Sie sind auf Seite 1von 9

IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 22, NO.

2, MAY 2007 791

Optimal Distribution System Horizon


Planning–Part I: Formulation
Robert H. Fletcher, Senior Member, IEEE, and Kai Strunz

Abstract—This paper is the first of a two-part paper on optimal and horizon-year (20+ years) planning periods. Prior research
distribution system planning. The horizon distribution planning addressed short-range and expansion planning by partitioning
problem and optimal distribution system model formulation the total planning problem into a set of sub-problems handled
are described. The horizon planning mission is to minimize
future costs by determining optimal design parameters given by available methods and techniques of total cost minimization
assumptions about the future. Prior work addressed short-range and optimization [1].
and expansion planning of subsets or combinations of design Power distribution cost minimization research has included
parameters. The many distribution requirements and associated optimal substation and feeder expansion planning [2]–[5],
constraints inhibited an all-inclusive evaluation of total horizon location and size of substations and feeders [6], [7], secondary
design requirements for the 20+ year period. The proposed model
and optimization formulation provides a generalized horizon system design [8], selection of feeder conductors [9]–[12],
planning approach and introduce a fully functioning comprehen- feeder reconfiguration [13], substation and feeder reliability
sive horizon planning model using a perspective that encompasses evaluation and design [14]–[16], and primary and secondary
all necessary parameters and constraints. Parameters include: design [17].
substation and distribution transformer capacities; number, size, Optimization methods used in short-range and long-range
and lengths of distribution feeders and secondary conductors;
and primary voltage class. Optimal design voltage drops and planning problems have included a variety of approaches
reliability indices are determined. The horizon planning optimiza- ranging from single-period models [8], [9] to multi-period
tion application is described and solved in the second companion models [2], [3]. The single-period models solve for system
paper using continuous constrained nonlinear programming parameters for a specific point in time. The multi-period models
methods. The application is demonstrated with Snohomish PUD solve a series of single-period models treating each period as an
case studies.
expansion unit. The distribution planning models include linear
Index Terms—Circuit optimization, losses, minimization or nonlinear, deterministic or heuristic, continuous or discrete,
methods, optimization methods, power distribution economics, or mixed-integer mathematical techniques [1], [18].
power distribution lines, power distribution planning.
The cited planning approaches do not include simultaneously
primary and secondary systems except as demonstrated in [17].
I. INTRODUCTION Furthermore, the primary voltage class is not considered a deci-
sion variable in prior planning approaches. Moreover, the final
built-out result of multi-period models is not always the lowest
cost horizon-year result. The prior model formulations require
T HE electric power distribution industry was built with
planning practices and design standards developed over
the past 100 years that will not be adequate in the future.
large existing distribution data structures and geographical at-
tributes. The large data requirement and associated constraints
Considerable change in utility environments is expected in the inhibited an all-inclusive valuation of horizon-year design pa-
next 20 to 50 years including consumer demand for services, rameters.
improved reliability, changing consumer load characteristics, The new planning model being introduced produces an
higher marginal cost of losses, and increased distributed re- optimal single-period horizon-year design encompassing all
sources. The distribution utility planning mission is to provide distribution design requirements for primary and secondary
long-range cost effective, affordable, and reliable service, while systems. The model uses a feeder layout viewed as a tree-like
ensuring consumers’ service voltages and power quality are circuit serving a circular sector of a round or hexagonal service
within standard range. The key horizon strategic need is to link area of uniform load density [19]. This arrangement abridges
short-range investment, operations, and maintenance decisions the input requirements. It provides an optimal horizon-year
to a long-range view of consumer comprehensive needs and design for a comprehensive set of parameters by minimizing
optimal design. The distribution planning function can be di- total cost per consumer given future assumptions. The opti-
vided into short-range (1 to 4 years), long-range (5 to 20 years), mization design approach uses a direct search single-period
continuous nonlinear constrained optimization methodology,
Manuscript received March 7, 2006; revised August 11, 2006. Paper no.
which has been implemented in MatLab® computer software.
TPWRS-00105–2006. The proposed planning model is also useful to investigate the
R. H. Fletcher is with Snohomish County PUD No.1, Everett, WA 98206 USA interplay and sensitivities between large numbers of design
(e-mail:rhfletcher@snopud.com). variables for educational purposes.
K. Strunz is with the Department of Electrical Engineering, University of
Washington, Seattle, WA 98195 USA (e-mail:strunz@ee.washington.edu). The horizon planning problem is described in Section II. In
Digital Object Identifier 10.1109/TPWRS.2007.895173 Section III, the horizon planning model is described including
0885-8950/$25.00 © 2007 IEEE
792 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 22, NO. 2, MAY 2007

