Professional Documents
Culture Documents
Energy Storage Planning in Electric Power Distribution Networks - A Stateof-The-Art Review
Energy Storage Planning in Electric Power Distribution Networks - A Stateof-The-Art Review
A R T I C L E I N F O A BS T RAC T
Keywords: In the past decade, energy storage systems (ESSs) as one of the structural units of the smart grids have
Distribution network experienced a rapid growth in both technical maturity and cost effectiveness. These devices propose diverse
Energy storage planning applications in the power systems especially in distribution networks. Despite offering numerous applications,
Energy storage allocation the ESSs are new devices characterized by high investment costs. Besides technological advancement, optimal
Optimal sizing
ESS planning and scheduling is one of the effective ways to reduce the costs and justifying high investment costs
Optimal sitting
by taking their benefits out as much as possible. During the past few years, various studies have been conducted
by the researcher to address the problem of optimal ESS planning in distribution networks. In this context,
various models, methods, and considerations have been proposed to enhance the functionality of optimal
planning process. The aim of this paper is to review the problem of optimal ESS planning including optimal bus
location, power rating, and energy capacity determination in the distribution networks. In order to facilitate
continuing and growing research in this field, a comprehensive literature survey and classification of the related
studies followed by research gaps and future opportunities is provided.
⁎
Correspondence to: Department of Electrical Engineering, Kermanshah University of Technology, P.O. Box: 63766-67178, Kermanshah, Iran.
E-mail addresses: h.saboori@kut.ac.ir (H. Saboori), r.hemmati@kut.ac.ir (R. Hemmati), smsghiasi@aut.ac.ir (S.M.S. Ghiasi), shahab.dehghan@ieee.org (S. Dehghan).
http://dx.doi.org/10.1016/j.rser.2017.05.171
Received 14 December 2016; Received in revised form 28 April 2017; Accepted 20 May 2017
1364-0321/ © 2017 Elsevier Ltd. All rights reserved.
H. Saboori et al. Renewable and Sustainable Energy Reviews 79 (2017) 1108–1121
technical limitations of the network and the ESSs. ESS planning in distribution networks, in addition to the contributions
This new optimization problem has been attracting more attention of the reviewed works, the review content is classified as follows. After
over the last years [21–51] and a variety of models, methods, and this introduction and in Section 2, various commercial ESS technolo-
consideration has been developed. In this context, this paper reviews gies and modeling details used by the researchers in the planning
the problem of optimal ESS planning in distribution networks. It problem are evaluated. In Section 3, those applications of the ESSs
should be noted that in the problem in hand the planning means not which are employed in the problem and the associated objective
only sizing in terms of power rating and energy capacity but also sitting functions are considered. Joining ESS planning with other solutions
of the ESSs in a distribution network with buses and lines. The ESS in the network, various objective types in terms of single or multi-
planning in a single bus network (single node model) which refers to objective, and also multi-stage planning concept are also dealt with in
only ESS sizing is not the focus of this paper because this problem has this section. Network modeling in addition to the solution methods and
already been widely reviewed [52–55]. The work in [56] performs a uncertainty modeling and management issues are investigated in
review of ESS planning in distribution networks, but, this work has Section 4. Finally, Section 5 offers conclusion remarks of the work.
been merged ESS planning in the network and ESS planning in the
microgrid. It should be kept in mind that most of the ESS applications 2. Energy storage technologies and modeling for planning
in the distribution network including voltage control, loss reduction,
congestion alleviation, and network expansion deferral will be achieved 2.1. Energy storage technologies
only in a network constrained framework. Therefore, combining this
problem with ESS planning in microgrid results in lessening the details Energy storage systems (ESSs) in the electric power networks can
of the work. In addition, numerous new works have been published be provided by a variety of techniques and technologies. The ESS
after that works. In this regard, this paper offers a detailed and updated technologies include pumped hydraulic storage (PHS), compressed air
review of the network constrained ESS planning in distribution net- energy storage (CAES), flywheel energy storage (FWES), superconduct-
work. ing magnetic energy storage (SMES), battery energy storage system
To this end, high quality research works are surveyed and classified (BESS), and supercapacitor or ultracapacitor energy storage (SCES).
