Está en la página 1de 13

Approximating electrical distribution networks via mixed-integer

nonlinear programming
Sanyogita Lakhera
a,1
, Uday V. Shanbhag
b,,1
, Michael K. McInerney
c
a
Citibank, New York City, NY
b
Department of Industrial and Enterprise Systems Engineering at the University of Illinois at Urbana-Champaign, 117 Transportation Building,
104 S. Mathews Ave., Urbana, IL 61801, USA
c
Construction Engineering Research Laboratory (CERL), USA
a r t i c l e i n f o
Article history:
Received 23 July 2009
Received in revised form 5 July 2010
Accepted 13 August 2010
Available online xxxx
Keywords:
Optimization
Nonlinear programming
Mixed-integer nonlinear programming
Distribution system design
GIS
a b s t r a c t
Given urban data derived from a geographical information system (GIS), we consider the problem of con-
structing an estimate of the electrical distribution system of an urban area. We employ the image data to
obtain an approximate electrical load distribution over a network of a prespeciced discretization.
Together with partial information about existing substations, we determine the optimal placement of
electrical substations to sustain such a load that minimizes the cost of capital and losses. This requires
solving large-scale quadratic programs with discrete variables for which we present a novel penaliza-
tion-smoothing scheme. The choice of locations allows one to determine the optimal ows in this
network, as required by physical requirements which provide us with an approximation of the distribu-
tion network. Furthermore, the scheme allows for approximating systems in the presence of no-go areas,
such as lakes and elds. We examine the performance of our algorithm on the solution of a set of location
problems and observe that the scheme is capable of solving large-scale instances, well beyond the realm
of existing mixed-integer nonlinear programming solvers. We conclude with a case study in which a
stage-wise extension of this scheme is developed to reect the temporal evolution of load.
2010 Elsevier Ltd. All rights reserved.
1. Introduction
Obtaining information pertaining to a citys utility networks is
crucial for purposes of planning, maintenance and redesign. Yet,
this information is often inaccessible to most agencies and a rel-
evant question is how one may approximate such information
through the usage of only image data. This provides a motivation
for estimating the underlying distribution system of an urban
area, given an image from a geographical information system
(GIS). Such an image captures the parcel data as well as a subset
of existing substations in the urban area. We consider a novel ap-
proach that comprises of two basic steps. First, the input image is
used to construct an electrical load distribution on a grid of pre-
specied discretization. Note that a grid, in this context, refers to
a regular fully connected network over which a distribution net-
work will be specied. Second, we solve an inverse problem that
estimates the set of lines that correspond to such a load distribu-
tion. Several issues complicate such an estimation. In general, the
true distribution system is a radial network and the resulting in-
verse problem falls within the realm of optimization problems in
function space and is, in general, intractable. Instead, we restrict
the set of possible networks to those that can be specied as
graphs on a grid of chosen size. Two additional complexities
emerge from modeling distribution systems in urban areas. First,
there are signicant areas that cannot be covered by the distribu-
tion system (such as lakes or elds). Therefore, the optimal solu-
tion has to reect these restrictions. Second, a clear evolution
pattern exists in the growth of the load and needs to be respected
in estimating the distribution system. For instance, if a particular
part of a township developed earlier than another, then the distri-
bution system would have such a structure.
The resulting problem can be recast as an mathematical pro-
gram in nite-dimensional space in which one seeks a set of ows
that satisfy the substation capacity constraints, Kirchhoffs conser-
vation equations and voltage bounds. Unfortunately, this problem
can be infeasible if the substation information is inaccurate. To
avert this possibility, we consider a problem in which we deter-
mine the installation of incremental substations as well as the
ows that emerge from the resulting system. This optimization
problem falls within a class of mixed-integer nonlinear problems
(MINLP) and has a size that grows with the level of discretization
and the number of substations. Currently no solvers exist for
0142-0615/$ - see front matter 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijepes.2010.08.020