formulation of planning perspective, secondary systems, con- sizing, transformer sizing, number of feeders, primary voltage
sumer load characteristics, voltage drop and losses, system reli- class, and system design voltage drops.
ability, input horizon-year assumptions, decision and dependent The proposed horizon model will aid the determination of the
variables, and economic evaluation. The optimization method- horizon-year design used in short-range expansion models. Sev-
ology used by the horizon planning model is briefly described eral papers have introduced short-range expansion models using
in Section IV. a technique that can be best described as a pseudo dynamic al-
gorithm, Sun [21], Ramirez-Rosado and Gönen [4] and Vaziri
II. HORIZON PLANNING PROBLEM [3]. In their distribution expansion approaches, the end expan-
sion stage identifies the set of expansion elements to be used for
The distribution utility planning mission is to provide long-
the entire planning period. A series of single stage algorithms
range cost effective, affordable, and reliable service, while en-
are employed for the intermediate short-range stages. The al-
suring consumers’ service voltages and power quality are within
gorithm for the short-range stages is formulated to select only
standard range. Moreover, the horizon planning process char-
elements from the end stage design i.e., conductor sizes, number
acterizes the needs of the horizon consumer and provides the
of feeders, size and location of power transformers. The horizon
underlying design assumptions and general guidance for short-
planning model developed in this paper provides a comprehen-
range planning decisions.
sive set of optimal end stage design parameters that can be used
A. Horizon Plan Comparative Assessments to improve the distribution short-range expansion models refer-
enced above.
The horizon plan provides a view of the future with general-
ized characteristics given a set of assumptions about the future. III. HORIZON PLANNING MODEL
As alternative scenarios are identified for consumer load charac-
Electric utility distribution systems are composed of many
teristics, cost of energy, consumer density, and distributed gen-
electrical elements connected in series originating from a
eration application, the generalized impacts on the distribution
substation. A generalized circular approach is used to describe
system can be observed. Because the future is uncertain, it is
a typical horizon-year substation service area. This approach
necessary to evaluate horizon plans over a large number of pos-
allows the solution of an optimal design with continuous
sible futures. The proposed horizon planning model provides the
constrained nonlinear programming methods. The optimal
core analysis simulation used in the horizon planning process
horizon-year design solution varies depending on assumptions
to determine the approximate relative consumer cost and design
about the horizon consumer density, cost of energy, cost of
for each specific future scenario.
reliability impacts, and consumer load characteristics. The
proposed horizon planning model determines the optimal dis-
B. Spatial Forecast Evaluations
tance between substations, primary system voltage, substation
Spatial land-use forecasts prepared by municipal, county, capacity, and number of feeders per substation. In addition, the
state governments, or regional planning agencies, provide model determines the typical size of distribution transformer,
information about the future. With this information, long-range number of consumers, and secondary conductors per trans-
spatial electric load forecasts are performed by utilities to iden- former. The inclusion of secondary systems in the optimization
tify the amount of power that must be delivered, and where and model is necessary due to its impact on the total cost function
when it will be needed. Most local governmental jurisdictions by limiting the available primary voltage drop.
routinely develop a comprehensive long-range plan which Dependent variables, such as primary and secondary design
identifies the horizon consumer types and zone densities for voltage drops, are derived from decision variables and horizon-
each quarter mile section [20]. year assumption input variables. The dependent and decision
The horizon planning model identifies the distribution system variables, and input variables are used to formulate the cost ob-
impact of proposed changes in spatial land-use forecasts. jective function and system constraints. The dependent variables
Horizon planning is performed by considering generalized ho- help functionally describe the distribution system horizon de-
mogeneous assessments of medium load densities for suburban sign. The coincidental sum of the primary and secondary voltage
or urban residential areas with uniform load distribution. The drops is constrained by the acceptable voltage standard limit de-
consumer loads, dispersed generation, and demand side man- scribed by ANSI Standard C84.1.
agement are assumed as a combined load mix of represented
load patterns. A. Planning Perspective for Distribution Feeders
The proposed planning model produces an optimal horizon
C. Short-Range Expansion Planning plan for a comprehensive set of design parameters, given system
The horizon planning model is used to test underlying de- constraints, by minimizing total cost per consumer given as-
sign assumptions and provide general guidance and strategic sumptions about the future. The model allows flexibility in the
direction for short-range planning, and to assure that design application to consider alternative horizon assumptions. Fig. 1
standards do not conflict with a long-range view of the future. represents typical substation service areas and feeder topologies
Horizon planning ensures that short-range decisions have con- as shown for the Snohomish County PUD City of Everett.
tinued and lasting value by supporting an optimal low cost dis- In utility practice, each mature substation feeder layout is
tribution system. The horizon plan only needs enough detail to seen as a reasonable approximation of a polygon or circle.
describe the general design element attributes such as conductor The final saturated build-out generally includes some overlap
FLETCHER AND STRUNZ: OPTIMAL DISTRIBUTION SYSTEM HORIZON PLANNING—PART I 793

