This is a self-archived – parallel published version of this article in the

publication archive of the University of Vaasa. It might differ from the original.

Energy management strategy in dynamic

distribution network reconfiguration
considering renewable energy resources and
storage

Author(s): Azizivahed, Ali; Arefi, Ali; Ghavidel, Sahand; Shafie-khah, Miadreza; Li,
Li; Zhang, Jiangfeng; Catalão, João P. S.

Title: Energy management strategy in dynamic distribution network

reconfiguration considering renewable energy resources and storage

Year: 2020

Version: Final draft (post print, aam. accepted manuscript)

Copyright ©202X IEEE. Personal use of this material is permitted. Permission

from IEEE must be obtained for all other uses, in any current or future
media, including reprinting/republishing this material for advertising or
promotional purposes, creating new collective works, for resale or
redistribution to servers or lists, or reuse of any copyrighted component
of this work in other works.

Please cite the original version:

Azizivahed, A., Arefi, A., Ghavidel, S., Shafie-khah, M., Li, L., Zhang, J.,
& Catalão, J.P. S., (2020). Energy management strategy in dynamic
distribution network reconfiguration considering renewable energy
resources and storage. IEEE transactions on sustainable energy 11(2),
662-673. https://doi.org/10.1109/TSTE.2019.2901429
 1

Energy Management Strategy in Dynamic

Distribution Network Reconfiguration considering
Renewable Energy Resources and Storage
Ali Azizivahed, Member, IEEE, Ali Arefi, Senior Member, IEEE,
Sahand Ghavidel, Student Member, IEEE, Miadreza Shafie-khah, Senior Member, IEEE,
Li Li, Member, IEEE, João P. S. Catalão, Senior Member, IEEE, and Jiangfeng Zhang, Member, IEEE

 system in distribution networks), distribution feeder

Abstract—Penetration of renewable energy sources (RESs) reconfiguration (DFR) is implemented on distribution systems
and electrical energy storage (EES) systems in distribution in the presence of RES and energy storages. The DFR process
systems is increasing, and it is crucial to investigate their impact is to apply changes to the topology of distribution networks in
on systems’ operation scheme, reliability and security. In this
paper, expected energy not supplied (EENS) and voltage stability order to optimize certain objective functions subject to all
index (VSI) of distribution networks are investigated in dynamic operational constraints [5]. The DFR is carried out by
balanced and unbalanced distribution network reconfiguration, managing the on/off states of tie-switches and sectionalizing-
including RESs and EES systems. Furthermore, due to the high switches in a distribution feeder without islanding any buses.
investment cost of the EES systems, the number of charge and
discharge is limited, and the state-of-health constraint is included The DFR problem can be formulated as a mixed-integer,
in the underlying problem to prolong the lifetime of these non-linear and non-convex optimization problem. Therefore,
facilities. The optimal charging/discharging scheme for EES traditional gradient based optimization algorithms are not
systems and optimal distribution network topology are presented suitable to solve DFR [6]. Accordingly, many researchers
in order to optimize the operational costs, and reliability and
security indices simultaneously. The proposed strategy is applied
adopted intelligent evolutionary optimization methods to solve
to a large-scale 119-bus distribution test network in order to the distribution network reconfiguration problem. For
show the economic justification of the proposed approach. example, in [7], an enhanced gravitational search algorithm is
implemented to solve the DFR in order to improve transient
Index Terms—Energy Management, Distribution Network
Reconfiguration, Energy Storage, PV Panels, Reliability.
stability and reduce operational cost and power losses. In [8],
a hybrid evolutionary algorithm based on particle swarm

O
I. INTRODUCTION optimization algorithm and Nelder–Mead simplex search
algorithm is developed to minimize the active power loss.
PERATIONAL strategies of distribution networks have
Furthermore, a modified genetic algorithm is proposed for
significantly changed over the past decade due to the
DFR in [9] where the variable population size is taken into
high penetration of renewable energy sources (RESs) and
account. In [10], the optimal sizing, location, and network
energy storages alongside automation systems [1]. The
topology are obtained simultaneously by using optimal power
stochastic nature of RES poses a serious challenge to supply
flow to minimize operational cost and power losses.
the demand in a reliable way. Accordingly, a lot of studies
Additionally, the deployment of RESs and electrical energy
have been carried out to optimally manage charging and
storages requires studies on the optimal management of these
discharging schedules of energy storage units, which play a
facilities. Many studies are carried out in order to obtain an
decisive role in the management of renewable energy sources
optimal management scheme for electrical energy storages in
within distribution networks [2-4]. In addition, as one of the
the fixed-topology distribution networks. For instance, the
prevalent techniques (due to the integration of automation
optimal charging and discharging pattern for energy storage in
the distribution network is obtained using a modified
This work was supported in part by International Science and Technology
Cooperation Project of Sichuan Province, China, under Grant evolutionary algorithm to improve reliability and reduce
2018HH0146. J.P.S. Catalão acknowledges the support by FEDER funds operational cost [11]. In [12], technical and financial benefits
through COMPETE 2020 and by Portuguese funds through FCT, under
SAICT-PAC/0004/2015 (POCI01-0145-FEDER-016434), 02/SAICT/2017 of electrical energy storage systems in distribution networks
(POCI-01-0145-FEDER-029803) and UID/EEA/50014/2019 (POCI-01-0145- are investigated. A dynamic model for the energy management
FEDER-006961).
A. Azizivahed, S. Ghavidel, L. Li and J. Zhang are with the Faculty of
of dispatchable distributed generation sources of micro-grids
Engineering and Information Technology, University of Technology Sydney, in the presence of wind farms and PV farms is formulated in
PO Box 123, Broadway, NSW 2007, Australia (e-
mails:{Sahand.GhavidelJirsaraie, Ali.Azizivahed}@student.uts.edu.au; {li.li,
[13] to balance the generation and demand. In [14], the energy
jiangfeng.zhang}@uts.edu.au). storage units are allocated in optimal places in a distribution
A. Arefi is with the School of Engineering and Information Technology, system integrated with wind power and PV sources in order to
Murdoch University, Perth, Australia. (email: [email protected]).
M. Shafie-khah is with School of Technology and Innovations, University prolong the lifetime of energy storage units. Moreover,
of Vaasa, 65200 Vaasa, Finland (e-mail: [email protected]). optimal investment cost of batteries is obtained in order to
J.P.S. Catalão is with INESC TEC and the Faculty of Engineering of the
University of Porto, Porto 4200-465, Portugal (e-mail: [email protected]). maximize the benefit [15].
 2

It is noteworthy that the above-mentioned literature II. PROBLEM DEFINITION AND FRAMEWORK
regarding DFR has ignored the daily load variation and solved In some cases, DISCO is the owner of some parts of
the DFR during a predetermined time interval. The DFR distribution network [18]. Here, it is assumed that only one
model for non-variable loads cannot demonstrate the real
DISCO is the owner of all facilities and operates the
scenarios and cannot achieve the optimal solution for 24-hour
distribution network. It solves the stochastic optimization
time scheduling for variable load distribution networks. To fill
problem by considering uncertainty in PV generation and
this gap, the DFR is determined in [16] for different time
horizons (year, season, month and day) in order to find the electricity price. The problem formulation is explained in the
most optimal switching cost. In [17], the DFR is applied to an following six parts; decision variables, objective functions
unbalanced distribution network over a 24-hour time horizon. (operational cost, reliability index and security index),
In addition, genetic algorithm is implemented as an operational limitations and constraints, and uncertainty
optimization tool for minimizing network energy losses. modeling and optimization tool.
Though the presented robust strategy in [17] considers the A. Decision Variables
uncertain price, load consumption and RES power generation
are treated deterministically based on a fixed prediction. In a The decision variables of underlying problem include graph
distribution system with high penetration of RESs, applying topology of the network, diesel generators active power and
stochastic programming in order to model intermittent batteries charge and discharge scheme as follows:
behavior of these uncertainty sources is a prevalent and
practical solution for distribution operations. In addition, the
𝑿 = [𝑿𝑆𝑤 𝑿𝐷𝐺 𝑿𝐵𝑎𝑡𝑡 𝑿 𝑇𝑎𝑝 ] (1)
solution for dynamic DFR in the presence of RESs integrated 𝑿𝑆𝑤 = [𝑋̅𝑆𝑤1 𝑋̅𝑆𝑤2 … 𝑋̅𝑆𝑤𝑁𝑠𝑤 ] (2)
with energy storages is not evaluated markedly in the
aforementioned literature. 𝑿𝐷𝐺 = [𝑋̅𝐷𝐺1 𝑋̅𝐷𝐺2 … 𝑋̅𝐷𝐺𝑁𝐷𝐺 ] (3)
Accordingly, an improved optimization model is expected 𝑿𝐵𝑎𝑡𝑡 = [𝑋̅𝐵𝑎𝑡𝑡1 𝑋̅𝐵𝑎𝑡𝑡2 … 𝑋̅𝐵𝑎𝑡𝑡𝑁𝐵𝑎𝑡𝑡 ]
(4)
to include these three aspects: dynamic distribution feeder
reconfiguration, optimal management scheme for energy (5)
𝑿 𝑇𝑎𝑝 = [𝑋̅𝑡𝑎𝑝1 𝑋̅𝑇𝑎𝑝2 … 𝑋̅𝑇𝑎𝑝𝑁𝑇𝑎𝑝 ]

storages integrated with RESs, and meeting demand in a

(6)
1 2 24
𝑋̅𝑆𝑤𝑘 = [𝑋𝑆𝑤 𝑋𝑆𝑤 … 𝑋𝑆𝑤
reliable and stable way considering uncertainties in RESs
𝑘 𝑘 𝑘
], 𝑘𝜖{1,2, … , 𝑁𝑠𝑤}

power generation, energy price and load consumption. The (7)