Corresponding author.
E-mail addresses: sanyo_dce@yahoo.com (S. Lakhera), udaybag@illinois.edu
(U.V. Shanbhag), Michael.K.McInerney@usace.army.mil (M.K. McInerney).
1
Authors have been partially supported by Award No. US Army/Construction and
Engineering Research Laboratory (CERL) W9132T-07-C-0010.
Electrical Power and Energy Systems xxx (2010) xxxxxx
Contents lists available at ScienceDirect
Electrical Power and Energy Systems
j our nal homepage: www. el sevi er . com/ l ocat e/ i j epes
Please cite this article in press as: Lakhera S et al. Approximating electrical distribution networks via mixed-integer nonlinear programming. Int J Electr
Power Energ Syst (2010), doi:10.1016/j.ijepes.2010.08.020
addressing problems arising from practical networks [17]; instead,
we provide a framework that can cope with large-scale instances.
In this paper, we make the following contributions:
1. By converting the image data into electrical load data, we recast
what was essentially an innite-dimensional problem as a
nite-dimensional discrete nonlinear optimization problem.
This problem is capable of encapsulating a variety of complex-
ities ranging from variable substation sizes, restricted areas and
variable feeder sizes.
2. Solving such discrete optimization problems is only possible
when the space of discrete variables is small. To address this
challenge, we present a penalization-smoothing scheme that
solves a continuous form of this problem. We observe that this
allows for solving large-scale instances to local optimality with
respect to the smoothed formulation. Furthermore, the scheme
is shown to scale well with the number of discrete variables.
3. The framework allows for modeling a variety of complexities.
We show that restricted or no-go areas can be accommodated
within the framework and the obtained results do indeed reect
the constraints.
4. Finally, we extend the framework to allow for a multi-stage
approach towards approximating the distribution system in
an urban area by using a sequence of images from the GIS
system.
The remainder of this paper is organized as follows. In the
remainder of this section, we review past work in this area. In Sec-
tion 2, we provide an outline of our methodological framework and
has three subsections. In Section 2.1, we provide some details per-
taining to the image data while in Section 2.2, we discuss the struc-
tural properties of the constraints arising in such problems. In
Section 2.3, we dene the nonlinear discrete optimization problem
and its smooth generalization. In Section 3, we compare the perfor-
mance of a standard commercial MINLP solver with the smooth
approximation, particularly from the standpoint of the nal value
as well as the scalability. We discuss the performance of our ap-
proach when attempting to estimate the distribution system of
the city of Champaign in Section 4. The paper concludes in Section
5 with a summary and a set of possible research directions.
1.1. Background
The network ow model for distribution system design was
adequately discussed by Willis et al. [24,25]. The use of an optimal
planning approach toward the design of electrical networks was
rst illustrated by Knight [13]. The idea here was to have a mini-
mum cost network design using a network ow algorithm based
on a set of geographical locations. Other work highlighting similar
approaches include [4,15,12,26,1,23]. A succinct review of network
ow algorithms is available in [2] while power systems analysis
and optimization can be found in [28].
Sun et al. [21] used the concept of a long range horizon year
(target) together with a time-phased expansion process beginning
with the base year and progressing on to the target year. The idea is
to determine an optimal static horizon year design using a xed
charge transshipment problem formulation (FNCP). The branch-
and-bound algorithm is then employed in the formulation which
also includes explicit modeling of xed charge and variable cost
components for improved accuracy.
An efcient algorithm for the static investment planning of
large radial distribution systems was presented by Fawzi et al.
[8]. The algorithm takes into account the xed costs, concave non-
linearities in the cost functions of all elements and the operational
constraints. It uses a concave xed cost model to represent
elements with large xed costs (substations and possibly some
feeders) while it assumes linear cost functions for the remaining
elements. The branch-and-bound algorithm was also used with
the bounding criteria dependent on the cost and operational con-
straints. An iterative procedure modifying the solution of the rst
step accounts for the xed costs of the remaining elements.
Sharif et al. [22] propose an approach which makes use of both
the mixed-integer linear programming (MIP) and spanning tree
methods to help estimate the future expansion paths for a radial
distribution network. The general methodology employed is to rst
generate the spanning trees to connect the source to the demand
nodes followed by a MIP approach to ascertain the spanning tree
which can be used as the optimal solution.
Modeling the planning of distribution substations as well as
methods for solving distribution planning problems are discussed
in [7,16]. The solution methodology adopted is that of maximizing
the Lagrangian dual using the NETOPT program. Yahav and Oron
[27] accounted for the nonlinear costs of losses and construction
through the solution of a nonlinear program using an off-the-shelf
solver called GINO (it uses the generalized reduced gradient solver
GRG2 by Lasdon and Waren [19]).
Boulaxis et al. [3] proposes a new algorithm for solving the opti-
mal feeder routing problem using the dynamic programming tech-
nique and GIS facilities. Some of the factors taken into
consideration include practical issues such as cost parameters
(investments, line losses, reliability), technical constraints (voltage
drops and thermal limits) and physical routing constraints (obsta-
cles, high-cost passages, existing line sections). In related work, Lin
et al. [14] determine substation locations and new feeders using
GIS data. This work seems closest to ours in motivation. While
the work by Lin et al. is far more detailed in terms of modeling
the distribution systems, our work allows far more generality
and proposes a exible optimization framework for determining
new locations and feeders. Related work pertaining to planning
of distribution systems is seen in [9,6].
Finally, Murray and Shanbhag [17] use a local relaxation ap-
proach for the problem of siting electrical substations. Through
the solution of a sequence of appropriately dened nonlinear pro-
grams, the authors are able to solve large-scale problems over
thousands of grid-points. Importantly, for cases where global solu-
tions are available, the algorithm compares favorably. Related
extensions of the work to nonlinear facility location problems is
found in [18]. The nonlinear facility location problem falls within
the realm of mixed-integer nonlinear programming problems, a
class of optimization problems that have signicant application
in distribution system planning. For instance, reconguration via
capacitor allocation has been modeled as a mixed-integer nonlin-
ear program by Oliveira et al. [5]. Here, as well a continuous func-
tion is employed for smoothing the discrete variables and a primal-
dual interior point scheme is utilized for solving the resulting
smoothed problem.
Finally, its worth noting that the presence of distributed gener-
ation has introduced both novel and challenging questions in the
context of distribution system design. Popovic et al. [20] consider
the question of siting distributed generation by conducting a sen-
sitivity analysis on the power ow equations. More recently, Ghosh
et al. [10] employ Newton-based iterative schemes for optimal siz-
ing and placement of such generation and test their schemes on a
modied set of IEEE 6, 14 and 30 bus systems.
2. Methodology
Our general approach begins with data acquisition. This step
takes image data and provides a load distribution over a prescribed
grid. Integral to such a transformation is the prescription of how
much load corresponds to a specic residential, commercial or
2 S. Lakhera et al. / Electrical Power and Energy Systems xxx (2010) xxxxxx
Please cite this article in press as: Lakhera S et al. Approximating electrical distribution networks via mixed-integer nonlinear programming. Int J Electr
Power Energ Syst (2010), doi:10.1016/j.ijepes.2010.08.020
industrial parcel. Using the grid-based load distribution, we con-
struct an evolutionary pattern that relates how the load has
evolved historically. This would require specifying a set of vectors
that represent the load distribution over specic time epochs.
At a particular time epoch, we solve a design subproblem that
determines the optimal lines and substations to meet the load
appearing in the next time epoch. Such a subproblem is in general
a large-scale nonlinear integer programming problem. In the sub-
sequent sections, we provide some insights of how local solutions
of such problems may be obtained. Given the location of substa-
tions with prescribed capacities, we may solve a full-rank set of
equations to obtain the appropriate ows. This set of ows, if fea-
sible , is taken as a proxy for the distribution system linkages. If
this system is infeasible with respect to substation capacities, then
we proceed to solve a modied problem, till we attain a feasible
integer solution. In the next three subsections, we describe various
aspects of this approach in greater detail.
2.1. Data acquisition
Our test data is the tax parcel data set of the city of Champaign.
It provides information about the land usage which can be catego-
rized into three major groups: residential commercial and indus-
trial. We further subdivided the categories as shown in Fig. 1.
Based on the given data, we mapped the land use information
with the energy consumption information for each type of land
usage (Table 4). Fig. 2 provides the electrical load distribution pro-
le. In reality, the building of a distribution system is an inherently
stage-wise activity that corresponds closely with the evolution of
the urban load prole. Using some prior information about the
evolution pattern, we determine a distribution system design. An
evolution pattern is an order over the set of subgrids (in terms of
age for instance). In reality, nding exact evolution patterns is
challenging. and may require study of historical city development
plan. Fig. 3 gives an example of such evolution pattern and its real-
ization on a grid. The schematic on the right provides an order for
the appearance of the subgrids.
To make a robust decision, one should be able to design over a
family of such evolution patterns for a city. The determination of
an appropriate evolution pattern is out of scope of this study, but
one option could be the change in the historical boundary of the
city. Even if we have this information, it is difcult to estimate
the historical electrical consumption. One option to over come this
would be to take a factor of current load distribution to represent
historical load prole. We use a similar approach for our case study
for the city of Champaign. Once we have come is with one such
evolution pattern we can nd out a evolving distribution system,
we show this in Fig. 4. It can be seen that in this schematic, the grid
design evolves in accordance with the stage-wise appearance of
new load.
2.2. Problem formulation
In this section, we formulate the optimization problem under
the assumption that the load information has been made available
from the GIS data set. Such an optimization problem species the
optimal location of substations and the related ows. In formulat-
ing the problem, several concepts need clarifying. First, the rele-
vant set of equations (namely the load-ow equations) are
discussed. This is followed by a brief discussion of the structure
of admittance matrix of the grid. Finally, an important complexity
in our setting is the requirement that we address restricted (or no-
go) areas. One approach for modeling such areas is provided.
2.2.1. Load-ow equations
Given a set of substation locations, nodal loads and a well-de-
ned branch impedance matrix, the load-ow equations provide
a unique solution vector of nodal voltages at nodes housing loads
and nodal currents at substation nodes. We denote the set of nodes
housing substations by SS while the set of load nodes is denoted by
L. Then the optimal ow is derived from the solution of the linear
equations dened by following load-ow equations:
Yv i 0
v
j
1; j 2 SS
i
j
l
j
j 2 L
X
j
i
j
0: 1
The rst equation relates the current vector i, the voltage vec-
tor v and the admittance matrix Y and originates from Kirchhoffs
constraint equations. Voltage at nodes j 2 SSis assumed to be 1 pu
(per-unit). Current at each node is given precisely by the load at
that node. For a given load distribution and set of substation
Fig. 1. Parcel data for city of Champaign (L) with superimposed grid (R).
S. Lakhera et al. / Electrical Power and Energy Systems xxx (2010) xxxxxx 3
Please cite this article in press as: Lakhera S et al. Approximating electrical distribution networks via mixed-integer nonlinear programming. Int J Electr
Power Energ Syst (2010), doi:10.1016/j.ijepes.2010.08.020
Fig. 2. Electrical load distribution for the city of Champaign (L) with 3-D plot (R).
Fig. 3. Evolution pattern of load distribution.
Fig. 4. Evolving distribution system.
4 S. Lakhera et al. / Electrical Power and Energy Systems xxx (2010) xxxxxx
Please cite this article in press as: Lakhera S et al. Approximating electrical distribution networks via mixed-integer nonlinear programming. Int J Electr
Power Energ Syst (2010), doi:10.1016/j.ijepes.2010.08.020
locations, we can solve this linear system of equations. In fact,
one can also show that this set of equations represents a square
nonsingular system (see [17]). Note that the notation of relevance
is summarized in Table 1.
2.2.2. The admittance matrix Y
As seen in Fig. 5, for an n n grid, every interior node is con-
nected to the four nodes while nodes on the boundaries have cor-
respondingly fewer connections. The admittance matrix Y captures
the relationship of a node with its neighbors. Each row of the rst
system of (1) can be thought of as a current-balance relationship in
which the current at a particular node is equal to the sum of the
current ows along the links connected to it. For an n n grid,
the admittance matrix Y can be represented as a symmetric block
tridiagonal matrix
Y
A
1
B
1
B
1
A
2
B
2
.
.
.
.
.
.
.
.
.
B
n2
A
n1
B
n1
B
n1
A
n
0
B
B
B
B
B
B
B
@
1
C
C
C
C
C
C
C
A
;
where the blocks A
j
,j = 1, . . . , n and B
j
,j = 1, . . . ,n 1, are tridiagonal
and diagonal in structure, respectively. Fig. 5 provides the sparsity
pattern for a 5 5 grid.
2.2.3. Restricted area
An important complexity seen in practical problems is that cer-
tain areas are not acceptable from the standpoint of building distri-
bution networks. Such areas are often referred to as restricted or
no-go areas. Our framework is exible enough to construct grids
that do not have connectivity in such areas. As a consequence,
the admittance matrix suffers some changes. In Fig. 6, we show
how a no-go area translates into a new grid. Note that any node
that remains unconnected is essentially removed from the problem
framework. Questions of infeasibility are removed by not allowing
the load to be positive at such points. Also shown in Fig. 6 is the
modied admittance matrix.
2.3. Design subproblem
The location of substations over a given grid with a specied
load distribution may be cast as a facility location problem, but
with an important difference. The cost function is nonlinear in nat-
ure, while the decision variables are integers, specically binary
Table 1
Variables and parameters.
Notation Description