area is a circular sector defined as . Each lateral line


is composed of both three-phase and single-phase portions with
the ratio of lengths being defined by an input variable.
The feeder main loading is, therefore, nonuniform with the
distance to the feeder main load center at ,
which is the center of gravity for a circular sector. All lateral line
loading is uniform with equally spaced distribution transformers
and represented as a rectangle load area. The lateral load center
is at one-half of the lateral length. The lowest primary voltage
approximation in the substation area is assumed to be at the end
of the longest lateral line connected at the end of the feeder main.
The area load density LD is the product of assumed consumer
Fig. 1. Typical substation area topology.
density CD, consumer peak demand , and consumer coin-
cidence factor or . The substation
load is a product of load density LD and substation area
. The number of consumers per substation area is
a product of consumer density CD and substation area .
The area served by the feeder laterals can be viewed as two tri-
angular areas on either side of the feeder main. The laterals are
tracing a grid of parallel roads separated by an equal distance
. These triangular areas can be reassembled as a rectangle
area for the purpose of approximating the total length of laterals.
For an efficient design, and to avoid unnecessary overlap of
laterals, the total length of lateral lines per substation area
is only a function of substation radius , the lateral separation
Fig. 2. Horizon distribution system design topology. distance , and independent of the number of feeders . It
is assumed that a typical lateral has a length as ex-
perienced in practice with laterals having similar lengths along a
of circles as well as nonserved areas. In order to reduce input feeder main. It is observed that if the number of laterals per main
data requirements, while providing flexibility in the application is , then the total length of laterals per main becomes
of different horizon assumptions, the model is based on an . To avoid that the total length of lateral lines
ideal feeder layout that views feeders as being tree-like circuits per substation area depends on angle and so guarantee that the
serving a circular sector of a round or polygonal service area total length of lateral lines is only a function of and ,
[19]. it becomes necessary to apply an adjustment to the number of
The horizon-year service area is defined as a circular area of laterals per feeder as follows. Feeder service area and maximum
uniform load and consumers per square mile, and, henceforth, length of laterals are
referred to as substation area. The solution of the decision vari-
ables varies depending on horizon-year assumption input vari- (1)
ables. The model is designed to characterize a typical substation (2)
area in terms of its size and number of feeders. At the same time,
it includes a large number of decision variables, including sec- The area served by the laterals for each main is
ondary systems and primary voltage class.
In Fig. 2, a description of a typical four-feeder substation area (3)
is shown. Each feeder is configured with one main. The length
The adjusted number of laterals per main is then
of feeder mains and laterals are formulated as a function of the
radius of the substation circle and angle of feeder circular sector.
All feeders are assumed radial with laterals branching perpen- (4)
dicular to mains.
This circular substation area arrangement allows for ease in The total adjusted lateral line length per main is then
determining system reliability cost impacts, system operational
loss costs, element voltage drops, and equipment installation (5)
costs. The radial arrangement will allow network configuration
and operation with adjacent like substation areas as required for The length of laterals per substation area is
feeder and substation contingency backup.
The length of the feeder main is . The maximum lateral (6)
length is and laterals are separated by a constant dis-
tance . The number of feeders is . Each feeder circular Hence, the total length of lateral lines per substation depends
sector angle is where in rad. The feeder service only on and .
794 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 22, NO. 2, MAY 2007

Fig. 3. Maximum number of consumers per transformer.