1 2 24
𝑋̅𝐷𝐺𝑘 = [𝑋𝐷𝐺 𝑋𝐷𝐺 … 𝑋𝐷𝐺 ], 𝑘𝜖{1,2, … , 𝑁𝐷𝐺}
mentioned model should give operators a decision-making
𝑘 𝑘 𝑘

strategy in which the most suitable DFR is achieved by proper (8)

24
𝑋̅𝐵𝑎𝑡𝑡𝑘 = [𝑋𝐵𝑎𝑡𝑡
1
𝑘
2
𝑋𝐵𝑎𝑡𝑡 𝑘
… 𝑋𝐵𝑎𝑡𝑡 𝑘
], 𝑘𝜖{1,2, … , 𝑁𝐵𝑎𝑡𝑡}

DG unit dispatching pattern, optimal energy storage

(9)
24
𝑋̅𝑇𝑎𝑝𝑘 = [𝑋𝑇𝑎𝑝
1 2
𝑋𝑇𝑎𝑝 … 𝑋𝑇𝑎𝑝 ], 𝑘𝜖{1,2, … , 𝑁𝑇𝑎𝑝}
charging/discharging control, and appropriate reliability and
𝑘 𝑘 𝑘

where, 𝑿 is the decision variable vector of the proposed

stability levels. Eventually, the problem can be completed
when the constraints related to the technical and operational problem which consists of four sub-decision variables: open
aspects are considered through DFR and energy management switches (𝑿𝑆𝑤 ), diesel generators output (𝑿𝐷𝐺 ), batteries
problem. To address the above-mentioned problem, this paper active power charge/discharge (𝑿𝐵𝑎𝑡𝑡 ) and tap position of the
presents reliability and stability oriented management scheme. tap-changer (𝑿 𝑇𝑎𝑝 ). Notations 𝑋̅𝑆𝑤𝑘 , 𝑋̅𝐷𝐺𝑘 , 𝑋̅𝐵𝑎𝑡𝑡𝑘 and 𝑋̅𝑇𝑎𝑝𝑘
The main contributions of this paper are summarized as are the 𝑘 𝑡ℎ set of open switch numbers, active power output of
follows. 𝑘 𝑡ℎ diesel generator, active power of charge/discharge of the
 The dynamic distribution networks and energy 𝑘 𝑡ℎ battery, and tap position of the 𝑘 𝑡ℎ tap-changer, at a 24-
management are modeled simultaneously. The hour time horizon, respectively.
obtained results include optimal switching plan,
optimal batteries charging/discharging schedule and B. Operational Cost
optimal daily diesel generators dispatching. The operational cost includes the cost of energy purchasing
 The energy not supplied is considered as a separate from substation, the fuel cost of diesel generators and
objective function based on graph theory. Similarly, switching which can be formulated as:
the voltage stability based on the distribution network
loadability is formulated.
24 𝑁𝑠 𝑁𝑆𝑢𝑏 𝑁𝐷𝐺

 Scenario reduction strategy and shuffled frog leaping

𝑠𝑠 𝑠𝑠 𝐷𝐺 𝐷𝐺
𝐶𝑜𝑠𝑡 = ∑ ∑ 𝜌𝑠 ( ∑ 𝐶𝑛,𝑡,𝑠 × 𝑃𝑛,𝑡,𝑠 + ∑ 𝐶𝑗,𝑡,𝑠 × 𝑃𝑗,𝑡,𝑠 )

algorithm (SFLA) are utilized in order to obtain the

𝑡=1 𝑠=1 𝑛=1 𝑗=1

optimal solutions for both balanced and unbalanced

24 𝑁𝑆𝑤

distribution networks. Also, the intermittent nature of

+ ∑ ∑ 𝐶 𝑆𝑤 × |𝑆𝑘𝑡 − 𝑆0,𝑘
𝑡
|

electricity price, load consumption and PV

𝑡=1 𝑘=1
(10)
generations are considered as uncertainty sources. where, , and 𝐶 are the price of energy from 𝑛𝑡ℎ
The remainder of this paper is organized as follows.
𝑠𝑠 𝐷𝐺 𝑆𝑤
𝐶𝑛,𝑡,𝑠 𝐶𝑗,𝑡,𝑠
substation at 𝑡 time interval for 𝑠 𝑡ℎ scenario, the price of
Section II presents the underlying problem formulation and
𝑡ℎ

fuel of 𝑗𝑡ℎ diesel generator at 𝑡 𝑡ℎ time interval for 𝑠 𝑡ℎ scenario

framework. It consists of decision variables, objective
and switching cost, respectively; 𝜌𝑠 is the probability of 𝑠 𝑡ℎ
functions, problem constraints and optimization methodology.
scenario; 𝑁𝑆𝑢𝑏 (𝑁𝐷𝐺 ) and 𝑁𝑆𝑤 are the number of substations
Section III presents case studies and numerical results, and
(diesel generators) and the number of switches, respectively;
finally, Section IV concludes the paper.
 3

∗
𝑃𝑖,𝑡,𝑠 − 𝑗𝑄𝑖,𝑡,𝑠 = 𝑉𝑖,𝑡,𝑠 × 𝐼𝑖,𝑡,𝑠 (15)
𝑡ℎ
𝑉𝑖,𝑡,𝑠 − 𝑉𝑖,𝑡,𝑠
(16)
∗
𝑃𝑖,𝑡,𝑠 − 𝑗𝑄𝑖,𝑡,𝑠 = 𝑉𝑖,𝑡,𝑠 × 𝑡ℎ 𝑡ℎ
𝑅𝑖,𝑡,𝑠 + 𝑗𝑋𝑖,𝑡,𝑠
Eq. (17) can be calculated from (16). Coefficients 𝐵𝑖,𝑡,𝑠 and
𝐶𝑖,𝑡,𝑠 are defined by (18) and (19), respectively.
Fig. 1. Thevenin equivalent system of bus i at 𝑡 𝑡ℎ hour for 𝑠 𝑡ℎ scenario (17)
4 2
|𝑉𝑖,𝑡,𝑠 | − 𝐵𝑖,𝑡,𝑠 . |𝑉𝑖,𝑡,𝑠 | + 𝐶𝑖,𝑡,𝑠 = 0
𝑁𝑠 is the number of scenarios; 𝑃𝑛,𝑡,𝑠𝑠𝑠
and 𝑃𝑗,𝑡,𝑠
𝐷𝐺
are the active 𝐵𝑖,𝑡,𝑠 = 𝑡ℎ 2
|𝑉𝑖,𝑡,𝑠 | 𝑡ℎ
− 2𝑃𝑖,𝑡,𝑠 𝑅𝑖,𝑡,𝑠 𝑡ℎ
− 2𝑄𝑖,𝑡,𝑠 𝑋𝑖,𝑡,𝑠 (18)
power from 𝑛 substation and active power from 𝑗𝑡ℎ diesel
𝑡ℎ
(19)
2 𝑡ℎ 2
generator at 𝑡 𝑡ℎ time interval for 𝑠 𝑡ℎ scenario, respectively;
2 2 𝑡ℎ
𝐶𝑖,𝑡,𝑠 = (𝑃𝑖,𝑡,𝑠 + 𝑄𝑖,𝑡,𝑠 ). ((𝑅𝑖,𝑡,𝑠 ) + (𝑋𝑖,𝑡,𝑠 ) )

and 𝑆𝑘𝑡 and 𝑆0,𝑘

𝑡
are the new and initial status of 𝑘 𝑡ℎ switch at In order to have a stable condition, constraint (20) is
𝑡 𝑡ℎ time interval. required. Therefore, the VSI will be defined as in (21).
(20)
C. Desire Reliability Index
2
𝐵𝑖,𝑡,𝑠 − 4𝐶𝑖,𝑡,𝑠 ≥ 0