j
Electrical load at node j
v
j
Voltage at node j
i
j
Current delivered at node j
Y Nodal admittance matrix (relates nodal current to nodal voltage)
z
j
Decision to install substation at node j
C
loss
Cost of losses
C
sub
Capital cost
S
cap
Substation capacity
SS Set of substation node
l Smoothing parameter
q Penalty parameter
Fig. 5. A 5 5 grid (L), nodal linkages (C) and admittance matrix Y (R).
Fig. 6. A 5 5 grid with restricted area (L) and admittance matrix Y (R).
S. Lakhera et al. / Electrical Power and Energy Systems xxx (2010) xxxxxx 5
Please cite this article in press as: Lakhera S et al. Approximating electrical distribution networks via mixed-integer nonlinear programming. Int J Electr
Power Energ Syst (2010), doi:10.1016/j.ijepes.2010.08.020
variables. Such problems are inordinately hard to solve and some
preliminary results using existing solvers have not been promising
[17]. In particular, for grid sizes beyond 20 20, the computational
effort is enormous and renders conventional integer programming
formulations useless.
Formally, the design subproblem is given by (INLP):
minimize
i;v;z
C
losses
v
T
Yv C
cap
e
T
z
Yv i 0
subject to v
L
1 v
L
z
j
6 v
j
6 1; 8 j
1 z
j