B. Secondary Systems C. Consumer Load Characteristics

The horizon planning model includes distribution trans- The consumer load characteristics are defined by a homoge-
formers and associated secondary lines. Since all lateral line neous uniform consumer density CD, coincidence factor ,
loading is assumed uniformly distributed with equally spaced peak load , and consumer load factor . The consumer
transformers, the typical transformer size can be included expected annual energy consumption is derived as
as a decision variable. The maximum number of consumers .
allowed for a given distribution transformer size The distribution system elements include distribution trans-
is constrained by spacing between consumers on a lateral former, secondary line, lateral line, main line, and substation
line and the maximum allowed secondary length . As the power transformer. To determine the demand loading on each
number of consumers per transformer is altered, the number particular system element, the number of consumer demands
of consumers per secondary conductor and number of served by the element are totaled, and diversified by the el-
secondaries per transformer are also impacted, and can ement’s associative coincidence factor. For a system element
be treated as decision variables. The number of consumers per serving a group of consumers , the Snohomish County
transformer is given as . PUD empirical developed element coincidence factor is
The number of consumers per each lateral is given as .
The peak demand is defined for the following six system
(7) elements: substation power transformer , main feeder
, lateral three-phase line , lateral single-phase line
The spacing between consumers on a lateral line is , distribution transformer , and secondary line
. The peak demand for each element is determined by
(8)
(9)
Consumers are assumed connected to the end of secondaries of The application of distributed generation is accomplished by
length which ensures that the maximum secondary voltage adjustment to typical consumer load characteristics and .
drop is identified. This generalization provides for a conserva- The distribution element phase current is represented as a
tive approximation of the maximum secondary voltage drop and function of element load demand and element base line-to-
secondary losses. neutral voltage .
For various secondary maximum allowed lengths, Fig. 3 The peak responsibility factors used to discount voltage
shows a typical relationship observed between the maximum drops, load factors, and loss factors can be derived for any
number of consumers per transformer, consumer separation system element. Peak responsibility factor and load
distance along laterals , and maximum secondary length factor for a given set of and are defined as
based on Snohomish PUD historical installation practices.

From data depicted in Fig. 3, the maximum number of con- (10)


sumers per distribution transformer can be represented as a con-
straint equation in the horizon model as a continuous empirical The effective element loss factor is less than load factor
function of consumer separation along laterals and max- and greater than , and is user-defined as a function
imum allowable secondary length . of load factor, i.e., [19].
FLETCHER AND STRUNZ: OPTIMAL DISTRIBUTION SYSTEM HORIZON PLANNING—PART I 795

D. Voltage Drops and Losses element load centers can be represented by using methods
described in [19] as
Voltage drop relationships are formulated for only five dis-
tribution system elements: main feeder, lateral three-phase and
(15)
single-phase lines, distribution transformers, and secondary
lines. The voltage drop for the substation power transformer
is not considered. System loss relationships are formulated (16)
for all six system elements including the substation power
transformer. The secondary line element equivalent load center distance
1) Voltage Drop Formulation: The general form of voltage is located at the load receiving end of the secondary line,
drop in Volts across the system line element when carrying therefore, . The distribution transformer voltage
the load current can be approximated by drop is similarly derived with the term replaced by
the transformer series resistance.
(11) The total coincidental per unit voltage drop formula-
tion used by the planning model is determined as the sum of the
where and are line resistance and reactance per unit weighted voltage drops of the five series elements defined
length respectively, is the consumer load power factor angle, in Section III-D
and represents the system line element. The horizon planning
model assumes that the consumer reactive load and reactive
(17)
line losses are effectively compensated for reactive power at
the load centers, which is an attribute of an optimally designed
distribution system. Thus the distribution system consumer The primary system voltage drop is formulated as the
load power factor is unity. By considering this unity power coincidental sum of the main and lateral voltage drops only.
factor design criterion for all consumer loads, the value of The service voltage drop is formulated as the coinci-
is zero. Thus, resistance predominantly influences the element dental sum of the distribution transformer and longest secondary
voltage drop if the power factor angle is assumed negligible. voltage drop.
Therefore, the voltage drop across any distribution system line 2) Loss Formulation: In order to determine the peak loss in
element is given as each for the six series elements, it is necessary to find the equiv-
alent loss center distances to fairly represent the associated
(12) network topology [19]. The general form of the peak loss of el-
ement is then
The maximum voltage drop for a distribution system element
is determined as a line-to-neutral per unit voltage drop of the (18)
applied base primary or secondary voltage. The base voltages
used in the per unit system are either the primary line-to-neutral Given a triangle load area of uniform load density, the loss
voltage , or the secondary line-to-neutral voltage . center can be represented as 8/15 of the total length from the
In order to determine the voltage drop in each line element vertex as described in [19]. Since a circular section is closely
for associated element phase current , it is convenient to find approximated by a triangular area, the main feeder element loss
the equivalent load center distance for each line element to center assuming a circular section load distribution can be ap-
fairly represent the associated network topology [19]. The per proximated as
unit voltage drop across any line element then becomes
(19)