Almost all power outages and blackouts are caused by

2
𝑡ℎ 2 𝑡ℎ 𝑡ℎ
𝑣𝑠𝑖𝑖,𝑡,𝑠 = (|𝑉𝑖,𝑡,𝑠 | − 2𝑃𝑖,𝑡,𝑠 𝑅𝑖,𝑡,𝑠 − 2𝑄𝑖,𝑡,𝑠 𝑋𝑖,𝑡,𝑠 ) −
faults in the transmission and distribution networks [19, 20]. 2 2

Accordingly, the operational problems should be carried out to

2 2 𝑡ℎ 𝑡ℎ
4. (𝑃𝑖,𝑡,𝑠 + 𝑄𝑖,𝑡,𝑠 ). ((𝑅𝑖,𝑡,𝑠 ) + (𝑋𝑖,𝑡,𝑠 ) )

optimize a reliability index such as minimization of EENS as

𝑖 = 2, 3, … , 𝑁𝑏𝑢𝑠 (21)
follows: All parameters in (14)-(21) are depicted in Error!
𝑡=1 ∑𝑖=2 𝐸𝐸𝑁𝑆𝑖,𝑡 ,
Reference source not found.. In order to reach a stable
(11)
𝑁

operation condition, VSI for all buses must be greater than

𝐸𝐸𝑁𝑆 = ∑24 𝑏𝑢𝑠

where, 𝐸𝐸𝑁𝑆𝑖𝑡 is the expected energy not supplied of 𝑖 𝑡ℎ bus

zero. In this regard, the third objective function is defined as
at 𝑡 𝑡ℎ time interval; and 𝑁𝑏𝑢𝑠 is the number of buses. Notation follow;
𝐸𝐸𝑁𝑆𝑖,𝑡 is defined as follows:
𝒗𝒔𝒊𝑡,𝑠 = [𝑣𝑠𝑖2,𝑡,𝑠 , 𝑣𝑠𝑖3,𝑡,𝑠 , … , 𝑣𝑠𝑖𝑁𝑏𝑢𝑠 ,𝑡,𝑠 ] (22)
(12)
𝑁
𝐸𝐸𝑁𝑆𝑖,𝑡 = ∑𝑠=1
𝑠
𝜌𝑠 𝑃𝑖,𝑡,𝑠 × (∑𝑙∈𝐻𝑖 𝑈𝑙 + ∑𝑘∈𝐻𝑖′ 𝑈𝑘′ ), 0 𝑣𝑠𝑖𝑖,𝑡,𝑠 > 0
where, 𝑃𝑖,𝑡,𝑠 is the total generation and consumption of active (23)
𝑏𝑣𝑠𝑖𝑖,𝑡,𝑠 = {
1 𝑣𝑠𝑖𝑖,𝑡,𝑠 ≤ 0
power of 𝑖 𝑡ℎ bus at 𝑡 𝑡ℎ time interval for 𝑠 𝑡ℎ scenario; 𝐻𝑖 and 𝑩𝒗𝒔𝒊𝑡,𝑠 = [𝑏𝑣𝑠𝑖2,𝑡,𝑠 , 𝑏𝑣𝑠𝑖3,𝑡,𝑠 , … , 𝑏𝑣𝑠𝑖𝑁𝑏𝑢𝑠 ,𝑡,𝑠 ] (24)
𝐻𝑖′ are the set of downstream and upstream branches of 𝑖 𝑡ℎ (25)
bus, respectively; 𝑈𝑙 and 𝑈𝑘′ are the service unavailability
𝑝𝑒𝑛𝑎𝑙𝑡𝑦 𝑓𝑎𝑐𝑡𝑜𝑟 = ℳ × 𝑠𝑢𝑚(𝑩𝒗𝒔𝒊𝑡,𝑠 )

related to the reparation time of 𝑙 𝑡ℎ downward branch of the

1
(26)
𝑉𝑆𝐼𝑡,𝑠 = + 𝑝𝑒𝑛𝑎𝑙𝑡𝑦 𝑓𝑎𝑐𝑡𝑜𝑟
𝑖 𝑡ℎ bus and the service unavailability associated to the
𝑚𝑖𝑛(𝒗𝒔𝒊𝑡,𝑠 )

restoration time of 𝑘 𝑡ℎ upward branch of the 𝑖 𝑡ℎ bus,

24 𝑁𝑠
1
respectively. Notations 𝑈𝑙 and 𝑈𝑘′ are defined as follows:
𝑉𝑆𝐼 = ∑ ∑ 𝜌𝑠 𝑉𝑆𝐼𝑡,𝑠
(27)
24
𝑡=1 𝑠=1

𝑉𝑆𝐼 is the third objective function. The parameter ℳ is a

𝑈𝑙 = 𝛽𝑙 × 𝑡𝑙
(13)
large number (for instance, 10100 ) which is used as a penalty
𝑈𝑘′ = 𝛽𝑘 × 𝑡𝑘′
where, 𝛽𝑙 and 𝛽𝑘 are the failure rate (fail/year) of 𝑙 𝑡ℎ and 𝑘 𝑡ℎ factor. The penalty factor is implemented to eliminate the
branch, respectively; and 𝑡𝑙 and 𝑡𝑘′ are the average reparation unstable decision variables during the optimization process.
of 𝑙 𝑡ℎ branch and restoration time of 𝑘 𝑡ℎ branch, respectively.
E. Technical and Operational Constraints
D. Voltage Stability Index
In this section, all the relevant technical and operational
Voltage collapse is one of the negative consequences of constraints are explained as below.
increasing load in distribution systems. In order to improve the
network loadability, the voltage stability index is included in  Distribution network radial structure
the underlying problem and defined based on “Thevenin- Almost all distribution networks are operated in radial
equivalent” [21]. The advantage of this strategy in comparison topology in order to simplify the protection systems. In order
with the previous methods in [22] is that it can be to satisfy the radial structure constraint, the bus branch
implemented on both mesh and radial networks. More details incidence matrix is implemented. More details regarding the
of this method are explained as follows. proposed matrix can be found in [23].
First, the Thevenin-equivalent will be obtained for all bases
of networks as shown in Error! Reference source not  Diesel generator limitation
found.. As mentioned, the proposed problem is solved
According to the load flow technique, (14) and (15) can be dynamically. In this regard, the ramp rate constraints should
obtained; and from these equations, (16) can be obtained [22]. be considered besides the maximum and minimum output
𝑡ℎ
𝑉𝑖,𝑡,𝑠 − 𝑉𝑖,𝑡,𝑠 limitations.
(14) (28)
𝐼𝑖,𝑡,𝑠 = 𝑡ℎ 𝑡ℎ
𝑅𝑖,𝑡,𝑠 + 𝑗𝑋𝑖,𝑡,𝑠 𝑃𝑗𝐷𝐺,𝑚𝑖𝑛 ≤ 𝑃𝑗,𝑡,𝑠
𝐷𝐺
≤ 𝑃𝑗𝐷𝐺,𝑚𝑎𝑥 ∀𝑗, 𝑡, 𝑠,
 4

historical data of solar irradiance are used to produce a beta

(29)
𝑈𝑝
𝑅𝑎𝑚𝑝𝑗𝐷𝑜𝑤𝑛 ≤ 𝑃𝑗,𝑡,𝑠
𝐷𝐺 𝐷𝐺

[24] distribution function as follows;

− 𝑃𝑗,𝑡−1,𝑠 ≤ 𝑅𝑎𝑚𝑝𝑗 ∀𝑗, 𝑡 , 𝑠,

where, 𝑃𝑗𝐷𝐺,𝑚𝑖𝑛 and 𝑃𝑗𝐷𝐺,𝑚𝑎𝑥 are the minimum and maximum

limitations of 𝑗𝑡ℎ diesel generator, respectively; and
Г(𝛼 + 𝛽) 0≤𝑠≤1
∙ 𝑠 (𝛼−1) ∙ (1 − 𝑠)(𝛽−1) ,
Г(𝛼). Г(𝛽) 𝛼, 𝛽 ≥ 0
𝑅𝑎𝑚𝑝𝑗𝐷𝑜𝑤𝑛 and 𝑅𝑎𝑚𝑝𝑗 are down and up ramp rates of 𝑗𝑡ℎ
𝑈𝑝 𝑓𝑏 (𝑠) =

diesel generator, respectively. { 0, (37)