in details in order to facilitate continuing research in this field. Table 1 Today, these technologies are very popular and used for various
shows list of the reviewed works together with the publication date and applications in the power systems [57].
the associated journal title. As the table shows, number of the In general, storage systems are categorized based on two factors
publications is increased year by year. The growing number of the namely storage medium (type of the energy stored) and storage
works demonstrates importance of the topic. (discharge) duration. In the first type classification, the ESSs are
Considering the components of the optimization problem in hand, divided to mechanical, chemical, and electrical storage systems based
on the form in which the energy is stored. The mechanical energy
storage can also be divided into kinetic and potential. Also, the
Table 1
electrical energy storage is divided into electrostatic and magnetic
List of the reviewed works.
energy storage. The pumped hydraulic storage and compressed air
Ref # Date of Publication Journal Title energy storage, flywheel energy storage, ultracapacitor, superconduct-
ing magnetic energy storage, and battery energy storage are belong to
[21] Nov 2010 IEEE Transactions on Power Systems
potential mechanical, kinetic mechanical, electrostatic electrical, mag-
[22] Jun 2013 IEEE Transactions on Smart Grid
[23] Jan 2014 IEEE Transactions on Power Systems netic electrical, and chemical storage categories, respectively.
[24] Jul 2014 International Journal of Electrical Power and Base on the second criterion, i.e. storage or discharge duration,
Energy Systems storage technologies are divided into two main categories including
[25] Sep 2014 IEEE Transactions on Power Systems long-term and short-term storage. The long-term storage, also known
[26] Sep 2014 Electronics Letters
[27] Sep 2014 IEEE Transactions on Smart Grid
as centralized bulk storage, large-scale, or grid-scale storage, is a
[28] Jan 2015 IEEE Transactions on Power Systems relatively large storage installation suitable for storing large amounts
[29] Nov 2015 Energy Conversion and Management of electricity. In this case, the storage capacity ranges from a few to
[30] Nov 2015 IEEE Transactions on Smart Grid hundreds of megawatts and the unit can supply power to the grid with
[31] Nov 2015 International Journal of Electrical Power and
discharge durations more than 8 h. The pumped hydro, compressed air
Energy Systems
[32] Dec 2015 Energy energy storage, and large-scale batteries belong to this category.
[33] Jan 2016 IEEE Transactions on Power Systems Considering the long discharge duration and energy capacity, this type
[34] Apr 2016a IEEE Transactions on Smart Grid of storage is fitted to the long-term energy management applications
[35] such as energy arbitrage, congestion management, expansion deferral,
[36] Apr 2016 Journal of Renewable and Sustainable
Energy
and long term voltage control. Second type of storage in this classifica-
[37] May 2016 IEEE Transactions on Power Systems tion is known as short-term or transient storage. The flywheel energy
[38] May 2016 IET Generation, Transmission & storage, superconducting magnetic energy storage, ultracapacitor, and
Distribution small-scale batteries fit in this category. Considering short-term
[39] Aug 2016 IEEE Transactions on Smart Grid
response, this type of storage is suitable for frequency regulation,
[40] Aug 2016 IEEE Transactions on Smart Grid
[41] Sep 2016 IET Renewable Power Generation short-term voltage control, transient renewable energy smoothing, and
[42] Oct 2016 Energy power quality improvement.
