Electrical Power and Energy Systems 33 (2011) 288–295

Contents lists available at ScienceDirect

Electrical Power and Energy Systems

journal homepage: www.elsevier.com/locate/ijepes

DG allocation with application of dynamic programming for loss reduction and

reliability improvement
N. Khalesi a, N. Rezaei a, M.-R. Haghifam b,⇑
a
Department of Electrical Engineering, Islamic Azad University, Tehran South Branch, Tehran, Iran
b
Department of Electrical Engineering, Tarbiat Modares University, Tehran, Iran

a r t i c l e i n f o a b s t r a c t

Article history: Distribution system companies intend to supply electricity to its customers in an economical and reliable
Received 8 December 2009 manner whereas customers in most distribution system are outspread and connect to distribution system
Received in revised form 2 July 2010 with different type of equipments. These equipment usually have various types and resistance together,
Accepted 13 August 2010
that produce highest loss and lowest reliability for distribution systems and customers that are not
appreciated in networks. Distributed generations (DGs) are one of the best reliable solutions for these
problems if they are allocated appropriately in the distribution system. This paper presents multi-objec-
Keywords:
tive function to determine the optimal locations to place DGs in distribution system to minimize power
Distributed generation
Dynamic programming
loss of the system and enhance reliability improvement and voltage profile. Time varying load is applied
Loss reduction in this optimization to reach pragmatic results meanwhile all of the study and their requirement are
Reliability improvement based on cost/benefit forms. Finally to solve this multi-objective problem a novel approach based on
Time varying load dynamic programming is used. The proposed methodology is successfully applied to a study case and
Voltage profile simulation results are reported to verify the proposed approach.
Ó 2010 Elsevier Ltd. All rights reserved.

1. Introduction power loss. Authors in [1,2] solve the problem by analytical ap-
proach [3], employs non-linear programming [4], uses combina-
Distribution system planners endeavor to supply economical tion of genetic algorithm and simulated annealing [5,6], present
and reliable electricity to customers. It is important to design, oper- genetic algorithm [7], submits tabu search method and [8] uses
ate and maintain reliable power systems with lowest cost and high- fuzzy approach for optimization of its algorithm [9,10], apply load
est benefit. Reliability improvement and loss reduction are two flow approaches [11], uses sequential optimization and [12] uses
important goals for electrical distribution companies. These compa- heuristic approach.
nies follow, consider and test a lot of technologies, optimization All papers presented in [1–12] deal important problems and
programs, etc. to bring above economic benefits and provide elec- weaknesses that are listed on below mentioned clauses:
tricity with high quality and reliability and prevent interruptions
in system because cost of interruptions and power outages can re-  All the simulations performed in [1–12] address a static load
sult severe economic impact on utility and customers. condition. Objective function optimization based on a single
With recent advances in technology, use of distributed genera- load point, such as the peak load, may not provide reliable
tion (DG) in the power distribution system can provide the most results.
economical solution and keep network in proper situation. A lot  Reliability aspects in above mentioned papers are not consid-
of Papers and studies have been carried out in recent years to pres- ered while applying DGs to a distribution system can contribute
ent methodologies in DG placement and sizing. to improving system reliability.
One of the criteria to search the optimal DG allocation is mini-  DG placement in network has not been considered with evalu-
mizing power loss or reliability improvement. ating reliability and loss at the same time.
Several papers have been published that address the use of arti-
ficial intelligence algorithms, analytical approaches or load flow Also some papers have appreciated approaches in their method-
approaches to optimize DG placement [1–12] based on minimizing ologies like [13], but considering static load condition in their con-
cepts may not lead to satisfactory results.
This paper tries to overcome above mentioned weakness and
⇑ Corresponding author. Tel./fax: +98 21 82884347.
proposes a novel algorithm to optimize objective function. To
E-mail addresses: [email protected], [email protected] (M.-R.
Haghifam).
follow this proper purpose, first time-varying loads are taken into