𝑂𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
 Current and voltage limitations where 𝑓𝑏 (𝑠) is the beta distribution function. 𝛼 and 𝛽 are the
The bus voltages and branch currents should be within their parameters of the beta distribution function and can be
maximum and minimum boundaries. obtained using historical data.
The continuous hourly beta PDFs are split into several
(30)
intervals with equal width. Each interval has a mean value and
𝑆𝐿𝑗,𝑡,𝑠 ≤ 𝑆𝐿𝑗𝑚𝑎𝑥

(31) a probability of occurrence which can be calculated as

𝑉𝑘𝑚𝑖𝑛 ≤ 𝑉𝑘,𝑡,𝑠 ≤ 𝑉𝑘𝑚𝑎𝑥

where 𝑆𝐿𝑗,𝑡,𝑠 and are the power flow magnitude of 𝑗𝑡ℎ

𝑆𝐿𝑗𝑚𝑎𝑥 follows.
branch at 𝑡 𝑡ℎ hour for 𝑠 𝑡ℎ scenario and its corresponding 𝑠𝑖+1

maximum power flow limitation, respectively; and 𝑉𝑘𝑚𝑖𝑛 and 𝜌𝑖𝑠

𝑉𝑘𝑚𝑎𝑥 are the minimum and maximum voltage of 𝑘 𝑡ℎ bus.
= ∫ 𝑓𝑏 (𝑠)𝑑𝑠𝑖
(38)
where 𝑠𝑖 and 𝑠𝑖+1 indicate the starting and ending points of the
𝑠𝑖

 Battery constraints
Because of the high investment cost of energy storages, interval i, respectively.
they should be operated in a secure environment to prolong The output power of a PV plant corresponding to specific
their lifetime. In this regard, the maximum permitted number solar irradiation can be calculated as follows.
of switching back and forth between charging and discharging
status must be considered besides the state of charge and other (39)
𝑆
𝑃𝑦,𝑡 (𝑠𝑦,𝑡 ) = 𝑁 × 𝐹𝐹 × 𝑉𝑦,𝑡 × 𝐼𝑦,𝑡
constraints. 𝑉𝑀𝑃𝑃 × 𝐼𝑀𝑃𝑃
(40)
1 𝐹𝐹 =
𝐸𝑖,𝑡 = 𝐸𝑖,𝑡−1 + 𝜌𝑐ℎ,𝑖 × 𝑃𝑐ℎ,𝑖,𝑡 × ∆𝑡 − 𝑃 × ∆𝑡, 𝑉𝑜𝑐 × 𝐼𝑠𝑐
𝜌𝑑𝑖𝑠,𝑖 𝑑𝑖𝑠,𝑖,𝑡
(41)
𝑐
𝑉𝑦,𝑡 = 𝑉𝑜𝑐 − 𝐾𝑣 × 𝑇𝑦,𝑡
𝑖 = 1,2, … , 𝑁𝐵𝑎𝑡𝑡,
(32) (42)
𝑡 = 1, … ,24, ∆𝑡 = 1ℎ 𝑐
𝐼𝑦,𝑡 = 𝑠𝑦,𝑡 [𝐼𝑠𝑐 − 𝐾𝐼 × (𝑇𝑦,𝑡 − 25)]
𝐸𝑖𝑚𝑖𝑛 ≤ 𝐸𝑖,𝑡 ≤ 𝐸𝑖𝑚𝑎𝑥 ; 𝑖 = 1,2, … , 𝑁𝐵𝑎𝑡𝑡 𝑡 = 1, … ,24 (33) 𝑁𝑂𝑇 − 20
(34)
𝑐

(43)
𝑇𝑦,𝑡 = 𝑇𝐴 + 𝑠𝑦,𝑡 ( )
𝑚𝑎𝑥
𝑃𝑐ℎ,𝑖,𝑡 ≤ 𝑃𝑐ℎ,𝑖 , 𝑖 = 1,2, … , 𝑁𝐵𝑎𝑡𝑡, 𝑡 = 1, … ,24 0.8
𝑚𝑎𝑥
𝑃𝑑𝑖𝑠,𝑖,𝑡 ≤ 𝑃𝑑𝑖𝑠,𝑖 , 𝑖 = 1,2, … , 𝑁𝐵𝑎𝑡𝑡, 𝑡 = 1, … ,24 (35) where 𝑐
is cell temperature (℃); 𝑇𝐴 is ambient temperature
𝑇𝑦,𝑡
(℃); 𝐾𝑣 and 𝐾𝑖 are voltage and current temperature coefficient
23
|𝑆𝑡+1 − 𝑆𝑡 |
(𝑉 ⁄℃ and A⁄℃), respectively; 𝑁𝑜𝑡 denotes nominal operating
∑ ⁄ ≤ 𝑁𝐶𝐻/𝐷𝐶𝐻
(36)
2
temperature of cell (℃); 𝐹𝐹 is fill factor; 𝐼𝑠𝑐 and 𝑉𝑜𝑐 indicates
𝑡=1

where 𝐸𝑖,𝑡 is the amount of energy storage in the 𝑖 𝑡ℎ battery at

short circuit current and open circuit voltage (A and V),
𝑡 𝑡ℎ hour. 𝑃𝑐ℎ,𝑖,𝑡 (𝑃𝑑𝑖𝑠,𝑖,𝑡 ) is the permitted rate of charge
respectively; 𝐼𝑀𝑃𝑃 and 𝑉𝑀𝑃𝑃 are, respectively current and
(discharge) of 𝑖 𝑡ℎ battery during a determinate period of time
voltage at maximum power point (A and V); 𝑃𝑦,𝑡 is output
(∆𝑡 = 1h). 𝜌𝑐ℎ,𝑖 (𝜌𝑑𝑖𝑠,𝑖 ) is the charge (discharge) efficiency
𝑆

power of the PV module; 𝑠𝑦,𝑡 solar irradiance; and 𝑡 and 𝑦 are

percentage of the 𝑖 𝑡ℎ battery. 𝐸𝑖𝑚𝑎𝑥 (𝐸𝑖𝑚𝑖𝑛 ) is the maximum
the indices of time periods and scenarios.
(minimum) amount of permitted energy storage in 𝑖 𝑡ℎ battery.
Constraints (34) and (35) impose the maximum charge rate 2) Market prices uncertainty modeling: In this paper, log-
𝑚𝑎𝑥
, and maximum discharge rate, 𝑃𝑑𝑖𝑠,𝑖
𝑚𝑎𝑥
, of the 𝑖 𝑡ℎ battery normal distribution function [25] is considered to characterize
the market price of each hour. Accordingly, the market prices
𝑃𝑐ℎ,𝑖
during a determinate period of time (∆𝑡 = 1h). In constraint
can be expressed as follows;
(36), 𝑁𝐶𝐻/𝐷𝐶𝐻 is the maximum permitted number of switching
back and forth between charging and discharging status and St 1 𝑝𝑟
(𝑙𝑛𝐸 − 𝜇)
2

represents the charging and discharging status which is equal

𝑓𝑝 (𝐸𝑝𝑟 , 𝜇, 𝜎) = 𝑒𝑥𝑝 ( − )
𝐸𝑝𝑟 𝜎√2𝜋
(44)
2𝜎2
to 1 and -1 for charging and discharging status, respectively.
where 𝜇 and 𝜎 represent mean value and standard deviation,
F. Uncertainty Characterization respectively, and 𝐸 𝑝𝑟 is the distribution function parameter
In this study, two sources of uncertainty are considered. (i.e., electricity market price).
 Electricity price of power supplied by substation In a similar way, the log-normal PDFs are sliced into
 Output power of PV plants. several intervals. Each interval yields a mean value and
1) PV power uncertainty modeling: For each time period,
 5

probability of occurrence.
Different realization of PV power production and market
prices can be modeled using a scenario generation process
according to roulette wheel mechanism [26].
In this method, a large number of scenarios is generated.
Higher numbers of scenarios result in a more precise modeling
of the system. However, the higher number of scenarios
causes higher computational burden. To this end, the number
of scenarios should be selected in a way that not only
diminishes the computational burden of the problem but also
maintains a good approximation of the uncertain parameters.
In order to reduce the number of scenarios and consequently
reduce the computational time; the backward method is
implemented to eliminate the duplicate scenarios or the
scenarios with minimum distance [27].
Fig. 3. The optimization framework of energy management and distribution
G. Multi-objective Strategy and Optimization Tool network reconfiguration

In this section, the multi-objective technique and the

1 𝑚𝑖𝑛
𝑂𝑏𝑗𝑚 ≤𝑂𝑏𝑗𝑚
shuffled frog leaping algorithm (SFLA) are introduced. The (
𝑚𝑎𝑥 -𝑂𝑏𝑗
𝑂𝑏𝑗𝑚 𝑚 𝑚𝑖𝑛 ≤𝑂𝑏𝑗 ≤𝑂𝑏𝑗 𝑚𝑎𝑥
µ𝑂𝑏𝑗𝑚 (𝑋)=
ALGORITHM and Fig. 2 describe the proposed multi-
𝑚𝑎𝑥 𝑚𝑖𝑛
𝑂𝑏𝑗𝑚 𝑚 𝑚