[43] Dec 2016 Applied Energy The energy storage used in the distribution networks should met
[44] Jan 2017a IEEE Transactions on Sustainable Energy some specific requirements in this network. Implementation of the
[45]
[46] Jan 2017 IEEE Transactions on Sustainable Energy
large-scale storage plants like pumped hydro storage and compressed
[47] Mar 2017 Applied Energy air energy storage involve special geographical and footprint require-
[48] Mar 2017 Journal of Modern Power Systems and Clean ments which cannot be achieved in distribution networks. Also, short-
Energy term storage technologies including flywheel, ultracapacitor, and
[49] Apr 2017 Solar Energy
superconducting magnetic energy storage are characterized by short
[50] May 2017 Renewable and Sustainable Energy Reviews
[51] Jul 2017 Electric Power Systems Research discharge duration and very low energy density. Therefore, these
technologies cannot provide long-term and high energy density re-
a
Two-part publication. quirements of the energy management solutions in the distribution
1109
H. Saboori et al. Renewable and Sustainable Energy Reviews 79 (2017) 1108–1121
Table 2
Economical characteristics of the ESS technologies [72].
Per unit of power rating (€/KW) Per unit of energy rating (€/KWh)
network. Table 3
In contrast, with the advancement of the high power and high Technical characteristics of the ESS technologies [72].
energy density, high efficiency, environmental friendly and grid scale
ESS Technology Configuration Round Trip Life Time
batteries, these devices are becoming one of the most potential storage Efficiency (Year)
technologies suitable for employment in the distribution networks.
Nowadays, the BESS have been adequately developed to reach a Min Max Min Max
reasonable level of technical maturity and cost competiveness making
Pumped Hydro Conventional 0.70 0.82 50 60
them the first option for storage applications in distribution networks Compressed Air Underground 0.70 0.89 20 40
[57]. Aboveground 0.70 0.90 20 40
Flywheel High Speed 0.93 0.95 15 20
Battery Lead-acid 0.70 0.90 5 15
2.2. Energy storage modeling Na-S 0.75 0.90 10 15
ZEBRA 0.86 0.88 15 15
The ESSs are characterized by various technical and cost data. Most Li-ion 0.85 0.95 5 15
of these properties can be defined almost for all types of the ESSs. VRFB 0.65 0.85 5 10
Zn-Br 0.60 0.70 5 10
Power density, energy density, life-time in cycles and years, self-
Fe-Cr 0.72 0.75 10 15
discharge rate, round-trip efficiency, storage durability, and discharge Ni-Cd 0.60 0.73 10 20
duration are of the common characteristics of the storage units. Also, PSB 0.65 0.85 10 15
cost data can be stated as installation cost per power rating, installation Zn-air 0.60 0.70 5 10
cost per energy capacity rating, operation and maintenance cost, and Supercapacitor Double Layer 0.85 0.95 10 20
SMES SMES 0.95 0.98 15 20
replacement cost [58]. Tables 2 and 3 represent typical technical and
Hydrogen Fuel Cell 0.33 0.42 15 20
economic data for current ESS technologies installed in the power Gas Turbine 0.33 0.42 15 20
systems [58]. The tables contain required information in order to use in
a planning problem. Detailed technical and cost data can be found in
Ch Ch
[58]. gross power discharged from the ESS will be equivalent to PESS / ηESS .
In general, operation mechanism of an ESS can be modeled by its This issue should be regarded in the ESS power and energy balance
power conversion unit (PCS) and storage unit, as depicted by Fig. 1 equations in bus at which the ESS will be installed. Eq. (1) simplifies
[58,59]. Both the technical and cost modeling of the ESS can be made the above issue and revises the power balance equation in order to
by this simple figure. As in the figure, a quantity of the charged power account for the ESS installation in the bus.
into and discharged power from the ESS will drop in the conversion
Di Ch
devices. The percent of the charge efficiency imposed by the power PG + PESS = PESS + PD + PI (1)
conversion unit in charging mode defines the relation between power
drawn from the grid and net power charged into the ESS. Strictly As in (1), an ESS unit can be modeled as a controllable load and a
Ch
speaking, if PESS Ch
and ηESS refer to charge power drawn from the bus and generating unit to account for charge and discharge states, respectively.