j
6 i
j
6 1 2z
j

j
z
j
S
cap
; 8 j
z
j
2 f0; 1g; 8 j;
where v
T
Yv represents the lifetime cost of losses while
C
cap
e
T
z C
cap
P
j
z
i
gives the cost of installing capacity. Note that C
cap
provides a ratio of the capacity cost to the cost of losses. The rst
constraint is the Kirchhoffs relationship between current and volt-
age. The second constraint prescribes voltage bounds at each node.
In particular, at a substation node, voltage is set to 1 while at other
node (namely load nodes), voltages are bounded between v
L
and 1
at a load node. The third constraint species current at non-substa-
tion nodes as being precisely the load and bounded by substation
capacity S
cap
at substation nodes. Finally, the decision variable per-
taining to the placement of a substation at node i, given by z
i
, is
specied to be binary.
3. A smoothing technique for binary nonlinear programs
As mentioned in the earlier section, the problem (INLP) is a
large nonlinear integer programming problem whose solvability
(to a global minimizer) is possible only for modest grid sizes. Given
that the problems could have arbitrary sizes, we relax the require-
ment of providing a global minimizer. Instead, we provide a local
solution with respect to a smoothed (and therefore continuous)
nonlinear program. Such ideas were employed by Murray and
Shanbhag [17] to obtain starting points but not for solving the ori-
ginal problem. Here, we use a different smoothing function, which
does have a close relationship with that in [17].
The smoothed nonlinear problem is given by NLP(l):
minimize
i;v;z
C
losses
v
T
Yv C
cap
X
j2N
2
1 e
lz
j
1

Yv i 0
subject to 1 z
j
v
L
z
j
6 v
j
6 1; 8 j
1 z
j

j
6 i
j
6 1 2z
j

j
z
j
S
cap
; 8 j
0 6 z
j
6 1; 8 j
where
C
cap
2
1 e
lz
j
1

represents the smoothed version of the step function representing
the installation of a substation and C
cap
is the ratio of the cost of
capital to that of losses. Finally, the relaxed decision variable per-
taining to the placement of a substation at node j is z
j
. The smooth-
ing of the step function using the exponential function is given by
Fig. 7:
One of the challenges in such a framework is that the relaxed
decision variables z
j
could be non-integral in nature. The likelihood
of such an event is reduced by adding a penalty function of the
form:
Pz; q
X
j2N
qz
j
1 z
j
;
which essentially penalizes non-integral decision variables. This
however has a disadvantage of introducing a large number of local
minimizers. The resulting optimization problem, nonetheless, is gi-
ven by NLP(l, q)
minimize
i;v;z
C
losses
v
T
Yv C
cap
X
j2N
2
1 e
lz
j
1