0142-0615/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijepes.2010.08.024
 N. Khalesi et al. / Electrical Power and Energy Systems 33 (2011) 288–295 289

account then multi-objective function are considered based on a inquiry and renovation cost. This cost is not related to placement
cost/benefit form that enhance benefits of DG allocation in system of DG and is equal for all DG placements. This cost can be evaluated
to compensate system loss, system reliability and cost of pur- by:
chased power from transmission line along the planning period. Fi-
X
NDG X
KDG
nally; to solve this multi-objective problem a novel approach based C2 ¼ Costmain;ik ð2Þ
on dynamic programming is used. In addition DGs are considered i¼1 k¼1
as constant power source such as photo cells, fuel cells or gas gen-
Present worth value of this annual cost with considering infla-
erators. Also in this paper, purchased active power price from
tion rate and interest rate [14] in planning period is calculated
transmission grid varies in different time of day and also cost of en-
below:
ergy not supplied for different customers (residential, commercial
and industrial) varies in different time of the day. T 
X t
1 þ InfR
In the following sections, load modeling is presented in Section CPVðC 2 Þ ¼ C 2 ð3Þ
t¼1
1 þ IntR
2, mathematical formulation is explained in Section 3, and objec-
tive function is submitted in Section 4, dynamic programming
method is illustrated in Section 5 and a case study is reported in 3.1.3. Operation cost of DG
Section 6. Finally, the conclusions of the paper are summarized Since distributed generation shall trace load demands therefore
in Section 7. it is required to have cost for its input source hence operation cost
is equivalent to fuel cost. This cost and its present worth value are
2. Load modeling evaluated by:
X
NDG X
KDG
Accurate optimization of objective function is resulted based on C3 ¼ T J  DGJ;ik  CGik ð4Þ
input data and correct analysis of this data. One important data is i¼1 k¼1

definition of load pattern. Distribution system load varies in differ- T 

X t
1 þ InfR
ent time of day, therefore in this paper, load condition is consid- CPVðC 3 Þ ¼ C 3 ð5Þ
t¼1
1 þ IntR
ered in three stages (light, medium and peak load). Passed time
in these three stages is registered and maximum load consumption where NDG: number of DG unit installed in network; KDG: capacity
in each load point is considered as input data for DG allocation of DG from 1 to 5 MW; Costinv: investment cost of DG sources ($/
algorithm. MW); Costmain: maintenance cost of DG ($/MW-year); DGJ,ik: Gener-
Table 1 shows abbreviations of above mentioned descriptions. ated power by DG source installed in network in identified load le-
‘‘A” parameter in second column of Table 1 presents percentage vel (MW); CGik: operation cost of DG sources ($/MW h); IntR: the
of minimum load to maximum load of network and other param- interest rate; InfR: the inflation rate; CPV(): cost present worth;
eters present situations of the network. TJ: passing time (h/year)

3. Mathematical model formulation 3.2. DG benefits evaluation

In this section, economical benefits and DG application costs are 3.2.1. Active power demand reduction from transmission line
submitted and modeled. In this modeling, distributions system In power system restructuring, electric utility distribution com-
companies are responsible for providing customer demand, DG pany purchases its power demand from transmission grid. Portion
operation and distribution system management. All of these of this power demand is for distribution system customers and an-
responsibilities are based on cost reduction and improving quality other one is spent in line and equipment loss. This power demand
and reliability of customer service. Therefore costs and benefits of is evaluated by:
DG allocation in network can be expressed as follows. PTNDG;J ¼ PDJ þ LossNDG;J ð6Þ

3.1. DG costs evaluation Distribution Company can supply portion of its power demand
with considering DG in network and gets lower electric power
3.1.1. Investment cost from transmission grid. In this case electric power demand is cal-
The cost of DG unit, investigation fee, site preparing for DG culated as below:
installation, construction, monitoring equipment, etc. are included X
NDG X Nloc 
KDG X 
in investment cost. These costs can be formulated as following PTDG;J ¼ PDJ  DGJ;ik þ LossDG
J;ikl ð7Þ
equation. i¼1 k¼1 L¼1

X
NDG X
KDG Therefore reduction of active power demand can be formulated
C1 ¼ Costinv;iK ð1Þ as following equation:
i¼1 K¼1
DPT ¼ PTNDG;J  PTDG;J
X
NDG X Nloc 
KDG X 
3.1.2. Maintenance cost
¼ LossNDG;J þ DGJ;ik  LossDG
J;ikl ð8Þ
Another yearly cost of DG allocation relates to maintenance
i¼1 k¼1 L¼1
cost. Maintenance cost includes annual mechanical and electrical