Ch Ch
the power conversion unit efficiency in charging mode, then the net In this equation, variables PESS and PESS represent charging power
power charged into the ESS will be equivalent to PESS Ch Ch
× ηESS . drawn from and discharging power injected to the bus by energy
Similar to the charge state, discharge state affects from the power storage s installed in bus i and at time period t, respectively, while PG ,
conversion unit in discharging mode. It is worth mentioning that, gross PD , and PI stand for generated power in the bus, bus demand, and
discharged power drawn from the ESS is equal to the net discharging injected power to the lines connected to the bus, respectively.
power injected to the grid at ESS installation bus in addition to the Besides charging and discharging losses in the power conversion
power conversion unit losses in discharging mode. In other words, if unit, some of the power stored in the storage unit will lost by the time.
Di
PESS Di
and ηESS stand for net discharge power injected to the bus and This effect is known as self-discharge and its amount is a function of
discharging power conversion unit efficiency in discharging mode, then storage technology and storage durability.
1110
H. Saboori et al. Renewable and Sustainable Energy Reviews 79 (2017) 1108–1121
1111
H. Saboori et al. Renewable and Sustainable Energy Reviews 79 (2017) 1108–1121
Table 4
ESS technologies and modeling details.
Ref # ESS technology Technology selection Life cycle (battery) ESS reactive power
this context, more detailed models can be used to better account for
the capability curve of the active/reactive powers of the system.
discharge action. This will result in flattening the load profile, shaving
the peak of the load profile, and consequently lowering cost of the CG = α + β PG + γ PG2 (7)
supplying load in the entire time period of the operation. The arbitrage
or load leveling practice is also referred as peak shaving along with In this equation,PG and CG represent power production and
valley filling, as demonstrated by Fig. 2 [60]. associated generation cost, respectively. Also,α , β , and γ denote first,
The required energy for charging the ESS during low demand second, and third order cost function coefficients, respectively. In order
periods come from the upstream network (HV/MV substation) in the to relax the nonlinearities of this function and enhance convergence
conventional distribution networks and/or the other sources in the process of the solution method, it is a common practice to use a piece-
active distribution networks. The active distribution network denotes a wise linear approximation as illustrated by Fig. 3. As the figure shows,
situation in which a distribution network contains small-scale dis- the quadratic cost function is approximated with a sequence of straight
patchable fuel fired (micro-turbine) and/or non-dispatchable renew- line segments. The details of such an approximation can be found in
able distributed generation (DG) resources. [60].
It should be noted that the reason of the obtained cost saving from Network congestion is a condition in which the existing distribution
the load leveling is the quadratic from of the cost function of the feeders are unable to accommodate connected load which is usually
electric power production. Considering the high share of thermal power occurs during periods of high demand or during emergency conditions
generation in the system, electric power production is a function of the [61]. Network congestion impacts reliability considerably since if
fossil fuel fired in the form of a quadratic function as stated by (7). feeders are congested and operated at (or near) their thermal limits,
1112
H. Saboori et al. Renewable and Sustainable Energy Reviews 79 (2017) 1108–1121
1113
H. Saboori et al. Renewable and Sustainable Energy Reviews 79 (2017) 1108–1121
(compensate forecast errors) [23] and also to procure reserve power in network upgrade and/or expansion planning which refers to define
the network [38]. the location and capacity of the new added or upgraded lines to cope
Abovementioned applications can be achieved, quantified, and with load growth and network congestion considering minimum
calculated by steady-state balanced studies. These applications are investment cost. This problem is investigated and well-reviewed in
considered as a goal to plan ESSs in the distribution networks by the conventional and active distribution networks [77–80].