Pz; q
Yv i 0
subject to 1 z
j
v
L
z
j
6 v
j
6 1; 8 j
1 z
j
l
j
6 i
j
6 1 2z
j

j
z
j
S
cap
; 8 j
0 6 z
j
6 1; 8 j
Finally, the solution to this problem may still be non-integral. If
that is indeed the case, then it is not clear if the load-ow equa-
tions are satised. To avert such a possibility, we dene
Fig. 7. Smoothed step function for a variety of ls.
6 S. Lakhera et al. / Electrical Power and Energy Systems xxx (2010) xxxxxx
Please cite this article in press as: Lakhera S et al. Approximating electrical distribution networks via mixed-integer nonlinear programming. Int J Electr
Power Energ Syst (2010), doi:10.1016/j.ijepes.2010.08.020
^
z
i

z
i
; z
i
2 f0; 1g
1; z
i
P
0; z
i
<
8
>
<
>
:
: 2
The integer-valued ^z may then be used to solve the load-ow
equations. If the solution to the load-owequations is feasible with
respect to the substation capacity constraints, we have a feasible
integer solution to (INLP). If not, we tighten the capacity con-
straints by reducing S
cap
and repeat this procedure. The framework
is formalized in specifying Algorithm 1. Note that in the current
framework, both l and q are kept xed but update rules for these
parameters are part of ongoing work.
Algorithm 1. Penalization-smoothing method
initialization S
cap
, l, q;
termin : false, k : 1, a : 1;
while termin = false do
(z
k
, v
k
, i
k
) solve NLP(c, l; aS
cap
);
z
k
roundz
k
according to (2);
Let v
k
;