Table 1
And loss reduction based on presence of DG is evaluated by:
Load characteristic levels. X
NDG X Nloc 
KDG X 
Load level (J) Percentage of Network condition Passing time
DLossJ;ikl ¼ LossNDG;J  LossDG
J;ikl ð9Þ
peak load (%) (h/year) i¼1 k¼1 L¼1

1 A–B Light load T1 Therefore Eq. (8) can be formulated as following equation:
2 B–C Medium load T2
3 C-100 Peak load T3 DPT ¼ PTNDG;J  PTDG;J ¼ DGJ;ik þ DLossJ;ikl ð10Þ
290 N. Khalesi et al. / Electrical Power and Energy Systems 33 (2011) 288–295

Based on above mentioned notes, active power reduction bene- for each year that Distribution Company can reach is expressed
fit for each year that Distribution Company can achieve is evalu- by Eq. (14):
ated by:
B2 ¼ C ENS  C ENS;DG ð14Þ
X
3
Present worth value of (14) is calculated below:
B1 ¼ C MWh;J  DPT  T J ð11Þ
J¼1 T 
X t
1 þ InfR
BPVðB2 Þ ¼ B2 ð15Þ
Present worth value of (11) is calculated below: t¼1
1 þ IntR
T 
X t where Nb: number of branches in the network; kb: branches failure
1 þ InfR
BPVðB1 Þ ¼ B1 ð12Þ rate (f/km-year); Lb: branch length (km); Cint,J: price of energy not
t¼1
1 þ IntR
supply in load level J. ($/MW h); Nres: number of nodes isolated dur-
where BPV (): benefit present worth; PTNDG,J: active power demand ing fault location; Nrep: number of nodes isolated during fault re-
when not considering DG in network (MW); PDJ: customer active pair; Pres: loads are restored during fault; Prep: loads are not
power demand (MW); LossNDG,J: system loss when not considering restored during fault; tres: duration of the fault location and switch-
DG in network (MW); PTDG,J: active power demand when consider- ing time; trep: duration of the fault repair; CEquip,J: cost of energy not
ing DG in network (MW); LossDG supply based on failure in equipments except of branches ($); CENS:
J;ikl : system loss when considering DG
in network (MW); CMWh,J: energy market price in load level J ($/ cost of energy not supply without DG ($); CENS,DG: cost of energy not
MW h). supply with DG ($).
It shall be noted that active power price is variant in different It shall be noted that price of energy not supply is not equal for
hours. This paper uses multi-level model for electricity price that different customer and different time of the day [16]. In this paper
is function of active power receiving from transmission grid. load points are divided into three groups: residential loads, com-
Fig. 1 shows the proposed electric price. mercial loads and industrial loads. Fig. 2 presents how price of en-
ergy not supply is calculated in this study.
3.2.2. Reliability improvement
Some of reliability indices are fundamentally important but 4. Objective function
they do not always give a complete representation of system
behavior and response. In order to submit the importance of a sys- In conclusion, cost and benefit view points which have been de-
tem outage, energy not supplied index (ENS) is evaluated. This in- scribed in previous sections are considered in one unique objective
dex reflects total energy not supplied by the system due to faults function that formulated below:
during study period. This reliability analysis is implemented by Max Z ¼ Benefits  Costs
analytical approach [15]. Therefore service disruption cost can be
¼ BPVðB1 Þ þ BPVðB2 Þ  ½C 1 þ CPVðC 2 Þ þ CPVðC 3 Þ ð16Þ
evaluated by using (13) which evolved out of [15] as below:
" !# Therefore, distributed generation allocation problem can be
XNb X
Nres X
Nrep
C ENS ¼ C int;J  kb  Lb  Pres t res þ Prep t rep solved by using dynamic programming which is appropriate opti-
b¼1 res¼1 rep¼1 mization technique for the proposed function.
Given function shall be optimized considering below constrains.
þ C Equip;J ð13Þ
This point shall be noted that Eq. (13) can be used for ENS cal-  Voltage limits
culation with and without presence of DG in network.
If DG is sited in distribution system, it is used as alternative
source to restore power to part of the loads that are failed based Optimization shall be done in order to find out network nodes
on faults on transmission grid and distribution system and system where DG can be installed and voltage profile is in the standard
reliability is improved therefore reliability enhancement benefit limits (i.e. 0.9–1.1PU) or in recovering case close to it.