network operation frameworks namely optimal power flow (OPF) and After that and with the purpose of improving network voltage
unit commitment (UC). Other applications of the ESSs in the distribu- profile, planning of the voltage source devices like capacitors and
tion network are related to nodal and/or transient short-term studies DFACTS (distribution flexible AC transmission systems) devices have
or unbalance network models which cannot combine with steady-state been the focus of the researchers [81]. In capacitor (or other voltage
balance applications including short-term renewable energy smoothing control devices) planning problem in the distribution network, the
[66,67], low-voltage and fault ride-through [68,69], load following and objective is to flatten voltage profile of the buses around one per-unit
regulation [70], voltage unbalance mitigation [71], short-term power with minimum planning costs.
quality improvement [72,73], and neutral-to-earth voltage (NEV) In the past decade and with the advent of small-scale local
mitigation [74,75]. generation resources in distribution networks, known as distributed
generation (DG), a new planning problem in distribution networks,
3.2. Planning objectives namely DG planning, has been emerged. This problem deals with
finding the optimal location and capacity of the fuel-fired and/or
The abovementioned applications of the ESSs are achieved by renewable DG sources in order that the desired objectives achieved.
defining objective function of the proposed model. In other word, the The objectives in this problem may consist of minimum power
defined objective function should capture considered ESS applications. generation cost, network losses reduction, flattening the voltage profile,
The objective function in the conventional vertically integrated systems and reliability [82–86].
is usually expressed as a minimization of sum of the several cost terms With this outlook, a number of the works have been coordinated the
as denoted by (10). planning of the ESSs with other equipment and solutions in the
network [22,26,27,29,34,35,37,42,44–46,50,51]. This practice can
Obj = Min (NPC ) (10)
potentially reduce overall cost of the planning multiple devices
In this equation, NPC refers to the total net present cost of the especially in the case in which the ESS cannot cope with the network
operation and planning costs in the network. The NPC can be defined shortages solely. In this context, ESS planning is jointed with the
based on a daily, yearly, or even life-time horizon. In the deregulated capacitor planning [22], switch placement [26], load shedding [27],
market environments, the objective function is expressed as a max- network expansion [29], DG allocation in addition to the network
imization of the distribution company (DISCO) net profit as stated by planning (feeder, substation, and transformer) [34,35], on-load tap
(11). changer (OLTC) adjustment as well as DG power factor (PF) control
[37], capacitor planning along with the OLTC adjustment [42], DG
Obj = Max (NPI −NPC ) (11)
planning, network expansion, and capacitor planning [44,45], repla-
In this equation, NPI denotes net present income from selling cing or adding lines (network reinforcement or expansion) [46], DG
energy to customers equal to the sold power multiplied by its price for planning [50], and switch placement [51].
each time period. It is worth mentioning that NPI and NPC in this It should be noted that, in the case of joint planning, investment
equation should be defined at a same operation horizon, usually daily and O & M costs associated with the other new installed devices apart
or yearly. from ESSs should be included in the objective function.
The NPC which denotes total cost of the system contains identical The objective function terms explained in the above can declare as a
cost terms for the both operation environments. The cost terms are single or multi-objective optimization problem. In the case of single-
related to the considered applications for the ESSs in addition to the objective modeling, all contributing terms in the objective function are
ESS itself, explained in the following. of identical type, to be exact, cost and based on a monetary value. The
Considering the nature of the problem, ESS planning, investment objective function in this case can be expressed as (12).
and operation and maintenance (O & M) costs are the main cost terms
whish should be considered in the problem. The ESSs investment and Min C Tot = ∑ (α CiInv + β CiO & M )
i (12)
O & M cost should be add up taking into account their different time
horizons. The investment cost is a life-time or one-shot cost while O & C Tot
In (12), , CiInv , CiO & M ,α ,
and β represent total planning cost for
M cost is usually expressed as a yearly cost. As denoted in Section 2, the desired planning horizon, total planning investment cost, yearly
appropriate coefficients should be used when adding these different planning O & M cost, and proper coefficients in order to convert total
cost terms. investment and yearly O & M cost to the desired planning horizon.