i
k
solve load-ow equations;
if

i
k

SS
6 S
cap
then
termin = true;
end
else
a : 0.9a
end
k = k + 1
end
4. Numerical results
This section will examine the workings of the suggested mod-
el as well as the underlying algorithms for obtaining solutions in
large-scale instances. In estimating the underlying distribution
system, we use the current ows from the substations as a basis.
In Section 4.1, we observe how changing the load distribution
leads to intuitively different ow patterns. A challenge faced in
such models is how one can deal with areas which cannot house
transmission lines. We prescribe one technique for addressing
precisely such a problem in Section 4.2 and provides some
examples to show the workings of the model In Section 4.3,
we provide some examples to show the suitability of our meth-
od for locating substations in the absence of such data. Finally,
in Section 4.4, we examine how the penalty-smoothing algo-
rithm works on a set of sample problems of steadily increasing
size. A comparison with the commercial solver CPLEX is also
given.
4.1. Load-ow models and their enhancements
In this section, we provide some preliminary results pertaining
to the estimation of distribution networks through the use of load-
ow equations. To aid in understanding the model, we consider a
relatively small area in which the substation location is known.
For instance, in Fig. 8, there is exactly one substation on the lower
left corner from which ow emerges. The positive ow species an
initial estimate of the distribution system. It is observed that the
ow from the substation ows symmetrically in the Northern
and Eastern directions. Now, consider a setting where a majority
of the load is situated in towards the top-left of the grid (North
of the substation). We model such a load by using a Gaussian load
distribution instead of a uniform load distribution. The resulting
ow pattern changes signicantly with a dominant ow pattern
towards the Northern direction to address the load requirements.
The new pattern is shown in Fig. 9. Further complexities in the load
distribution (as exemplied by adding two Gaussian distributions)
are handled well by the existing framework (see Fig. 10). Increas-
ing the scaling of the problem provides a richer characterization
of the underlying distribution system design. In Fig. 11, we use
the load distribution from our test data with a single substation.
The size of the grid is increased to 25 25 and the resulting ow
shows a concentration in the load areas and the connections be-
tween the load center and the substation locations.
4.2. Modeling of restricted areas
We have modeled restricted areas as Fig. 12 shows. Specically,
in an 8 8 grid, we introduce a 3 3 region that cannot house
transmission lines. We see that ow passes around the no-go areas,
precisely as required.
4.3. Optimal location of substations
This subsection provides some of the results of using such an
approach to locate substations. In most settings, we may have
either no data (or possibly incomplete data) pertaining to the sub-
stations in an existing area. The optimal placement of these substa-
tions would then specify the ow along the wires.
In each gure, we display the load distribution (negative) on the
left and the current ows on the right. In Fig. 13C, we have 10 10
grid with C
cap
and l set at 1. The load centers are at locations (3, 3),
(3, 8), (8, 3) and (8, 8). Since the cost of capacity is low with respect
to the cost of losses, it is optimal to locate a substation at each load
center and our smoothing approach nds what is conceivably the
global solution. Clearly, such a strategy is not optimal as the cost
of capital increases. This is observed in Fig. 13R as the number of
Fig. 8. (L) Uniform load distribution, (C) current ows and (R) nodal current distribution.
S. Lakhera et al. / Electrical Power and Energy Systems xxx (2010) xxxxxx 7
Please cite this article in press as: Lakhera S et al. Approximating electrical distribution networks via mixed-integer nonlinear programming. Int J Electr
Power Energ Syst (2010), doi:10.1016/j.ijepes.2010.08.020
Fig. 9. Gaussian load distribution.
Fig. 10. Multiple Gaussians load distribution.
Fig. 11. Test area load distribution with a single substation location.
Fig. 12. Flow distribution in a 8 8 grid having 3 3 restricted area.
8 S. Lakhera et al. / Electrical Power and Energy Systems xxx (2010) xxxxxx
Please cite this article in press as: Lakhera S et al. Approximating electrical distribution networks via mixed-integer nonlinear programming. Int J Electr
Power Energ Syst (2010), doi:10.1016/j.ijepes.2010.08.020
substations reduces to two as the cost of capital grows by a factor
of 10.
Next, we consider two scenarios (Figs. 14 and 15 in each of
which a 25 25 grid is employed with multiple Gaussian load dis-
tributions placed at a variety of locations. In each instance, the
smoothing approach leads to placement of the substations at the
load centers or near a clump of load centers.
4.4. Scalability of the algorithm
As noted in the earlier sections, crucial to the solution of the
problem is the ability to solve large-scale instances of the design
subproblem given that the grid sizes can be arbitrarily high. In this
section, we provide some details on several aspect of our smooth-
ing approach. First, we show how the computational effort scales
with grid size. However such techniques lead to a local minimizer
of a smoothed problem which need not be a global minimizer.
Obtaining global minimizers would require the use of cutting-
plane or branch-and-bound techniques which tend to scale poorly
with the size of the problem. To obtain such solutions, we use the
commercial solver CPLEX. Table 2 charts the performance of Algo-
rithm 1 for obtaining a feasible integer solution to the problem
NLP(l, q) where l and q are kept at values of 10 and 10,000,
respectively. Note that our approach for solving this problem
Fig. 13. 10 10 grid with load distribution (L), substation locations with low capital costs (C
cap
= 1) (C) and high capital costs (C
cap
= 10) (R).
Fig. 14. 25 25 grid with symmetrically placed Gaussian loads.
Fig. 15. 25 25 grid with asymmetrically placed Gaussian loads.
S. Lakhera et al. / Electrical Power and Energy Systems xxx (2010) xxxxxx 9
Please cite this article in press as: Lakhera S et al. Approximating electrical distribution networks via mixed-integer nonlinear programming. Int J Electr
Power Energ Syst (2010), doi:10.1016/j.ijepes.2010.08.020
requires solving NLP(l, q), given the original substation capacity.
From Table 2, several observations may be made:
1. The number of major and minor iterations is relatively insensi-
tive to the size of the grid. We use a sequential quadratic pro-
gramming algorithm (SNOPT) [11] for solving the smoothed
problem. Such an approach solves a sequence of quadratic pro-
gramming problems. The bottleneck in such an approach is the
solution of the quadratic programming subproblem. As the grid
size grows, the real challenge lies in solving such problems ef-