Network in
Peak load Industrial
load
CMW h, J [$\MW h]

Commercial
Cint, J [$\MWh]

load
Network in
medium load
Residential
load
Network in light
load

Network in Net work in Network in

light load medium load Peak load

PT [MW] PT [MW]
Fig. 1. Electricity price of transmission grid. Fig. 2. Estimation of energy not supply price in network.
 N. Khalesi et al. / Electrical Power and Energy Systems 33 (2011) 288–295 291

V min < V ij < V max ð17Þ  Problem will be solved by finding optimized policy for each
state from the last state which named backward solution.
where Vij is the voltage of bus ‘‘i” in load level j.
Response to this stage is evident because process will be pur-
sued from destination.
 Optimized policy for all states of stage ‘‘n” can be determined by
 Capacity of feeders
a backward function and by assuming that optimized policy for
all ‘‘n + 1” stages has been defined.
Maximum flowing power at network feeders shall be limited to
 Solution will be applied by using backward function from one
tolerance of conductors.
stage to previous stage running from end. In each stage, opti-
Sij < Smax i ð18Þ mized policy for all states of that stage will be specified and
finally optimized policy for first stage will be determined.
where Sij is the power flow at feeder ‘‘i” in load level j; Smax i is max-
imum flowing power at feeder ‘‘i”.

fn ðSn ; X n Þ ¼ Z n ðX n Þ þ fnþ1 ðSn  X n Þ ð20Þ
 Maximum capacity of installed DGs at network X
Nloc
fn ðSn ; X n Þ ¼ Z n ðX n Þ þ Max Z L ðX l Þ
Total installed capacity of distributed generations is limited by L¼nþ1

following constrain: X
Nloc
fn ðSn Þ ¼ Max f n ðSn ; X n Þ and X l ¼ Sn
X
NDG X
KDG L¼nþ1
DGJ;ik < DGJ; max ð19Þ
i¼1 k¼1 
where Sn: are states of stage n; fnþ1 optimized value of function in
where DGJ; max is maximum capacity of DG in load level j. stage n + 1; Xn decision at stage n.
Proposed algorithm based on dynamic programming for DG
4.1. Calculation of variables and indices placement has been accomplished by programming with MATLAB
application and it is extendable for different distribution network.
Calculation of objective function variables, which has been de- The general diagram of the algorithm that follows for DG alloca-
rived from reliability and power loss indices of the network, is the tion is shown in Fig. 3.
main part of optimization problem. There are different methods
and software for calculation of reliability and power loss. In this pa-
6. Case study
per, for calculation of power loss, Mat Power application of MATLAB
software with load flow capability has been used with minor mod-
Test system for case study has been shown in Fig. 4 [18]. For
ification and for calculation of reliability index; analytical method
testing of proposed technique, distributed generations have been
which has been programmed in Microsoft Excel has been used.
considered as negative loads, 1–5 MW, 0.9 lag power factors. Dis-
tribution test network includes high voltage distribution substa-
5. Dynamic programming tion 132–33 kV which feeds eight load points and each branch
has been separated from network by an isolator switch. Maximum
At most practical problems, sequential applications shall be capacity of each branch of the network is 25 MV A. Network daily
proceed in different time for solving a problem. Problems which active loads has been shown in Fig. 5. Power factor of all points is
shall be solved by sequential decisions are named sequential deci- 0.9 lag and all load points of the network has been considered as
sion problem. Dynamic programming in one kind of multi-stages candidate for installation of DG.
sequential decision problem which is an efficient mathematical
method for study and optimization of multi-stages sequential deci-
sion making problems [17]. Start

5.1. Optimization algorithm

Input Load Data
The major steps of the algorithm are:

 Problem will be divided into stages. For each stage, a decision Load Flow Analysis and ENS
policy will be required. In the other hand, each stage indicates Evaluation in Presence of DG
a part of problem which needs required decision. Number of on Candidate Points
stages in DG allocation problems equals number of candidate
locations for DG installation. Decision making methods at each
stage includes loss reduction and reliability improvement. Run Dynamic Programming
 Each stage contains related states. In current research, number Algorithm to Determine Optimal
of distributed generations with specified capacity that can be Allocation
allotted to mentioned stages, will be considered as state of
the problem.
 In each stage, current state of the stage will be transferred to No Check Next
related state in next stage by making a decision.
Are all Constraints Satisfied? Priority
 Autonomous policy for the remained stages can be followed by
knowing current state. Totally for optimization with dynamic Yes
programming, current state information transfers all required Display Results
information for previous behaviors which will be required for
identification optimized policy from current state to the next. Fig. 3. Optimal DG allocation methodology.
292 N. Khalesi et al. / Electrical Power and Energy Systems 33 (2011) 288–295

