated from the simplified linearized model ( A V ~ )
(18)
and from values obtained in solving a non-linear load flow
problem (AV;'):
fA=O, vx:=o, m€QA (20)
-vrvsv (21)
I
460
~
variables are shown, indicating the possibility of installing for solving linear mixed integer problems (BDMLP, CPLEX,
two different kinds of cable). CPLEXPAR, OSL, OSL2, XA, XAPAR, XPRESS, ZOOM).
x; x;
I
B. 16node network
The 18-node network has I I branches in the initial
3@- / \ '\ '. 1 x6" configuration (2 can be substituted by a line of different
? gauge; 9 removed or substituted by lines of different gauge)
'.'\
1
and I3 new candidate branches (4 of them having two gauge
options), as shown in Fig. 3. The nominal network voltage is
13.8 kV, operation between 13.11 kV and 14.49 kV being
permitted. At each of the network nodes numbered from 1 to
16 a demand of 50 A exists; the capacities of the sub-stations
are 500 A. The line data are shown in Table 111, using the
following costs: 40 for removal; 20 for substitution (cable 4/0
d2 by cable 336.4); -20 for substitution after removal (in this
Fig. 2. S-nodenetwork.
case the total cost is 20, since there would be a cost of 40 for
removal plus -20 for substitution after removal); 100 for an
Table 1 gives the data for the nodes, namely: demand,
addition (cable 4/0); 120 for an addition (cable 336.4);
generation limits, and voltage limits.
additions of 200 for sub-station output modules to each new
TABLEI line (cables 410 or 336.4). The impedance of the reaches
NODE DATAFOR THE 5-NODE NETWORK. corresponds to a distance of 2 km in a structure with
geometric mean spacing of 1.322 m (typical for a standard
structure). Taking all the possible alterations into account, the
number of possible combinations is about 18 million.
Table I1 shows the data for the lines, namely: the initial and j x;'
final node, the impedance (initial and final, in the case of a
substitution), maximum current (initial and final, in the case of
a substitution) and cost, preceded by the associated decision
variables. As the proposed model is more general than that
described in [ 6 ] , some options for alteration and additional
expansion were also added (only the lines corresponding to
the decision variablesx: , x;", x,", e X: form part of the
original problem; the others were introduced with unattractive
costs so as not to alter the solution to the problem), thus
increasing the number of possible combinations from 8 to 384.
s," I I I 3 I 0.05 250 1 1OD The solution obtained, shown in Fig. 4, has a cost of 560
for the following alterations: substitution of the initial reaches
I . ! I I I I I
of feeder lines already installed ($ =x: = I), removal and
substitution of the reach 12-16 (x; = I > e addition of 5
x i 1 2 1 4 I 0.0113 100 8.7 17"
new reaches (x; =I,"= = x t =x: = 1). As the cost of
substitution is considerably less than the cost of installing a
new feeder, increasing the capacity of the feeders was the
option selected, with the load equally divided amongst them,
The solution found is given in [6], with.$ = X: = I and the since the nominal capacity of each is 400 A. Because the
remainder zero, after using all the solvers available on the maximum voltage drop allowed, the conductors in the reach
GAMS platform (Version 2.0.13.0) [13] that are appropriate 12-1 6 were changed. in addition, least-cost cables were
46 1
6
always selected for making connections to nodes that were Table IV shows a comparison between magnitudes of
previously isolated. nodal voltages obtained using the simplified model and using
the load flow solution, considering three types of load
TABLE111
LINE DATAFOR THE 18-NODENETWORK (constant power, constant impedance, and 50% of each). The
x From To
I
TABLE IV
COMPARISONBETWEEN NODALVOLTAGES OBTAINEDUSING THE SIMPLIFED
I I
It is seen that the greatest differences occur when the load
$ 9 10 1.0873 250 100 is modeled by an injection of constant power (third column of
4 9 13 1 .U873 250 100 Table IV), but that all values less than 0.5%. Obviously, if the
x& 9 17 1.0873 250 300 loads were represented by injections of constant current the
I -- ! I
0.8457
differences would be smaller still.