ciently. It can be seen that while the number of grid-points
grows by a factor of 25, the corresponding growth in the cpu-
time is far more modest.
2. An important concern that can possibly plague Algorithm 1 is
that the number of substations added can be rather large, while
ensuring feasibility of the resulting integer solutions. Our preli-
minary tests however show that the number of substations
added is slightly larger than that seen in the globally optimized
setting.
The dominant reason for developing an alternate approach for
solving the nonlinear integer programming problem is that exist-
ing commercial solvers for such problems obtain global solutions
but at enormous computational cost. Table 3 details the perfor-
mance of CPLEX for a set of steadily increasing grid sizes. An
immediate observation is the growth in all metrics of computa-
tional effort is the rapid exponential growth. For instance, if the
problem size grows by 10 (from 25 to 225), the CPU time grows
from 12 s to over 30,000 s. Fig. 16 compares the growth in effort
of the two solvers using the logarithmic scale. A closer look at
the nature of the solutions shows that while CPLEX always nds
a global solution with 4 substations, the smoothing method termi-
nates with 5 or 6 substations in all the tested cases.
5. Case study: city of Champaign
In this section, we use image data of the city of Champaign to
approximate the underlying distribution system. Importantly, we
employ a stage-wise algorithm that solves a smoothed problem
at each stage. Such an extension allows adherence to an underlying
temporal evolution of the load.
5.1. Obtaining the load distribution
To approximate the test area distribution system, we began by
approximating the load distribution by using the GIS parcel data
obtained the Champaign County Regional Planning Commission.
This data has information about the tax parcel polygons, parcel
boundary lines and subdivision polygons. The parcel data set pro-
vides information about the land, usually with some implication
for land ownership or land usage. By aggregating the land usage
into three major categories that is, residential commercial and
industrial. We further subdivided the categories. Based on the gi-
ven data, we mapped the land use information with the energy
consumption information provided in Appendix A. Using this map-
ping, we created an additional mapping between, land use infor-
mation and approximation of site energy consumption for
residential/commercial/industrial buildings.
5.2. Evolution pattern
We assumed an evolution of the distribution network corre-
sponds with the evolution of the test area. We based the estima-
tion of the city development as provided the Champaign County
Regional Planning Commission which maps the city boundaries
over the last 60 years at certain key points on the basis of old maps.
Using the GIS les provides, the evolution of the city is shown in
Fig. 17.
Using this evolutionary data, we constructed two sets of distri-
bution patterns. The rst test case used the existing set of substa-
tions which were sequentially introduced within the evolution.
The second test case assumed that such data was not available
and prescribed an optimal placement.
5.3. Test case 1: Partial substation information
In the rst test case, we assumed that information about the
substation provided in the image data. Using the load distribution,
and assuming a three stage evolution pattern we come us with fol-
lowing distribution system. Here we assumed that these substa-
tions have enough capacity to take care of any load requirements.
Fig. 18 shows the growth in load over the three stages. Given
the load in stage 1, the bottomschematic in Fig. 18 shows the ows
based on the grid placement. Using this set of lines as given and
new load emerging from the stage 2 expansion, the second sche-
matic in Fig. 18 shows how the lines may be extended. Note that
we may insert a variety of no-go areas at this instance. Here, we
merely introduce a grid of a specic size. Finally, in Fig. 19, we
Table 2
Performance of the smoothing approach.
Size Iteration Minor iteration Time Loss Substation
25 269 5338 1.03 0.0094 6
64 89 5314 0.72 0.0679 6
100 47 4790 0.72 0.0615 5
225 4 994 0.28 0.3780 5
289 44 4657 1.83 0.0531 5
400 8 2642 1.21 0.2252 6
625 11 5814 4.27 0.3137 5
Table 3
Performance of CPLEX.
Size Iteration Minor iteration Time Loss Substation
25 8159 30,596 12.20 0.107 4
64 87,937 361,307 168.98 0.740 4
100 627,456 2,614,568 2053.83 0.470 4
225 3,768,278 16,079,112 31742.30 1.034 4
Fig. 16. Time comparison.
10 S. Lakhera et al. / Electrical Power and Energy Systems xxx (2010) xxxxxx
Please cite this article in press as: Lakhera S et al. Approximating electrical distribution networks via mixed-integer nonlinear programming. Int J Electr
Power Energ Syst (2010), doi:10.1016/j.ijepes.2010.08.020
show how the designed distribution system compares with the ori-
ginal area.
5.4. Test case 2: No substation information
In the second test case, we assumed that no information about
the location of substations is provided. Therefore, the algorithm
must nd the best possible location of the substations and estimate
the distribution system. Using the load distribution, and by assum-
ing a three stage evolution pattern, our algorithm provides us with
the following distribution system. Here we assumed that these
substations have a specied capacity. The distribution system de-
sign varies signicantly with the earlier one. In particular, we
sequentially add substations of nite capacity and the algorithm
Fig. 17. Evolution of the city of Champaign.
Fig. 18. Load prole (top) and distribution system (bottom) across stages 1 (top), 2 (middle) and 3 (bottom) (with substation information).
S. Lakhera et al. / Electrical Power and Energy Systems xxx (2010) xxxxxx 11
Please cite this article in press as: Lakhera S et al. Approximating electrical distribution networks via mixed-integer nonlinear programming. Int J Electr
Power Energ Syst (2010), doi:10.1016/j.ijepes.2010.08.020
chooses to add several substations in certain areas that represent
load centers. Fig. 20 provides a schematic of how the system
evolves.
6. Summary
We have presented a framework for obtaining approximations
of utility networks by employing GIS images. Our basic framework
has relied on obtaining an electrical load distribution over a pre-
specied grid. In the presence of possibly incomplete substation
data and restricted areas, we present an optimization-based for-
mulation for obtaining ows on the network, the latter usable as
a proxy for the lines in the distribution system. Addressing large-
scale instances of the location and ow problem requires solvers
that scale well with the grid size. Unfortunately, these problems
are discrete nonlinear problems and existing solvers are capable
of solving problems no more than a few hundred variables. Instead,
we present a penalization-based smoothing scheme that scales
well with the grid size and provides solutions that are comparable