Technical information of the test network and loads in three

3 1 levels has been shown in Table 3. Minimum and maximum loads
2
of the network are 48.15 and 68.2 MW, respectively. Therefore load
demands of the network in this study changes from 70.6% to 100%
of the peak load.
Cost of purchased active power from transmission system in
6 specified level has been mentioned in Table 4. Time of network
5 being in three position light load, medium load and peak load will
7 4 be specified from the load curve which has been shown in Fig. 5.
Information for active load during 24 h of the day has been mod-
9 eled. In other hand, these curves indicate daily loading of the sys-
8
tem during whole year. For modeling of annual load change, these
: Isolator switch curves have been repeated 365 times.
: Branch
In addition, commercial information regarding DGs has been
: 132/33 kV substation : Load points specified in Table 5 that evolved out of [14,18].

Fig. 4. Studied test network.

6.1. Numerical studies and results analysis

Table 6 has been resulted from simulation based on effect of DG

80
allocation on optimization of power loss and reliability of the sys-
tem. In this table first column in each condition of the network re-
70
fers to loss reduction point of view and acquired benefit during
planning period. Second column in each condition of the network
60 refers to best location and size of DG from improvement of reliabil-
ity view point and its profit based on DG efficient in planning per-
50
Load Point 2 iod that has been exchanged to present worth.
Demand (Mw)

Load Point 3 Study results obtained from first column of light load (from loss
Load Point 4 reduction point of view in DG allocation) shows that there is
40
Load Point 5 0.584 MW decrease in loss which is equal to 1278.96 MW h/year
Load Point 6
30 Load Point 7

Load Point 8 Table 3

Technical characteristics of branches and load data.
20 Load Point 9

total Section R X Length of Max. load Max load Max load

(X) (X) branch in level 1 in level 2 in level 3
10 (km) (MW) (MW) (MW)
From To

0 1 3 1.4 1.5 1.5 5 6 8

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 3 7 2.78 5.5 5.5 7.5 8.8 9.2
Time of Day 1 2 2 4 4 8.3 11.2 9
2 6 2.8 5.5 5.5 4 5 7
Fig. 5. Curve of daily demands for high voltage distribution substation and eight 1 5 1.7 1.7 1.7 7.5 8.8 9.2
load points. 5 9 2.1 4 4 7.3 10.2 8
1 4 2.26 4.5 4.5 6 7 9
In this study, fault rate for 33 kV overhead lines, is 0.046 (f/km- 4 8 2.4 5 5 7.5 8.7 9.2

year) and for high voltage distribution substation 132/33 kV is

0.05 f/year and all other devices of the network has been consid-
ered to be 100% reliable. Maintenance time for overhead lines is
Table 4
8 h and for high voltage distribution substation is 15 h. Time for
Technical and commercial information.
fault location and isolating faulted zone and connection of DG to
the network has been considered 1 h [19]. In addition, price of en- Level Percentage of Network Time duration Market price ($/
peak load situation (h/year) MW h)
ergy not supply based on load type and network condition has
been mentioned in Table 2. 1 70.6–80 Light load 2190 35
2 80–95 Medium 4745 49
load
Table 2 3 95–100 Peak load 1825 70
Required information for reliability evaluation.

Load points Load type Network situation Price of not supplied

energy ($/kW h)
2, 3, 6 Residential Light load 0.053 Table 5
Medium load 0.073 Commercial information of DGs.
Peak load 0.105
Parameter Unit Value
4, 7, 8, 9 Commercial Light load 2
DG investment cost $/MW 318,000
Medium load 2.8
DG operation cost $/MW h 29
Peak load 3.6
DG maintenance cost $/MW h 7
5 Industrial Light load 6 Interest rate % 12.5
medium load 8.4 Inflation rate % 9
Peak load 11.050 Planning period (DG life time) Year 20
 N. Khalesi et al. / Electrical Power and Energy Systems 33 (2011) 288–295 293

Table 6
Profit analysis from studied viewpoints.