Other contributed cost terms in the objective function are asso- In contrary to single-objective planning, multi-objective modeling
ciated with the considered ESS applications in the network which are contains various contributing terms in the objective function with
explained in the previous subsection. These cost terms include opera- different and usually conflicting economic and technical nature. A
tion cost or energy purchase (from up-stream substation or local micro- limited number of the works employ a multi-objective framework for
turbines) [21–51], reliability index or load curtailment cost including considering various conflicting objectives [24,31,38,51]. In most of
ENS [32,33,42,43,48], EENS [27,30,35], MAIFI [51], and SAIDI [51], these works, a weighed sum approach, i.e., analytic hierarchy process
penalty factors for voltage and line flow limit devia- (AHP), is used to convert the problem to a single-objective one
tions [24,25,27,31,33,42,43,49], network losses [21– [24,28,31]. Also, in [51] a Pareto Front approach is used to handle
23,25,26,28,30,38,39,41,47], emission cost [26,41,44,45,47,48], and the multi-objective model.
curtailment cost [21,26,37,41]. It ought to be noted that all of the cost Finally, the optimization problem can be formulated as a single-
terms included in the objective function irrespective of their original stage or multi-stage problem. In the single-stage planning only
time horizon should be converted to a base operation horizon, i.e., location, power rating, and energy capacity of the ESSs will be defined
daily, yearly, or life-time by using proper coefficients [76]. optimally [21–28,31–33,36–43,46–51]. In the multi-stage planning
Planning of the ESSs in the distribution network can be combined the optimal time of the installations is also determined among a
with the planning of the other equipment, devices, and solutions. First predefined planning stages usually years [29,30,34,35,44,45].
and primary planning problem in the distribution network is the Table 5 shows a summary of the considered ESS applications and
1114
H. Saboori et al. Renewable and Sustainable Energy Reviews 79 (2017) 1108–1121
Table 5
ESS applications and objective function terms and considerations.
Ref # ESS application Objective function Objective Type Joint planning with Multi-Stage
Planning
1115
H. Saboori et al. Renewable and Sustainable Energy Reviews 79 (2017) 1108–1121
Table 5 (continued)
Ref # ESS application Objective function Objective Type Joint planning with Multi-Stage
Planning
1116
H. Saboori et al. Renewable and Sustainable Energy Reviews 79 (2017) 1108–1121
Table 5 (continued)
Ref # ESS application Objective function Objective Type Joint planning with Multi-Stage
Planning
VAR support
[48] Arbitrage Operation (energy purchase) Single objective – Single-Stage
Emission reduction cost
Reliability improvement ESS investment and OM cost
Network expansion deferral ESS replacement (life cycle) cost
Emission cost
Reliability cost (ENS)
Expansion deferral
[49] Voltage support ESS investment cost Single objective – Single-Stage
Voltage limit deviation cost
[50] Arbitrage Operation (energy purchase) Single objective DG Planning Single-Stage
cost
ESS investment and OM cost
[51] Reliability improvement ESS investment cost Multi Switch placement Single-Stage
Reliability cost (MAIFI and Objective
SAIDI) (Pareto Front)
objective function terms and considerations in the problem. genous ESS applications is another way to cover their high invest-
ment cost. Various multi-objective models for instance Pareto Front
3.3. Future works or ε-constraint models are not focused adequately.
– New joint planning models of the ESSs can be proposed including
As stated above, various applications of the ESSs are considered in adding new and/or replacing overhead conductor and underground
the planning process by defining related objectives. In addition, some cable feeders, transformers, capacitors, phase-shifters, auto-boos-
consideration like active networks, multi-stage planning, multi-objec- ters, DFACTS, PHEV charging lots, dispatchable and non-dispatch-
tive planning, and joint planning are proposed by the researchers. able DGS, and other conventional and new solutions and devices in
Suggestions for the future works in this context can be declared as medium and low voltage levels.
follows.