(13) The three-phase and single-phase lateral element loss centers


can be formulated as a function of . For example, the lat-
where is the line-to-neutral base voltage or eral single-phase line element loss center assuming a rectan-
. The main feeder line element load center distance is gular load area of uniform load density is given by
represented as the center of gravity of a circular section where
the radius is and arc angle is and given as (20)

(14) The secondary loss center is represented as . The


distribution transformer loss demand is similarly derived with
the term replaced by the transformer series resistance.
The horizon planning model assumes that lateral line elements
are composed of three-phase and single-phase sub-elements
E. System Reliability
connected in series. The factor is the ratio of the length
of three-phase lateral line element to total lateral line length Reliability assessment methods fall into two classes: simula-
. For the lateral line sub-element three-phase line cur- tion and analytical [22]. Simulation is the most flexible method,
rent and single-phase line current , the lateral but suffers in computation time and uncertainty of precision.
796 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 22, NO. 2, MAY 2007

TABLE I used by the model without any protection devices installed


DECISION VARIABLES WITH EXAMPLE BOUNDS along the main feeders. Given that is the main failures per
year per mile and is lateral failures per year per mile, then
for a typical lateral , the annual interruptions are given as a
continuous function by

(21)

The substation area annual consumer-interruptions is

(22)

The substation area SAIFI per year is then


Analytical methods can be further divided into network mod-
eling and Markov modeling. The horizon planning model ap- (23)
plies a network modeling approach [23].
Network modeling has been the most popular technique The horizon model assumes a similar derivation for SAIDI
for distribution system reliability analysis from the first EPRI using a network modeling approach assuming a radial feeder
programs to programs recently developed and used by major configuration with main fault isolation and with or without
utilities. This popularity stems from the generalization of the a backup tie switch [19], [23]. Depending on the number of
method, and the natural similarities between the network mod- main feeder protection devices and feeder backup capability,
eling and the distribution system topology. Network modeling the SAIDI formulation is a function of , , ,
can handle basic dependent events such as fault isolation and , , failure rates, fault repair times, and switch op-
load restoration required with horizon planning. eration durations. Network modeling formulations are derived
The optimal horizon model includes traditional generalized assuming fault isolation and main feeder backup from adjacent
radial network modeling. The design of the horizon model pro- like feeders served from adjacent substations. Alternate distri-
vides a framework to determine the optimal level of reliability bution feeder arrangements can be formulated and evaluated as
for a chosen protection and fault isolation scheme, and assump- needed depending on type of service area and utility’s system
tions about consumer sustained interruption costs of frequency protection standard practices [27]. All reliability formulations
and duration. are continuous.
The substation area reliability indices for sustained outages
SAIFI “System Average Interruption Frequency Index” and F. Decision Variables
SAIDI “…Duration Index” [24] are a function of the substation The selection of distribution system elements represented by
circular arrangement. For most utilities, these two indices can the proposed horizon planning model is adequate to define a fu-
adequately represent reliability for sustainable interruptions. ture standard design. These elements are termed decision vari-
The horizon model objective cost function includes the net ables and represented as vector , and listed in Table I. Included
present value of interruption costs per consumer. The annual are the example bound limits for each of the variables.
cost of interruption frequency and durations is a function of
the number of consumers , the reliability indices, and G. Horizon Planning Model Assumptions
user defined reliability impact costs [25]. The reliability indices
SAIFI and SAIDI are formulated as a continuous function of The horizon planning model allows some generalizations
user defined element annual failure rates, repair rates, and with which to characterize a typical horizon substation area.
manual/automatic sectionalizing times. These generalizations enable the inclusion of a large number of
The feeder layout shown in Fig. 2 is assumed with laterals decision variables. The decision variables, dependent variables,
protected with a single fuse or recloser at the connection with and input variables have a clear relationship. The circle perspec-
the main. Protection devices may be added in the main line. tive allows for a horizon abstraction of costs and distribution
Manually or automatically operated switches are located on design standards. All of the decision variables are assumed
the main line adjacent to each lateral connection, immediately continuous, which aids sensitivity analyses as detailed in [28].
downstream of each lateral connection, which provides a The model assumptions are listed in Table II.
two-stage switching restoration capability [26]. Each main is
H. Horizon-Year Assumption Input Variables
assumed to have a normally open tie switch for backup from
a similar adjacent substation main. All protection devices are The proposed planning model is capable of a wide range of
operated as nonreclose. The main backup tie switch can be utility applications for a variety of future assumptions. Modifi-
either enabled or disabled to assess reliability impacts. Feeder cations of fixed and variable cost functions must be performed to
fault isolation automation can be assessed by reducing main represent each utility’s own particular cost drivers and assump-
sectionalizing or backup tie switch operating times to zero. The tions about the future. The assumptions about the future include
following is the SAIFI formulation of the radial feeder layout comprehensive land use consumer load densities consumer and
FLETCHER AND STRUNZ: OPTIMAL DISTRIBUTION SYSTEM HORIZON PLANNING—PART I 797