Network condition Peak load Medium load Light load

DG allocation From loss reduction From reliability From loss reduction From reliability From loss reduction From reliability
point of view point of view point of view point of view point of view point of view
2 MW at node 6 1 MW at node 5 2 MW at node 6 1 MW at node 5 2 MW at node 6 5 MW at node 8
3 MW at node 8 4 MW at node 8 1 MW at node 7 4 MW at node 8 1 MW at node 7
2 MW at node 8 2 MW at node 8
Benefit of loss reduction 1,685,881 1,404,571 2,792,897 2,259,740 679,784 571,531
($)
Benefit of reliability 358,797 587,431 237,525 456,890 169,669 326,349
improvement ($)
Total benefits ($) 2,044,678 1,992,008 3,030,422 2,716,630 849,453 897,880

and 679,784$ benefit acquired during DG life time. In addition, Table 8

profit due to reliability improvement of the network for DG locat- Economical costs for DG allocation in the network.
ing in order to reducing loss is 169,669$. From first column of mid Economical Network DG operation (MW h/ Costs ($)
loading condition it is concluded for DG placement from loss costs condition year)
reduction viewpoint, there is 0.791 MW decrease in loss of the sys- Investment Low load 10,950 1,590,000
tem that equal to 3753.3 MW h/year and its present worth is Operation 4,822,346
2792,897$. Moreover profit due to reliability enhancement of the Maintenance 116,401
network is 237,525$. For peak loading condition, loss will be lessen Total 6,528,747

0.869 MW which is equal to 1585.93 MW h/year considering peak Investment Medium load 23,725 1,590,000
loading interval which is worth 1685,881$ for DG life time and Operation 10,448,416
Maintenance 252,203
benefit for reliability improvement is 358,797$. Total 12,290,619
From second column of light loading, it can be seen 0.491 MW
Investment Peak load 9125 1,590,000
will be reduced from loss of the network which is equal to Operation 4,018,622
1075.29 MW h/year considering light load interval by DG alloca- Maintenance 97,001
tion for reliability improvement. Then present worth of this energy Total 5,705,623
in DG life time equals 571,531$. In the same condition during mid
load, loss reduction is 0.64 MW which is equal to 3036.8 MW h/
year and its worth is 225,9740$ for DG life time. During heavy
Table 9
loading, loss will be decreased to 0.724 MW, multiplying to heavy Net benefit resulted in study case.
loading duration it is equal to 1321.3 MW h/year. Then present
Network condition Peak load Medium Light load
worth is 1,404,577$.
load
It is worth mentioning that for allotted capacity to three condi-
Net benefit ($) 6,039,177 8,394,026 189,207
tions, distribution companies will purchase less power from trans-
DG allocation view Loss Loss Reliability
mission grid then related profit during DG life time shall be point reduction reduction improvement
considered too. Benefit regarding this condition has been summa-
rized in Table 7.
The remained problem is to calculate investment, maintenance
1.01
and operation costs of distributed generations for the three condi-
tions of loading. These costs considering commercial information 1
Without DG
of Table 5 has been shown in Table 8 for planning period. 0.99
With DG
At last, net benefit resulted of allocation DG based on proposed 0.98
Voltage profile

objective function in accordance with equation (16) shall be com-

0.97
puted. This profit considering benefit of optimized DG allocation
0.96
minus total costs from Tables 6 to 8 can be expressed in Table 9.
According to results presented in Table 9, in light loading condi- 0.95

tion optimized DG allocation emphasizing on reliability improve- 0.94

ment approaches more benefit comparing to loss reduction 0.93
effort. In this condition optimization constrains are in appropriate
0.92
range which has been shown in Figs. 6 and 7. In mid loading con-
0.91
dition, loss reduction view point achieves more benefit to distribu- 0 1 2 3 4 5 6 7 8 9 10
tion companies and as it can be seen from Figs. 8 and 9, there is load points
improvement in voltage profile and reduction on power flowing
Fig. 6. Voltage profile in light load.