I .oan
VI. CONCLUSIONS
3 1.0873 250 100
The optimization model presented here for planning the
1.0873 250 300
expansion of distribution systems has been shown to be
0.8457 400 320
I ~- ! I flexible and considerably general in application, allowing for
x i 13 17 1.0873 250 300 the various kinds of possible alterations: removal of a reach
x; 13 17 0.8457 400 320 (sectioning the feeder), change in conductor gauge (re-
capacitation), and the option of gauge changes. The possibility
of representing these different kinds of alteration results in
final costs being obtained which are less than those found
using simpler models, in which only some of these
alternatives can be contemplated.
a-0 xq =1
The main constraints of the real-world problem were
considered: namely, the capacity of conductors, the limit in
voltage drop and maintenance of network radiality. The
%
simplification used in representing Kirchhoff s Voltage Law
was found to be very effective, as the results obtained are very
close to those for non-linear load flow, so that an optimization
12 ='
@ xi=1 model is derived for which all constraints are linear, by means
of the linear disjunctive model.
The efficiency of the model was tested using a 5-node
network whose solution is already known, and another with
18 nodes whose solution can be easily interpreted and
Fig. 4 . Result for the 18-node network. evaluated. Given some adaptation, the model can be further
462
7
extended to consider various stages, so that long-term [6] K. Aoki, K. Nara, T. Satoh, M. Kitagawa, K. Yamaoaka, “New
Approximate Optimization Method for Distribution System Planning”,
planning horizons can be allowed, and benefits obtained IEEE Trans. on Power Systems, Vol. 5. No. 1 , pp. 126132, February
thorough economies of scale that result from anticipating 1990.
important future investments. /7] K. Nara, T. Satoh, H. Kuwabara, K. Aoki, M. Kitagawa, T. Ishiham,
“Distribution Systems Expansion Planning by Multi-Stage Branch
Exchange”, IEEE Trans. on Power Systems, Vol. 7 , No. 1, pp. 208-214,
February 1992.
[SI K. “a, T. Satoh, K. Aoki, M. Kitagawa, “Multi-Year Expansion
VII. REFERENCES Planning for Distribution System”, IEEE Trans. Power System, Vol. 6,
NO. 3, pp. 952-958, August 1991.
[9] I. J. Ramirez-Rosado, T. Gonen, “Pseudodynamic Planning for
H. K. Temrar, V. H. Quintana, “Distribution System Expansion Expansion of Power Distribution System”, IEEE Tram. Power Systems,
Planning Models: An Overview”, Elecfric Power Sysrems Reseurch, Vol. 6, No. I, pp. 245-254, February 1991.
Vol. 26, pp. 61-70, 1993. [lo] H. Kuwabara, K. Nara, “Multi-Year and Multi-State Discrjbutiod
S. K. Khawr, L. C. Leung, “Power Distribution Planning: A Review of Systems Expansion Planning by Multi-Stage Branch Exchange”, IEEE
Models and Issues”, IEEE Trans. on Power Systems, Vol. 12, No. 3, pp. Trans. Power Delivery, Vol. 12, No. I , pp. 457463. January 1997.
1151-1159, August 1997. [ll] R. Romro, A. Monticelti, A. Garcia, S. Hafier, “Test Systems and
E. Lakervi, E. J. Holmes, Elecfriciry disfributfoion network design, 2nd Mathematical Models for Transmission Network Expansion Planning”.
ed., IEE Power Series 21, London, Peter Peregrinus Ltd., 1995, p. 325. IEEProc-Gener. Trunsm Distrib., Vol. 1491, No. I , pp. 27-36, January
D. I. Sun, D. R. Farris, P. J. Cote, R. R. Shoults, M. S. Chen, “Optimal 2002.
Distribution Substation and Primary Feeder Planning Via the Fixed [12] D. Chattopadhyay “Application of General Algebraic Modeling System
Charge Network Formulation”, IEEE Trans. Power Appurafw and to Power System Optimization”. IEEE Truns. Power Sysrems, Vol. 14,
System, Vol. PAS-101, No. 3, pp. 602-609, March 1982. No. 1, February 1999.
M. A. El-Kady, “Computer-Aided Planning of Distribution Substation 1131 A. Brooke, D. Kendrick, A. Meeraus, R. Rsman, “GAMS A User’s
and Primary Feeders”, IEEE Trans. Power Apparum and Systems, Vol. Guide’’, GAMS Development Corporation, 1998, Washington, USA.
PAS-103, No. 6,pp. 1183-1189, June 1984. [14] ETAP PowerStation Version 4.7 OE,Operation Technology, Inc.(OTI),
USA, http://www.etap.cod.
463