TABLE II TABLE III


HORIZON PLANNING MODEL ASSUMPTIONS INPUT VARIABLES WITH EXAMPLE VALUES

load characteristics, costs of energy and consumer interruptions,


system parameters, transformer characteristics, load multipliers, TABLE IV
DEPENDENT VARIABLES AND FUNCTIONAL RELATIONSHIPS
economic factors, and reliability factors. Loading multipliers are
used to define the system element maximum loading constraint.
For example, the distribution transformer loading upper limit is
the transformer size times the load multiplier factor .
Table III lists horizon assumption data used to represent a dis-
tribution system urban area.

I. Dependent Variables
Dependent variables are derived from decision variables and
horizon-year assumption input variables. The dependent vari-
ables, decision variables, and input variables are used to formu-
late the cost objective function and system constraints. The de-
pendent variables functionally describe the distribution system
horizon design. Table IV lists significant dependent variables
derived with the functional relationships.

J. Net Present Value and Discounting the Future


Adequate planning can ensure that the future consumer cost represented by a uniform series present value factor [29] which
of service is minimized. Fixed and annual costs are represented is found in engineering economy textbooks
as a net present value (NPV) of life-cycle investment costs and
variable costs respectively. Annual costs are expenditures repre- (24)
sented by system loss cost and consumer reliability cost impacts
[25]. The NPV factor is a function of the assumed net dis- A critical component of the net present value analysis is the
count rate and equipment life-cycle duration in years and is assumed annual investment rate and inflation rates. A gen-
798 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 22, NO. 2, MAY 2007