with the global solutions obtained by CPLEX. Our framework can
be seen to address a variety of load distributions and restrictions
on line placements. We conclude with a case study in which a
stage-wise approach is adopted to the planning problem.
Acknowledgements
The authors gratefully acknowledge the data provided by
Champaign County Regional Planning Commission and some preli-
minary effort by Kevin Waicekauskas.
Appendix A
See Table 4.
Fig. 19. Given location of substation.
Fig. 20. Load prole (top) and distribution system (bottom) across stages 1 (top), 2 (middle) and 3 (bottom) (without substation information).
12 S. Lakhera et al. / Electrical Power and Energy Systems xxx (2010) xxxxxx
Please cite this article in press as: Lakhera S et al. Approximating electrical distribution networks via mixed-integer nonlinear programming. Int J Electr
Power Energ Syst (2010), doi:10.1016/j.ijepes.2010.08.020
References
[1] Adams RN, Laughton MA. Static and time phased network synthesis. Proc IEE
1974;121(2):13947.
[2] Ahuja RK, Magnanti TL, Orlin JB. Network ows. Theory, algorithms, and
applications. Englewood Cliffs (NJ): Prentice Hall Inc.,; 1993.
[3] Boulaxis NG, Papadopoulos MP. Optimal feeder routing in distribution system
planning using dynamic programming technique and gis facilities. IEEE Trans
Power Deliv 2002;17:2427.
[4] Crawford DM, Holt SB. A mathematical optimization technique for locating and
sizing distribution substations and deriving their optimal service areas. IEEE
Trans Power Appar Syst 1975:2305.
[5] de Oliveira LW, Carneiro Jr S, de Oliveira EJ, Pereira JLR, Silva Jr IC, Costa JS.
Optimal reconguration and capacitor allocation in radial distribution systems
for energy losses minimization. Int J Electr Power Energy Syst
2010;32(8):8408.
[6] El-Fouly THM, Zeineldin HH, El-Saadany EF, Salama MMA. A new optimization
model for distribution substation siting, sizing, and timing. Int J Electr Power
Energy Syst 2008;30:30815.
[7] El-Kady MA. Computer-aided planning of distribution substation and primary
feeders. IEEE Trans Power Appar Syst 1984;PAS-103(6):118392.
[8] Fawzi TH, Ali KF, El-Sobki SM. A new planning model for distribution systems.
IEEE Trans Power Appar Syst 1983;PAS-102:30107.
[9] Farrag MA, El-Metwally MM, El-Bages MS. A new model for distribution
system planning. Int J Electr Power Energy Syst 1997;21(7):52331.
[10] Ghosh S, Ghoshal SP, Ghosh S. Optimal sizing and placement of distributed
generation in a network system. Int J Electr Power Energy Syst
2010;32(8):84956.
[11] Gill PE, Murray W, Saunders MA. SNOPT: an SQP algorithm for large-scale
constrained optimization. SIAM J Optim 2002;12(4):9791006.
[12] Hindi KS, Brameller A. Design of low-voltage distribution networks:a
mathematical programming method. Proc IEE 1977;124(1):548.
[13] Knight UGW. The logical design of electricity networks using linear
programming methods. Proc IEE 1970;117:211727.
[14] Lin MW, Tsay MT, Wu SW. Application of geographic information system for
substation and feeder planning. Int J Electr Power Energy Syst
1996;18(3):75183.
[15] Masud E. An interactive procedure for sizing and timing distribution
substations using optimization techniques. IEEE Trans Power Appar Syst
1974;93(5):12816.
[16] Marshall AC, Boffy TB, Green JR, Hague H. Optimal design of electricity
networks. Proc IEEE C 1991;138(1):6977.
[17] Murray W, Shanbhag UV. A local relaxation approach for the siting of electrical
substations. Comput Optim Appl 2006;33(1):749.
[18] Murray W, Shanbhag UV. A local relaxation method for nonlinear facility
location problems. In: Multiscale optimization methods and applications.
Nonconvex Optim Appl, vol. 82. New York: Springer; 2006.
[19] More JJ, Wright SJ. Optimization software guide. SIAM Phila 1993:879.
[20] Popovic DH, Greatbanks JA, Begovic M, Pregelj A. Placement of distributed
generators and reclosers for distribution network security and reliability. Int J
Electr Power Energy Syst 2005;27(56):398408.
[21] Sun DI, Farris DR, Cote PJ, Shoults RR, Chen MS. Optimal distribution substation
and primary feeder planning via the xed charge network formulation. IEEE
Trans Power Appar Syst 1982;PAS-101:6029.
[22] Sharif SS, Salamat MA, Vanelli A. Model for future expansion of radial
distribution networks using mixed integer programming. IEEE; 1996. p. 1525.
[23] Thompson GL, Wall DL. A branch and bound model for choosing optimal
substation locations. IEEE Trans Power Appar Syst 1981;PAS-100(5):26837.
[24] Willis HL, Tram H, Engel MV, Finley L. Optimization applications to power
distribution. IEEE Comput Appl Power 1995;10:127.
[25] Willis HL, Tram H, Engel MV, Finley L. Selecting and applying distribution
optimization method. IEEE Comput Appl Power 1996;1:127.
[26] Wall DL, Thompson GL, Northcote-Green JED. An optimization model for
planning radial distribution networks. IEEE Trans Power Appar Syst 1979;PAS-
98(3):10616.
[27] Yahav K, Oron G. Optimal locations of electrical substations in regional energy
supply systems. IEEE; 1996. p. 30710.
[28] Zhu JZ. Optimization of power systems operation. Wiley-IEEE Press; 2009.
Table 4
Land usage and energy consumption.
APROP Usage Btu per square foot VA per square foot
1000 Developer held residential lots 0.00 0.00
1100 Single family rental dwelling 115.10 36.20
1150 Owner/occupied single family dwelling 115.10 36.20
1200 Duplex rental dwelling 123.30 36.20
1250 Owner/occupied duplex dwelling 123.30 36.20
1300 Apartment 37 153.40 44.90
1350 Apartment 37 153.40 44.90
1400 Apartment 8 or more 131.50 38.60
1450 Apartment 8 or more dwelling units 131.50 38.60
1500 Group home/fraternities/sorority 123.30 36.20
1700 Mobile home parks 194.50 57.00
1800 Condominium rental 123.30 36.20
1850 Owner/occupied condominium dwelling 123.30 36.20
2000 Industrial use 2000.00 746.67
2100 Developer held industrial lots 0.00 0.00
3000 Commercial use 128.80 37.70
3050 Owner/occupied commercial use 128.80 37.70
3100 Developer held commercial lots 0.00 0.00
4000 Communications or utilities use 2000.00 746.67
5000 Hotels and motels use 126.60 37.10
6000 Property exempt from taxation 123.30 36.20
6001 Property exempt from taxation 123.30 36.20
6002 Property exempt from taxation 123.30 36.20
6003 Property exempt from taxation 123.30 36.20
6005 Property exempt from taxation 123.30 36.20
6006 Property exempt from taxation 123.30 36.20
6007 Property exempt from taxation 123.30 36.20
6800 Property partially exempt from taxation 123.30 36.20
6900 Veterans and fraternal organizations 123.30 36.20
7000 Land used as a commons area 0.00 0.00
7400 Open space valuation 0.00 0.00
7500 Open space valuation 0.00 0.00
8100 Agriculture (10+ acres) 5.00 5.00
8150 Agriculture use with owner occupied dwelling 6.00 6.00
S. Lakhera et al. / Electrical Power and Energy Systems xxx (2010) xxxxxx 13
Please cite this article in press as: Lakhera S et al. Approximating electrical distribution networks via mixed-integer nonlinear programming. Int J Electr
Power Energ Syst (2010), doi:10.1016/j.ijepes.2010.08.020