(45)
𝑂𝑏𝑗𝑚 -𝑂𝑏𝑗𝑚
objective technique. According to this method, all populations
𝑚𝑎𝑥
{ 0 𝑂𝑏𝑗𝑚 ≥𝑂𝑏𝑗𝑚
are sorted in ascending order of the first objective function, where 𝑂𝑏𝑗𝑚 and 𝑂𝑏𝑗𝑚 𝑚𝑎𝑥
(𝑂𝑏𝑗𝑚
𝑚𝑖𝑛
) are the 𝑚𝑡ℎ objective
then an eliminating zone is defined for each individual (i.e. function and its corresponding upper (lower) bound,
Fig. 2), and based on that, some populations are eliminated. respectively, and µ𝑂𝑏𝑗𝑚 is the fuzzy set for 𝑚th objective
This process is applied to reach a set of non-dominated function.
solutions. The fitness function is determined for each individual as
The ranges of objective function values are different in the follows:
multi-objective programming. Therefore, the fuzzy decision-
making technique (i.e. based on trapezoidal fuzzy membership
𝑁
𝑂𝑏𝑗
∑𝑚=1 𝜔𝑚 × µ𝑂𝑏𝑗 (𝑿𝑖 )
functions) is employed in order to have the same range for all (
𝑚
𝜑𝑖 =
of them. (46)
𝑁𝑛𝑑 𝑂𝑏𝑗𝑁
∑𝑘=1 ∑𝑚=1 𝜔𝑚 × µ𝑂𝑏𝑗 (𝑿𝑖 )
where 𝑁𝑛𝑑 is the number of non-dominated solution; 𝜑𝑖 and
𝑚

𝜔𝑚 are the fitness function of 𝑖 𝑡ℎ non-dominated solution and

the weighting factor (i.e. the priority grade from the decision
makers point of view) of 𝑚𝑡ℎ objective function, respectively.
In this study, 𝜔1 = 𝜔2 = 𝜔3 = 0.33 is chosen.
SFLA is implemented to solve the above optimization
problem. This algorithm models the social life of group of
frogs when they are searching food. The details of this method
are available in [28].
The frogs are divided equally into several memeplexes. In
Fig. 2. Eliminating zone for each solution this algorithm, the worst frog (𝑿𝑤𝑜𝑟𝑠𝑡 ) in each memeplex is
ALGORITHM updated according to two strategies based on the best frog in
1. Input: 𝒑𝒐𝒑 = [population & objective functions]. the memeplex (𝑿𝑏𝑒𝑠𝑡 ) and the best frog in all population
2. Sort the population in ascending order of the first objective (𝑿𝐺𝑏𝑒𝑠𝑡 ) as follows;
function.
3. 𝑧=1, 𝑁𝑝𝑜𝑝 =𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑝𝑜𝑝𝑢𝑙𝑎𝑡𝑖𝑜𝑛.
𝑿𝑛𝑒𝑤
(47)
𝑤𝑜𝑟𝑠𝑡 = 𝑿𝑤𝑜𝑟𝑠𝑡 + 𝑟𝑎𝑛𝑑 × (𝑿𝑏𝑒𝑠𝑡 − 𝑿𝑤𝑜𝑟𝑠𝑡 )
4. While 𝑧 ≤ 𝑁𝑝𝑜𝑝
(48)
new
5. Constitute the eliminating zone (𝐸𝑍𝑧 ) for 𝑧 𝑡ℎ population
𝐗worst = 𝐗worst + rand × (𝐗Gbest − 𝐗worst )
(red shadow in Fig. 2). where 𝑟𝑎𝑛𝑑 is random number between 0 and 1. In the
6. Eliminate the populations that are in the eliminating zone optimization process, the worst frog is updated by the best
of 𝑧 𝑡ℎ population. frog in the memeplex as in (47). If the fitness function for
𝑤𝑜𝑟𝑠𝑡 is better than of 𝑿𝑤𝑜𝑟𝑠𝑡 , then 𝑿𝑤𝑜𝑟𝑠𝑡 will be replaced
𝒑𝒐𝒑𝒛𝒆𝒍 ⊆ 𝒑𝒐𝒑 & 𝑿𝑛𝑒𝑤
𝒑𝒐𝒑𝒛𝒆𝒍 ={𝑠𝑒𝑡 𝑜𝑓 𝑝𝑜𝑝𝑢𝑙𝑎𝑡𝑖𝑜𝑛𝑠 𝑖𝑛 𝑡ℎ𝑒 𝐸𝑍𝑧 𝑎𝑟𝑒𝑎} by 𝑿𝑛𝑒𝑤
𝑤𝑜𝑟𝑠𝑡 ; otherwise 𝑿𝑤𝑜𝑟𝑠𝑡 will be updated by the best frog
𝒑𝒐𝒑 = 𝒑𝒐𝒑 ∖ {𝒑𝒐𝒑𝒐𝒛𝒆𝒍 } in all population as in (48). Similarly, if the fitness function
7. Update 𝑧=𝑧+1, 𝑁𝑝𝑜𝑝 =𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑝𝑜𝑝𝑢𝑙𝑎𝑡𝑖𝑜𝑛.
for 𝑿𝑛𝑒𝑤
𝑤𝑜𝑟𝑠𝑡 is better than 𝑿𝑤𝑜𝑟𝑠𝑡 , 𝑿𝑤𝑜𝑟𝑠𝑡 will be replaced by
8. end 𝑤𝑜𝑟𝑠𝑡 , and if not, 𝑿𝑤𝑜𝑟𝑠𝑡 will be replaced by a new randomly
𝑿𝑛𝑒𝑤
9. Output: 𝑠𝑒𝑡 𝑜𝑓 𝑛𝑜𝑛 𝑑𝑜𝑚𝑖𝑛𝑎𝑡𝑒𝑑 𝑠𝑜𝑙𝑢𝑡𝑖𝑜𝑛𝑠 = 𝒑𝒐𝒑. generated frog.
 6

This process will be applied for all memeplexes and for a TABLE I
RESULTS OF DIFFERENT METHOD FOR MINIMIZING THE OPERATIONAL COST
predetermined number of iterations. The framework of the
Operational Cost ($)
proposed strategy is depicted in Error! Reference source not Method
Best Mean Worst STD C.T.
found..
GA 13068.06 13206.7 13355.0 109.81 201
PSO 12835.64 12985.2 13355.0 212.32 164
III. SYSTEM MODEL AND SIMULATION RESULTS DE 12835.64 12903.5 13161.1 99.79 178
SFLA 12820.85 12841.2 12897.1 25.32 167
In order to assess the performance of the proposed method,
C.T.: Computational Time (Second)
a case study based on the 119-bus distribution network is
provided in this section. B. Simulation and Results
A. Case Study The three objective functions (Eqs. 10, 12 and 27) are
important for the reliable and cost-effective operation of the
The 119-bus distribution network under study is shown in Fig.
distribution system; however, it is usually impossible to
4(a), consisting of three feeders, 15 tie-switches and 11 kV
substation [29]. The average hourly forecasted active and achieve the optimal results for all of them simultaneously.
reactive load profile is shown in Fig. 4(b). Four 500 kW PV Also, the operation of diesel generators, batteries, and network
panels and their relevant batteries are installed at bus#31, switches are different in each case. Hence, various analyses
bus#42, bus#96 and bus#109. Five 500 kW diesel generators are performed to explore the best compromise solution as will
with unit power factor and 50 kW/h up and down ramp rate be described in the following.
are located at bus#20, bus#28, bus#71, bus#74 and bus#111. 1) Single objective case study: SFLA is compared with
30 scenarios are implemented in order to simulate the three different heuristic methods such as genetic algorithm
uncertainty parameters. The simulation is done in MATLAB (GA), particle swarm optimization (PSO) and differential
R2011b environment using a core-i5 processor laptop with 2.4 evolution (DE), each is run 10 times to solve the DFR
GHz clock frequency and 4.0 GB of RAM. optimization problem, and the comparison results are detailed
in Error! Reference source not found.. Although the PSO
reaches the optimal solution in a slightly shorter
computational time compared to the SFLA, the obtained
results by the SFLA are better than those obtained by the other
methods. And also, it has a more robust performance than
other algorithms. Therefore, in the following, only SFLA is
used.
TABLE II shows the results of the optimization results for
individual objectives and the initial condition (distribution
network without DG units, PV panels and energy storages).
Firstly, it is worth mentioning that the objective values of all
objective functions are improved due to the positive impact of
these facilities. In addition, the targeted objective (for
example, operational cost) is minimized, and subsequently, the
values of the other two objectives are calculated in each case.
According to this table, the amount of optimal operational
cost, optimal EENS, and optimal VSI are $12820.85, 340.73
kWh/year and 2.53 p.u., respectively. Also, it can be seen that
the operational cost is sharply in conflict with two other
objective functions. In other words, by minimizing operational
cost, the amount of ENS and VSI are increased (368.646
kWh/year and 3.397 p.u.). And similarly, by minimizing the
ENS and VSI, the amount of operational cost is increased to
a) Single line 119-bus distribution network
$13413.68 and $12939.96, respectively. Similarly, the
operational cost conflicts with VSI. The optimal VSI amount
is 2.1, and in this condition, the amount of operational cost is
$13278.1.
The list of open-switches for minimizing operational cost is
shown in 0. it is worth mentioning that the radiality constraint
is satisfied in each hour. furthermore, Fig. 5 and Fig. 6 show
the diesel generators active power output and batteries active
power charge/discharge. as can be seen, almost all the diesel
generators are operated at their minimum level in order to
b) Active and reactive load profiles minimize the operational cost. also, from Fig. 6 it can be seen
Fig. 4. Data for 119-bus standard distribution system that the maximum permitted number of switching back and
 7