Table 7
in the feeders. In peak load condition, network condition is similar
Reduction in purchased energy from transmission network.
to mid loading and constrains are in suitable range which can be
Network condition Reduction in purchased Benefit ($) seen in Figs. 10 and 11.
energy (MW h/year)
Finally, for comparing benefits of proposed approach with other
Light load 10,950 5,820,073 settled methodologies, two papers [14,20] have been considered
Medium load 23,725 17,654,221 with the same case study but DGs have been located and sized
Peak load 9125 9,700,122
on network with different methodologies and concepts.
294 N. Khalesi et al. / Electrical Power and Energy Systems 33 (2011) 288–295

18 1.02

16
Power flow in feeders (MW)

Without DG
1 Without DG

14 With DG With DG
0.98

Voltage profile
12
0.96
10
0.94
8
0.92
6
0.9
4
0.88
2
0.86
0
0 1 2 3 4 5 6 7 8 9 10
1 to 3 3 to 7 1 to 2 2 to 6 1 to 5 5 to 9 1 to 4 4 to 8
Load points
From/To Bus
Fig. 10. Voltage profile in peak load.
Fig. 7. Power flow in feeders in light load.

20
1.02
Without DG

Power flow in feeders (MW)

1 Without DG With DG
15
With DG
0.98
Voltage profile

0.96
10
0.94

0.92 5

0.9

0.88 0
1 to 3 3 to 7 1 to 2 2 to 6 1 to 5 5 to 9 1 to 4 4 to 8
0 1 2 3 4 5 6 7 8 9 10
load points From/To Bus

Fig. 8. Voltage profile in medium load. Fig. 11. Power flow in feeders in peak load.

Table 10
20 Results comparison.
Power flow in feeders (MW)

Without DG
Methodology Network Total capacity of Loss reduction
With DG condition DGs to network after DG
15 (MW) installation (%)
Proposed approach Light 5 24.5
load
10
Medium 5 21.5
load
Peak load 5 21.2
5
Ant colony optimization Light 0 0
[14] load
Medium 19 43.1
0
load
1 to 3 3 to 7 1 to 2 2 to 6 1 to 5 5 to 9 1 to 4 4 to 8
Peak load 20 36.8
From/To Bus
Comprehensive Peak load 18 MV A 53.6
optimization model and
Fig. 9. Power flow in feeders in medium load.
planners experience
[20]

But it shall be noted that reliability assessment and its benefits

for reliability improvement has not been studied on those papers.
Moreover, since cost dada in two mentioned papers have minor therefore loss has not been reduced in this network condition
differences with respect to this paper therefore they can be com- but this matter can be removed in case of considering benefits of
pared just on loss reduction point of view. reliability improvements in presence of DGs in the system. This pa-
Table 10 compares percentage of loss reduction and total capac- per by considering 1/4 of total DG capacity compared with [14] can
ity of DGs on network with these papers. reduce optimized amount of losses and attain remarkable benefits.
Authors in [14] have estimated and located about 20 MW DGs Authors in [20] in case of DG versus substation expansion (plan-
in network hence costs of investment, maintenance and operation ner decision) have settled DGs in distribution system with 18 MV A
of DGs are increased by enhancement on DGs capacity which can capacity which is about 3.5 times of total DG capacity in respect of
decrease final benefits. In addition, DGs are not located in light load this paper but important problem is that all analysis and calcula-
of network because costs of DGs are higher than their benefits tions in [20] is based on peak load condition and other situations
 N. Khalesi et al. / Electrical Power and Energy Systems 33 (2011) 288–295 295