eral inflation rate is applied to equipment life-cycle replace- [2] M. Vaziri, K. Tomsovic, A. Bose, and T. Gönen, “Distribution ex-
ment, and annual reliability cost impacts. Annual maintenance pansion problem: Formulation and practicality for a multistage glob-
ally optimal solution,” in Proc. IEEE Power Engineering Soc. Winter
is not addressed, and is assumed to be a constant fixed charge. Meeting Conf., Feb. 2001, vol. 3, pp. 1461–1466.
A separate inflation rate is applied to the marginal cost of en- [3] M. Vaziri, K. Tomsovic, and A. Bose, “A directed graph formulation
ergy. The effective discount rate is the investment rate of the multistage distribution expansion problem,” IEEE Trans. Power
Delivery, vol. 19, no. 3, pp. 1335–1341, Jul. 2004.
minus inflation rate or . [4] I. J. Ramirez-Rosado and T. Gönen, “Pseudo-dynamic planning for
expansion of power distribution systems,” IEEE Trans. Power Syst.,
vol. 6, no. 1, pp. 245–254, Feb. 1991.
IV. OPTIMIZATION METHODOLOGY [5] M. Blanchard, L. Delorme, C. Simard, and Y. Nadeau, “Experience
The optimal design parameters are determined using a with optimization software for distribution system planning,” IEEE
Trans. Power Syst., vol. 11, no. 4, pp. 1891–1898, Nov. 1996.
direct search single-period continuous nonlinear constrained [6] V. Parada, J. A. Ferland, M. Arias, and K. Daniels, “Optimization of
optimization methodology [18], [30], which has been imple- electric distribution feeders using simulated annealing,” IEEE Trans
Power Delivery, vol. 19, pp. 1135–1141, Jul. 2004.
mented in MatLab® computer software. The solution requires [7] M. S. Carneiro, P. M. Franca, and P. D. Silverira, “Long-range planning
minimization of a single cost objective function relating infra- of power distribution systems: Primary networks,” Elect. Power Syst.
structure installations and operation cost as a function of the Res., vol. 27, no. 3, pp. 233–231, Aug. 1993.
[8] ——, “Long-range planning of power distribution systems: Secondary
decision variables and user supplied service area characteris- networks,” Comput. Elect. Eng., vol. 22, no. 3, pp. 179–191, May 1996.
tics. The horizon-year system consists of nonlinear inequality [9] S. Manda and A. Pahwa, “Optimal selection of conductors for distri-
constraints, such as voltage limits, primary and secondary bution feeders,” IEEE Trans. Power Syst., vol. 17, no. 1, pp. 192–197,
conductor ampacity, and transformer loading limits. All of Feb. 2002.
[10] H. N. Tram and D. L. Wall, “Optimal conductor selection in planning
the decision and dependent variables, objective function, and radial distribution systems,” IEEE Trans. Power Syst., vol. 3, no. 1, pp.
constraint equations are formulated as continuous. The cost 200–206, Feb. 1988.
objective function is divided into two main parts: fixed invest- [11] R. Ranjan, A. Chaturvedi, P. S. Solanki, and D. Das, “Optimal
conductor selection of radial distribution feeders using evolution pro-
ment and variable operation. The constraint equations describe gramming,” in Proc. TENCON 2003 Conf. Convergent Technologies
system element limits for loading, total system voltage drop, for Asia-Pacific Region, Oct. 2003, vol. 1, pp. 456–459.
[12] Z. Wang, H. Liu, D. C. Yu, X. Wang, and H. Song, “A practical ap-
and physical constraints. proach to the conductor size selection in planning radial distribution
system,” IEEE Trans. Power Delivery, vol. 15, no. 1, pp. 350–354, Jan.
V. CONCLUSIONS 2000.
[13] E. Diaz-Dorado, J. Cidrás, and E. Miguez, “Application of evolutionary
In this, the first of a two-paper series, the horizon planning algorithms for the planning of urban distribution networks of medium
voltage,” IEEE Trans. Power Syst., vol. 17, no. 3, pp. 879–884, Aug.
model formulation and optimization approach is described. The 2002.
value of horizon planning is the ability to identify the general [14] Y. Tang, “Power distribution system planning with reliability mod-
electrical power needs of the future consumer and approximate eling and optimization,” IEEE Trans. Power Syst., vol. 11, no. 1, pp.
181–189, Feb. 1996.
the lowest cost distribution system infrastructure design re- [15] L. Goel and R. Billinton, “Determination of reliability worth for dis-
quired to serve the future consumer. Distribution horizon plans tribution system planning,” IEEE Trans. Power Delivery, vol. 9, no. 3,
provide the underlying design assumptions and general guid- pp. 1577–1583, Jul. 1994.
[16] R. Billinton, “Economic cost of non-supply,” in Proc. Power Eng. Soc.
ance for short-range planning decisions. The horizon planning Winter Meeting Conf., Feb. 2002, vol. 2, pp. 959–962.
model approach will enable determination of a complete set of [17] P. C. Paiva, H. M. Khodr, J. A. Domínguez-Navarro, J. M. Yusta, and
optimal minimum cost future designs. A. J. Urdaneta, “Integral planning of primary – Secondary distribution
system using mixed integer linear programming,” IEEE Trans. Power
The main scientific contributions of the horizon planning Syst., vol. 20, no. 2, pp. 1134–1143, May 2004.
model are shown to be three-fold. First, the advantages of [18] N. S. Rau, Practical Optimization Methods. New York: IEEE Press/
adopting a horizon planning philosophy and embracing the Wiley, 2003.
[19] T. Gönen, Electric Power Distribution System Engineering. New
need to identify long-range distribution system impacts are York: McGraw-Hill, Inc., 1986.
given. Second, the model provides both engineering formula- [20] 10 Year Update Comprehensive Plan Snohomish County, WA, 2005,
tion and economic evaluation approach of a circular substation The Future Long-Range Map [Online]. Available: www.co.sno-
homish.wa.us
area perspective. By allowing generalizations, the model can [21] D. I. Sun, E. R. Farris, P. J. Cote, R. R. Shoults, and M. S. Chen, “Op-
facilitate the solution of a comprehensive set of horizon-year timal distribution substation and primary feeder planning via the fixed
decision variables that include design parameters for primary charge network formulation,” IEEE Trans. Power Apparat. Syst., vol.
101, no. 3, pp. 602–609, Mar. 1982.
and secondary lines, and substation and distribution trans- [22] R. E. Brown, S. Gupta, R. D. Christie, S. S. Venkata, and R. Fletcher,
formers. This comprehensiveness uniquely distinguishes it “Distribution system reliability assessment using hierarchical Markov
from prior formulations. Third, the formulation is such that the modeling,” IEEE Trans. Power Delivery, vol. 11, no. 4, pp. 1929–1934,
Oct. 1996.
model is compatible with continuous nonlinear gradient search [23] R. N. Allan, E. N. Dialynas, and I. R. Homer, “Modeling and evaluating
methods. The horizon planning model approach can lead to a the reliability of distribution systems,” IEEE Trans. Power Apparat.
more efficient utilization of the distribution system in the long Syst., vol. 98, no. 6, pp. 2181–2189, Dec. 1979.
[24] IEEE Guide for Electric Power Distribution Reliability Indices, , May
term. 2004, IEEE Standard 1366-2003.
[25] R. F. Ghajar and R. Billinton, “Economic costs of power interruptions:
A consistent model and methodology,” J. Elect. Power Energy Syst.,
REFERENCES vol. 28, no. 1, pp. 29–35, Jan. 2006.
[1] S. K. Khator and L. C. Leung, “Power distribution planning: A review [26] R. E. Brown and A. P. Hanson, “Impact of two-stage service restoration
of models and issues,” IEEE Trans. Power Syst., vol. 12, no. 3, pp. on distribution reliability,” IEEE Trans. Power Syst., vol. 16, no. 4, pp.
1151–1159, Aug. 1997. 624–629, Nov. 2001.
FLETCHER AND STRUNZ: OPTIMAL DISTRIBUTION SYSTEM HORIZON PLANNING—PART I 799