forth between the charging and discharging status and the

charging and discharging status constraint are satisfied for all
batteries.
The list of open-switches for minimizing operational cost is
shown in 0. It is worth mentioning that the radiality constraint
is satisfied in each hour. Furthermore, Fig. 5 and Fig. 6 show
the diesel generators active power output and batteries active
power charge/discharge. As can be seen, almost all the diesel
generators are operated at their minimum level in order to
minimize the operational cost. Also, from Fig. 6 it can be seen
that the maximum permitted number of switching back and Fig. 6. Active power of batteries during charge and discharge for the
forth between the charging and discharging status and the optimal operational cost.
charging and discharging status constraint are satisfied for all
2) Multi-objective case study: As mentioned, all three
batteries.
TABLE II objective functions are in conflict, and it is impossible to find
RESULTS OF MINIMIZING THE OBJECTIVE FUNCTIONS SEPARATELY a solution to have optimal operational cost, ENS and VSI
Method Cost ($) ENS (kWh/year) VSI (p.u.) simultaneously. Therefore, the best strategy is to find a
Initial Condition 15633.87 544.7785 4.0509 compromise among the three conflicted objective functions. In
Cost Minimization 12820.85 368.646 3.3974 this regard, “Pareto optimal strategy” is applied in order to
ENS Minimization 13413.68 340.7324 3.3831 obtain a set of non-dominated solutions, and then decision
VSI Minimization 12939.96 367.9338 2.5336 makers would be able to have a tradeoff among the objective
functions according to their considered priority. Fig. 7, Fig. 8
TABLE III and Fig. 9 show the two-dimension Pareto optimal solutions
LIST OF OPEN SWITCHES FOR MINIMIZING OPERATIONAL COST
for operational cost-ENS, operational cost-VSI and ENS-VSI,
Hour Open Switches
respectively. The percentages in these figures show the
1 43-11- 23- 51- 47- 61- 38- 56- 72- 73- 98- 82- 85- 131- 32
2 45- 12- 17- 53- 122- 36- 39- 57- 66- 73- 128- 105- 101- 115- 33
amount of conflict between the objective functions. The most
3 43- 24- 20- 53- 46- 36- 27- 54- 71- 127- 98- 105- 102- 116- 33 conflict is observed in the operational cost-VSI case with
4 43- 12- 120- 52- 48- 123- 27- 56- 126- 127- 96- 81- 102- 115- 33 70.69%, while the minimum conflict is between ENS-VSI
5 118- 16- 7- 50- 48- 61- 124- 55- 70- 127- 76- 106- 101- 131- 34 with 5.3%. The conflict between operational cost-ENS is
6 42- 25- 120- 51- 48- 123- 38- 55- 71- 87- 128- 105- 103- 113- 34 9.69%. In order to have a reliable and secure operational cost,
7 118- 14- 23- 49- 46- 36- 37- 57- 89- 127- 128- 105- 103- 115- 32 it is better than to find a set of three-dimension non-dominated
8 44- 12- 120- 121- 48- 61- 37- 56- 66- 127- 76- 82- 85- 108- 132
solutions.
9 118- 14- 23- 53- 122- 61- 39- 57- 68- 73- 75- 105- 103- 109- 33
10 44- 24- 21- 53- 47- 58- 37- 56- 72- 74- 96- 129- 130- 131- 31
Three-dimension Pareto front is shown in Fig. 10. The best
11 41- 13- 23- 53- 122- 59- 38- 125- 70- 87- 96- 82- 102- 113- 30 compromise solution with the equal priority weight (𝑤1 =
12 45- 15- 19- 52- 48- 58- 124- 57- 89- 87- 97- 129- 85- 108- 34 𝑤2 = 𝑤3 = 0.33) is highlighted with a red star.
13 41- 12- 23- 50- 47- 36- 124- 57- 70- 87- 76- 105- 85- 115- 31 For this solution, the amounts of operational cost, ENS and
14 41- 16- 17- 121- 48- 58- 39- 57- 126- 87- 128- 81- 85- 116- 34 VSI are $13933.064, 373.96 kWh/year and 3.213 p.u.,
15 118- 14- 120- 50- 46- 58- 38- 56- 88- 74- 128- 105- 102- 116- 34 respectively.
16 42- 14- 20- 52- 122- 123- 124- 55- 72- 127- 97- 105- 103- 117- 33
The list of open switches for best compromise solution is
17 42- 12- 23- 49- 48- 61- 39- 56- 71- 74- 76- 80- 100- 116- 34
18 45- 15- 18- 121- 48- 58- 124- 57- 71- 73- 128- 81- 102- 116- 34
listed in TABLE IV. According to these results, the crucial
19 45- 15- 20- 51- 48- 36- 124- 57- 72- 87- 128- 105- 85- 109- 30 constraint for the radial structure is satisfied for each topology.
20 45- 16- 7- 50- 122- 58- 27- 54- 71- 86- 97- 82- 103- 109- 30 Furthermore, Fig. 11 and Fig. 12 depict the active power
21 41- 10- 18- 121- 122- 58- 38- 57- 90- 127- 96- 106- 101- 108- 33 output of diesel generators and active power of
22 43- 14- 21- 51- 46- 60- 38- 57- 72- 127- 97- 81- 85- 115- 34 charging/discharging of batteries, respectively. From Fig. 11,
23 43- 25- 22- 121- 47- 58- 39- 54- 69- 86- 75- 82- 100- 114- 34 it can be observed that the diesel generators are operated close
24 118- 14- 120- 53- 48- 58- 124- 54- 68- 87- 76- 106- 103- 109- 33
to their middle levels in order to have a secure and reliable
operation plan.
According to the energy not supplied formulation, it is
better to feed the load consumption locally instead of feeding
them through the transmission system. Then with respect to
ENS improvement, the diesel generators tend to operate at
their maximum level to feed the load consumption locally.
The same analysis can be done for VSI. In order to increase
the distribution load-ability, it is better to feed the load
consumption locally. In the other way, diesel generators tend
to operate at their minimum level in order to minimize the
operational cost, and then the best compromise makes a
Fig. 5. Diesel generator power scheduling for the optimal operational cost. tradeoff between these two different tendencies.
 8