of network conditions have not been considered and another prob- References
lem is that authors have not applied reliability benefits in objective
function of problem. [1] Wang C, Nehrir MH. Analytical approaches for optimal placement of
distributed generation sources in power systems. IEEE Trans Power Syst
But this point shall be considered that over increasing of distrib- 2004;19(4):2068–76.
uted generation total capacity in network that has been regarded [2] Gozel T, Hocaoglu MH. An analytical method for the sizing and siting of
in [14,20], may not have significant effects on the transmission sys- distributed generators in radial systems. Int J Electr Power Syst Res
2009;79:912–8.
tem, but its impacts at the lower voltage distribution system could [3] Le DT, Kashem MA, Negnevitsky M, Ledwich G. Optimal distributed generation
be important particularly in respect of fault current levels, the parameters for reducing losses with economic consideration. In: IEEE general
magnitude and direction power flow, the system voltage (both meeting, 24–28 June 2007. p. 1–8.
[4] Gandomkar M, Vakilian M, Ehsan M. A combination of genetic algorithm and
steady-state and transient) and the system stability under various
simulated annealing for optimal DG allocation in distribution networks. In:
small and large signal transient conditions. These impacts and 2005 Canadian conference on electrical and computer engineering, 1–4 May
interactions may be having negative impacts on the distribution 2005, p. 645–8.
network operating characteristics. [5] Celli G, Pillo F. Optimal distributed generation allocation in MV distribution
networks. In: Proceedings of the IEEE international conference on power
engineering society, May 2001. p. 81–6.
[6] Carpinelli G, Celli G, Mocci S. Optimization of embedded generation sizing and
7. Conclusions sitting by using a double trade-off method. IEE Proc Generat Transm Distrib
2005;152(4):503–13.
[7] Nara K, Hayashi Y, Ikeda K, Ashizawa T. Application of tabu search to optimal
In this paper, introducing dynamic programming as an optimi- placement of distributed generators. In: IEEE power engineering society winter
zation tool, a novel method has been presented to find best loca- meeting; 2001. p. 918–23.
tion for distributed generation installation in the network with [8] Cano Edwin B. Utilizing fuzzy optimization for distributed generation
allocation. In: IEEE region 10 Conference, TENCON 2007; 30 October 2007.
variable load model which results maximum profit. Load of the [9] Rau NS, Wan Y-H. Optimum location of resources in distributed planning. IEEE
network has been modeled in different level and the problem has Trans Power Syst 1994;4:2014–20.
been optimized considering existed constrains on permanent oper- [10] Ghosh S, Ghoshal SP, Ghosh Sa. Optimal sizing and placement of distributed
generation in a network system. Int J Electr Power Energy Syst
ation of the distribution system. From studied results it has been
2010;32:849–56.
derived that reliability and loss of the network are drastically de- [11] Sebastian Rios M, Marcelo Rubio S. Sequential optimization for sitting and
pends on location of the consumer, demanded power of the net- sizing distributed generation (DG) in medium voltage (MV) distribution
networks. In: IEEE power tech conferences; 2007.
work, type and capacity of the distributed generations and their
[12] Porkar S, Poure P, Abbaspour-Tehrani-fard A, Saadate S. A novel optimal
location in the network. In addition network condition have great distribution system planning framework implementing distributed generation
effects on DG allocation in system. in a deregulated electricity market. Int J Electr Power Syst Res
DG allocation has another advantage in addition to mentioned 2010;60:828–37.
[13] Borges CLT, Falcao DM. Optimal distributed generation allocation for
benefits which cannot be neglected. These beneficial effects in- reliability, losses and voltage improvement. Int J Electr Power Energy Syst
clude improvement in voltage profile of the load point and locating 2006;28:413–20.
network buses in allowable limit. Another advantage is reducing of [14] Falaghi H, Haghifam MR. Aco algorithm for distributed generation source
allocation and sizing in distribution systems. In: IEEE power tech conference;
power flow in feeders because of compensating loss and part of re- 1–5 July 2007. p. 555–60.
quired power of load points of the network. It decrease stress of the [15] Billinton R, Allan RN. Reliability evaluation of power system. 2nd ed. New
feeders especially feeders that they are next to high voltage distri- York: Plenum Press; 1996.
[16] Dugan RC, Price SK. Issues for distributed generation in the US. In: IEEE winter
bution substation. This increases duration of life time of the meeting, vol. 1; 27–31 January 2002. p. 121–6.
equipment. [17] Belman R, Dreyfus S. Applied dynamic programming. Princeton, New
Therefore DGs can get technical and financial benefits as indi- Jersey: Princeton University Press; 1962.
[18] Haghifam MR, Falaghi H, Malik OP. Risk-based distributed generation
cated in simulation results if allocated in proper locations with
placement. IET Generat Transm Distrib 2008:252–60.
appropriate sizes. [19] Allan RN, Billinton R, Sjarief I, Goel L. A reliability test system for educational
In future works there are extra considerations that must be purposes – basic distribution system data and results. IEE Trans Power Syst
1991;6(2).
careful attention about network upgrading problem and reduction
[20] El-Khattam W, Hegazy YG, Salama MMA. An integrated distributed generation
of purchased reactive power in presence of DG in network. This optimization model for distribution system planning. IEEE Trans Power Syst
mentioned factors may affect the results of DG allocation. 2005;20(2):1158–65.