[27] Y. Mao and K. N. Miu, “Switch placement to improve system relia- Kai Strunz received the Dipl.-Ing. degree from the
bility for radial distribution systems with distributed generation,” IEEE University of Saarland, Saarbrucken, Germany, in
Trans. Power Syst., vol. 18, no. 4, pp. 1346–1352, Nov. 2003. 1996, and the Dr.-Ing. degree with (summa cum
[28] R. H. Fletcher and K. Strunz, “Optimal distribution system horizon laude) from the same university in 2001.
planning part II: Application,” IEEE Trans. Power Syst., 2006. Currently, he is an Assistant Professor with the
[29] D. E. Grant and W. G. Ireson, Principles of Engineering Economy, 4th University of Washington, Seattle. He has pur-
ed. New York: Ronald Press, 1960. sued research at Brunel University, London, U.K.,
[30] Optimization Toolbox for Use with MATLAB® The Math Works, Inc., working with the National Grid Company in the
Natick, MA, 2004 [Online]. Available: http://www.mathworks.com fields of power system stabilization, electromechan-
ical transients, and power system modeling. He was
Robert H. Fletcher (M’79–SM’04) received the with the Division Recherche et Developpement of
B.Sc., M.Sc., and Ph.D. degrees in electrical engi- Electricite de France (EDF), Paris, France, where his main research work was
neering from the University of Washington, Seattle. the creation of efficient numerical methods for real time digital simulation.
He is a Principal Engineer in System Planning Dr. Strunz received the national Science Foundation CAREER award in 2003
and Protection with Snohomish County Public and the Outstanding Teaching Award of the Department of Electrical Engi-
Utility District No.1, Everett, WA. He has over 38 neering at University of Washington in 2004.
years of electric utility industry experience as an
Electrical Engineer. He is an author or co-author of
several papers on the subject of distribution system
efficiency and reliability. He is currently engaged
in conservation voltage regulation and system effi-
ciency research. He is Technical Chair of the Distribution System Efficiency
Initiative (DSEI) conducted by the Northwest Energy Efficiency Alliance
involving 20 electric utilities.
Mr. Fletcher is a member of the IEEE PSACE Committee, RRPA Subcom-
mittee, and Chair of the PSPI Distribution Planning WG. He is a registered Pro-
fessional Engineer in the States of Washington and Oregon.

Das könnte Ihnen auch gefallen