TABLE IV
LIST OF OPEN SWITCHES FOR BEST COMPROMISE

Hour Open Switches

1 40- 14- 22- 49- 47- 60- 27- 56- 65- 73- 128- 129- 104- 114- 30
2 118- 24- 7- 52- 122- 58- 39- 125- 66- 87- 76- 106- 104- 113- 30
3 45- 25- 6- 121- 122- 36- 27- 57- 70- 73- 98- 105- 85- 113- 33
4 44- 25- 7- 52- 47- 36- 37- 56- 71- 87- 128- 129- 103- 131- 31
5 44- 12- 21- 50- 48- 59- 38- 57- 69- 127- 76- 105- 130- 108- 32
6 45- 119- 23- 49- 122- 36- 124- 125- 70- 73- 98- 106- 103- 117- 32
7 118- 13- 21- 121- 122- 35- 124- 55- 90- 86- 128- 129- 103- 114- 33
Fig. 7. Two-dimension Pareto front for operational cost-EENS 8 40- 10- 18- 121- 46- 36- 38- 57- 70- 74- 97- 129- 101- 114- 33
9 44- 24- 4- 53- 48- 59- 38- 125- 70- 73- 96- 129- 104- 114- 34
10 41- 13- 7- 49- 122- 58- 38- 56- 69- 86- 75- 129- 130- 116- 132
11 41- 16- 20- 53- 122- 60- 124- 54- 89- 74- 97- 129- 101- 115- 31
12 43- 25- 3- 51- 48- 61- 37- 57- 90- 74- 96- 106- 101- 117- 34
13 118- 11- 21- 49- 46- 60- 38- 54- 65- 74- 98- 80- 100- 117- 30
14 42- 15- 19- 53- 48- 36- 37- 54- 69- 74- 128- 81- 85- 115- 132
15 42- 14- 23- 51- 46- 59- 37- 57- 66- 127- 98- 81- 101- 116- 30
16 44- 25- 18- 51- 47- 61- 27- 56- 88- 74- 96- 82- 101- 108- 33
17 43- 26- 21- 49- 47- 123- 37- 125- 126- 87- 76- 82- 130- 115- 33
18 42- 14- 20- 53- 48- 36- 39- 55- 69- 87- 76- 129- 130- 116- 32
19 118- 13- 23- 52- 47- 36- 37- 57- 68- 74- 128- 106- 100- 109- 132
20 44- 11- 19- 53- 47- 58- 124- 55- 69- 127- 76- 82- 85- 117- 34
Fig. 8. Two-dimension Pareto front for operational cost-VSI 21 118- 119- 7- 49- 48- 123- 39- 57- 69- 87- 98- 106- 104- 117- 32
22 118- 26- 18- 52- 47- 59- 38- 54- 70- 127- 96- 82- 103- 108- 32
23 41- 24- 6- 53- 122- 35- 27- 54- 65- 73- 128- 106- 103- 109- 30
24 42- 119- 120- 53- 48- 59- 124- 56- 65- 127- 98- 107- 101- 131- 132

Fig. 9. Two-dimension Pareto front for EENS-VSI

Fig. 11. The diesel generator power scheduling for the best compromise

While, the amount of VSI represents the amount of desire

voltage stability index for the bus with the lowest stability
margin. Furthermore, the conflict between the objective
functions is obvious in the unbalanced case as well as the
single-line case, and the best compromised solution provides
the trade-off between the objective functions.
Fig. 10. Three-dimension Pareto front for operational cost-ENS-VSI
The list of open-switches and the tap-changer positions are
3) Unbalanced case study: In this subsection, the proposed tabulated in TABLE VI. In accordance with these results, it is
approach is tested on an unbalanced version of the 119-bus worthwhile to note that the crucial radial constraint is satisfied
distribution network. Compared to the original balanced test at all time intervals. Moreover, it is evident that the load
case, the total loads are 22654 kW and 16980 kVar at phase A, profile is followed by the tap-positions. In other words, during
21568 kW and 16038 kVar at phase B, and 21646 kW and the peak load period, the tap-positions tend to increase the
16106 kVar at phase C. Furthermore, three voltage regulators voltage magnitude for two reasons; the first reason is to
are encompassed on the three main feeders. In this case, the TABLE V
tap positions are considered in the decision variables and RESULTS OF OPTIMIZATION FOR THE UNBALANCED CASE
optimized with the other variables simultaneously. Method Cost ($) ENS (kWh/year) VSI (p.u.)
TABLE V shows the extracted results for individual Cost Minimization 37118.89 1088.581 3.1539
objective as well as the multi-objective case. According to ENS Minimization 38118.83 1073.687 6.0478
these results, the amount of operational cost and ENS are VSI Minimization 37578.26 1153.294 2.7322
Best Compromise 37175.70 1092.425 3.1532
almost tripled in comparison with the single-line case study.
 9

TABLE VI
OPEN SWITCHES AND TAP CHANGER POSITION FOR THE BEST COMPROMISE
Hour Open Switches Tap Position
1 44, 24, 19, 51, 122, 58, 39, 55, 89, 74, 97, 107,
(-13), (-15), (-12)
100, 115, 30
2 40, 11, 6, 49, 122, 35, 124, 54, 67, 74, 75, 107,
(-15), (-7), (-10)
104, 115, 34
3 118, 14, 18, 53, 122, 58, 37, 57, 71, 74, 75,
(-12), (-12), (-16)
105, 100, 109, 132
4 43, 24, 22, 51, 122, 36, 27, 57, 66, 127, 96,
(-12), (-15), (-10)
105, 102, 117, 132
5 40, 14, 6, 52, 46, 123, 124, 57, 67, 127, 76, 82,
(-9), (-15), (-14)
104, 117, 31
6 44, 13, 17, 53, 46, 35, 39, 56, 90, 86, 97, 82, Fig. 12. Active power of batteries during charge and discharge for the best
(3), (10), (11) compromise
103, 109, 32
7 42, 11, 20, 49, 46, 123, 37, 55, 68, 74, 76, 82,
(15), (14), (14)
101, 113, 30
8 43, 16, 19, 51, 48, 58, 37, 57, 67, 87, 128, 81,
(-8), (-14), (-16)
104, 116, 132
9 45, 15, 18, 53, 122, 59, 38, 125, 88, 127, 98,
(-7), (-8), (-15)
80, 104, 117, 33
10 42, 15, 20, 121, 122, 60, 37, 56, 70, 86, 98,
(-16), (-10), (-14)
106, 100, 131, 32
11 42, 11, 7, 52, 47, 35, 39, 125, 88, 87, 75, 82,
(-8), (-11), (-10)
104, 115, 30
12 40, 15, 23, 49, 122, 35, 39, 125, 69, 127, 76,
(10), (10), (9)
82, 130, 116, 132
13 118, 16, 21, 50, 47, 60, 39, 57, 71, 74, 96, 82,
(10), (9), (8)
103, 109, 30
14 43, 26, 23, 52, 122, 35, 38, 57, 68, 74, 98, 81,
(10), (11), (9)
102, 109, 31
15 45, 10, 19, 50, 46, 36, 124, 57, 72, 74, 97, 105, Fig. 13. Distribution network VSI profile (upper), distribution network bus-
(14), (10), (8) vsi profile at hour #19 (lower)
130, 116, 30
16 41, 26, 120, 50, 122, 36, 38, 54, 66, 73, 98, 81,
(14), (11), (10)
85, 131, 132
17 44, 16, 18, 53, 122, 36, 39, 125, 69, 73, 97, 80,
(15), (16), (13)
104, 114, 33
18 118, 14, 19, 52, 48, 60, 124, 57, 89, 86, 128,
(15), (16), (14)
80, 85, 116, 34
19 40, 26, 17, 53, 47, 58, 124, 54, 90, 87, 75, 106,
(15), (16), (14)
104, 114, 30
20 40, 13, 18, 49, 46, 58, 38, 57, 88, 73, 97, 80,
(-1), (3), (-2)
130, 109, 33
21 41, 11, 23, 49, 122, 36, 37, 54, 88, 87, 98, 81,
(-1), (-1), (-2)
85, 113, 33
22 44, 15, 120, 53, 48, 35, 27, 56, 126, 74, 75,
(-3), (-2), (-2)
129, 130, 131, 33 Fig. 14. Daily voltage profile for bus 55
23 43, 24, 21, 53, 46, 35, 39, 54, 70, 127, 98, 107,
(-4), (-5), (-4)
101, 114, 34
24 45, 13, 21, 53, 46, 59, 38, 57, 69, 73, 76, 106,
(-5), (-7), (-9)
101, 116, 33
avoid voltage drop, and the second one is to decrease power
losses. In the off-peak period, the tap changers will reduce the
voltage magnitude in order to avoid the over-voltage issue.
The total energy transaction of batteries during their charge
and discharge and the average active and reactive power of
DGs are listed in 0. The results indicate that the batteries at
buses 31 and 109 have more penetration in comparison with
the batteries located at buses 42 and 96.
TABLE VII Fig. 15. P-V and Q-V curves of bus 55 as the result of load variations
THE PENETRATION OF BATTERIES AND DGS IN THE UNBALANCED
Batteries Energy Transaction Also, the results depict that the DGs at buses 20 and 74 have
Battery No. Bus# 31 Bus# 42 Bus# 96 Bus# 109 the most and the least commitment in the distribution network,
E (kWh) 1002.3 751.3 399.6 1190.6 respectively. The DGs at buses 28, 71 and 111 have
The Average DGs Power Transaction
approximately the same commitment with a value less than the
DG No. Bus# 20 Bus# 28 Bus# 71 Bus# 74 Bus# 111
P (kW) 468.92 413.93 432.425 248.16 418.98
DG at bus 28.
Q (kVar) 167.12 135.55 181.84 92.54 136.05 Aiming to have a better understanding of the improvement
in terms of VSI based on distribution network loadability,
some analysis is performed as follows.
 10

Error! Reference source not found.Fig. 13 (upper) shows [7] E. Mahboubi-Moghaddam, M. R. Narimani, M. H. Khooban,
the network voltage stability profile during the 24-hour time and A. Azizivahed, "Multi-objective distribution feeder
horizon. According to this figure, the worst voltage stability reconfiguration to improve transient stability, and minimize
power loss and operation cost using an enhanced evolutionary
happens at hour #19. Furthermore, the bus-vsi profile at hour algorithm at the presence of distributed generations,"
#19 and the worst bus-vsi are shown in Fig. 13 (lower)Error! International Journal of Electrical Power & Energy Systems,
Reference source not found.. The worst bus-vsi happens at vol. 76, pp. 35-43, 2016.
the bus 55 with 0.31684 p.u., and the daily voltage profile of [8] T. Niknam, E. Azadfarsani, and M. Jabbari, "A new hybrid
this bus is depicted in Fig. 14. The P-V and Q-V curves for evolutionary algorithm based on new fuzzy adaptive PSO and
presented strategies are plotted in Fig. 15. The continuous- NM algorithms for distribution feeder reconfiguration," Energy
Conversion and Management, vol. 54, no. 1, pp. 7-16, 2012.
power-flow method is used to plot these curves for the [9] M. Abdelaziz, "Distribution network reconfiguration using a
weakest bus at the critical hour (19 o’clock) in the case study. genetic algorithm with varying population size," Electric Power
Obviously, by improving the VSI, the stability margin is Systems Research, vol. 142, pp. 9-11, 2017.
increased. [10] M. Nick, R. Cherkaoui, and M. Paolone, "Optimal Planning of
Distributed Energy Storage Systems in Active Distribution
IV. CONCLUSION Networks Embedding Grid Reconfiguration," IEEE
Transactions on Power Systems, 2017.
This paper proposed a new energy management strategy in [11] A. Azizivahed, E. Naderi, H. Narimani, M. Fathi, and M. R.
dynamic distribution network reconfiguration considering Narimani, "A New Bi-Objective Approach to Energy
renewable energy resources and energy storage to improve the Management in Distribution Networks with Energy Storage
distribution network security and reliability besides Systems," IEEE Transactions on Sustainable Energy, 2017.
[12] S. R. Deeba, R. Sharma, T. K. Saha, D. Chakraborty, and A.
minimizing operational cost. The simulation results showed Thomas, "Evaluation of technical and financial benefits of
that the proposed strategy obtained the reasonable and high- battery-based energy storage systems in distribution networks,"
quality schedules for switching, batteries charging and IET Renewable Power Generation, vol. 10, no. 8, pp. 1149-
discharging, and the active power values of diesel generators 1160, 2016.
in both single-objective and multi-objective frameworks. [13] M. Falahi, S. Lotfifard, M. Ehsani, and K. Butler-Purry,
Furthermore, the exact energy not supplied index and voltage "Dynamic model predictive-based energy management of DG
integrated distribution systems," IEEE Transactions on Power
stability index as separate objective functions are considered Delivery, vol. 28, no. 4, pp. 2217-2227, 2013.
to have an optimal operation in a reliable and secure [14] K. Khawaja, S. U. Khan, S.-J. Lee, Z. M. Haider, M. K. Rafique,
environment. Numerical results for various cases were and C.-H. Kim, "Optimal sizing and allocation of battery energy
performed to demonstrate the ability of the proposed strategy storage systems with wind and solar power DGs in a distribution
in achieving the optimal solutions from the perspective of the network for voltage regulation considering the lifespan of
DISCO. The proposed distribution network voltage stability batteries," IET Renewable Power Generation, 2017.
[15] B. Lin and W. Wu, "Economic viability of battery energy
assessment using PV and QV curves analysis distinguished the storage and grid strategy: A special case of China electricity
proposed study from other studies in this area. market," Energy, vol. 124, pp. 423-434, 2017.
[16] Z. Li, S. Jazebi, and F. De León, "Determination of the optimal
V. REFERENCES switching frequency for distribution system reconfiguration,"
IEEE Transactions on Power Delivery, vol. 32, no. 4, pp. 2060-
[1] E. Dall'Anese, S. V. Dhople, and G. B. Giannakis, "Optimal
2069, 2017.
dispatch of photovoltaic inverters in residential distribution
[17] F. Ding and K. A. Loparo, "Feeder reconfiguration for
systems," IEEE Transactions on Sustainable Energy, vol. 5, no.
unbalanced distribution systems with distributed generation: a
2, pp. 487-497, 2014.
hierarchical decentralized approach," IEEE Trans. Power Syst,
[2] A. Azizivahed, E. Naderi, H. Narimani, M. Fathi, and M. R.
vol. 31, no. 2, pp. 1633-1642, 2016.
Narimani, "A new bi-objective approach to energy management
[18] H. Khazaei, B. Vahidi, S. H. Hosseinian, and H. Rastegar,
in distribution networks with energy storage systems," IEEE
"Two-level decision-making model for a distribution company
Transactions on Sustainable Energy, vol. 9, no. 1, pp. 56-64,
in day-ahead market," IET Generation, Transmission &
2018.
Distribution, vol. 9, no. 12, pp. 1308-1315, 2015.
[3] R. Li, W. Wang, and M. Xia, "Cooperative planning of active
[19] R. Billinton and R. N. Allan, Reliability evaluation of
distribution system with renewable energy sources and energy
engineering systems. Springer, 1992.
storage systems," IEEE Access, vol. 6, pp. 5916-5926, 2018.
[20] M. Gitizadeh, A. A. Vahed, and J. Aghaei, "Multistage
[4] W. Yi, Y. Zhang, Z. Zhao, and Y. Huang, "Multiobjective
distribution system expansion planning considering distributed
Robust Scheduling for Smart Distribution Grids: Considering
generation using hybrid evolutionary algorithms," Applied
Renewable Energy and Demand Response Uncertainty," IEEE
energy, vol. 101, pp. 655-666, 2013.
Access, vol. 6, pp. 45715-45724, 2018.
[21] D. A. Bell, Fundamentals of Electric Circuits: Lab Manual.
[5] M. R. Narimani, A. A. Vahed, R. Azizipanah-Abarghooee, and
Oxford University Press, Inc., 2009.
M. Javidsharifi, "Enhanced gravitational search algorithm for
[22] M. Chakravorty and D. Das, "Voltage stability analysis of radial
multi-objective distribution feeder reconfiguration considering
distribution networks," International Journal of Electrical
reliability, loss and operational cost," IET Generation,
Power & Energy Systems, vol. 23, no. 2, pp. 129-135, 2001.
Transmission & Distribution, vol. 8, no. 1, pp. 55-69, 2014.
[23] J. Aghaei, K. M. Muttaqi, A. Azizivahed, and M. Gitizadeh,
[6] Q. Peng, Y. Tang, and S. H. Low, "Feeder reconfiguration in
"Distribution expansion planning considering reliability and
distribution networks based on convex relaxation of OPF," IEEE
security of energy using modified PSO (Particle Swarm
Transactions on Power Systems, vol. 30, no. 4, pp. 1793-1804,
Optimization) algorithm," Energy, vol. 65, pp. 398-411, 2014.
2015.
 11

[24] M. Barani, J. Aghaei, M. A. Akbari, T. Niknam, H. Farahmand,

and M. Korpås, "Optimal Partitioning of Smart Distribution
Systems into Supply-Sufficient Microgrids," IEEE Transactions
on Smart Grid, 2018.
[25] A. J. Conejo, F. J. Nogales, and J. M. Arroyo, "Price-taker
bidding strategy under price uncertainty," IEEE Transactions on
Power Systems, vol. 17, no. 4, pp. 1081-1088, 2002.
[26] T. Niknam, M. Zare, and J. Aghaei, "Scenario-based
multiobjective volt/var control in distribution networks
including renewable energy sources," IEEE Transactions on
Power Delivery, vol. 27, no. 4, pp. 2004-2019, 2012.
[27] L. Wu, M. Shahidehpour, and T. Li, "Stochastic security-
constrained unit commitment," IEEE Transactions on Power
Systems, vol. 22, no. 2, pp. 800-811, 2007.
[28] M. M. Eusuff and K. E. Lansey, "Optimization of water
distribution network design using the shuffled frog leaping
algorithm," Journal of Water Resources planning and
management, vol. 129, no. 3, pp. 210-225, 2003.
[29] D. Zhang, Z. Fu, and L. Zhang, "An improved TS algorithm for
loss-minimum reconfiguration in large-scale distribution
systems," Electric Power Systems Research, vol. 77, no. 5-6, pp.
685-694, 2007.