4. Network modeling, solution methods, and uncertainty
– Employing multiple synergic applications of the ESSs as much as management
possible will augment their cost effectiveness compared to the other
solutions in the network. For instance, application of the ESSs to act 4.1. Modeling of the network
as a backup source in island mode operation of the network is less
addressed. In order to constitute a network-constrained ESS planning pro-
– Developing new multi-objective models from synergic non-homo- blem, a model based on the power flow (PF) equations should be
developed. In general, two main power flow equations are developed
and used in the literature namely AC and DC power flow equations. In
AC power flow equations, besides active power, bus voltages, reactive
power, and network losses will be modeled and calculated as a non-
linear model resulting to an iterative solution method at the expense of
high execution time and risk of divergence. In contrary, DC power flow
equations, ignore network losses and consider bus voltages equal to
one per-unit resulting to ignoring reactive power. These approxima-
tions mean that only active power will be modeled and calculated
through a linear model, very fast, and non-iterative solution [87].
Although some DC power flow models are modified to consider
network losses or voltages while preserving linear nature, but, their
results are still far from AC power flow ones [88,89]. Each power flow
model possesses its pros and cons for the specific applications in
operation or planning problems.
In the reviewed problem in this work, ESS planning in distribution
network, the AC power flow model in the form of full or linear
approximated is used in the most of the works. The reason is behind
the fact that calculating some applications of the ESS requires knowing
bus voltages (for instance voltage support application) or network
losses (for instance loss reduction application) and this will be achieved
only via an AC power flow model. The works presented in [34,35,44–
46] use a DC power flow model which is enhanced with linear
expressions for voltage and losses, but, the considered approximated
models contain some errors compared to the full AC power flow as
stated by the authors.
Fig. 5. Flowchart for solving constrained optimization problems by Meta-heuristic Some works use a specific AC power flow model for radial
optimization algorithms. distribution networks known as DistFlow equations and converted it
1117
H. Saboori et al. Renewable and Sustainable Energy Reviews 79 (2017) 1108–1121
Table 6
Power flow, model types, and solution methods.
voltage losses B LP heuristic classic size Uncertain parameters Uncertainty management method
U NLP
to a second order cone programming model (SOCP) [25,31,38,40]. The 4.2. Solution methods and uncertainty management
SOCP problem is a type of non-linear optimization problems in which a
linear objective function will be minimized over the intersection of an It should be noted that utilizing AC or DC power flow model will
affine-linear manifold and the product of second order cones (quadratic result in a non-linear or linear model, respectively. The type of the
functions). This problem can be solved by efficient primal-dual interior model, non-linear or linear, will have direct effect of the solution
point methods. method options, in turn. The models presented in [34,35,44–46] are
In a typical distribution network impedances of the lines are usually the only proposed linear models which consequently uses a commercial
unequal because of the uneven spaces between the conductors in the mixed integer linear programming (MILP) solver, namely CPLEX.
distribution poles and also lack of the transposition practice. In Other works uses a non-linear model to formulate the problem which
addition, even if distribution lines be balanced ideally, as a conse- several of them adopt a Meta-heuristic solution method [22–24,26–
quence of supply various single phase loads the network is unbalance. 30,32,33,36,39,41–43,47–51] and others a classic optimization solver
To address this issue in the problem, an unbalance three phase power [25,31,34,35,37,38,40,44–46]. Meta-heuristic optimization algorithms
flow model should be utilized. All reported works in the field of ESS are optimization techniques that mainly work based on the random and
planning in distribution networks uses a balance single phase power iterative procedures, as shown in Fig. 5. The algorithm starts with
flow model result in ignoring this matter. generating a random initial population. The initial population is a
matrix and each row of the matrix is called a particle or chromosome.
1118
H. Saboori et al. Renewable and Sustainable Energy Reviews 79 (2017) 1108–1121
Each particle indicates a possible solution of the problem. At next step, 5. Conclusion remarks
algorithm selects one particle in the population and objective function
of the problem is calculated for this particle. Then constraints of the The problem of the ESS planning in distribution networks has been
problem are checked for current particle. If the constraints are gained an increasing attention by the researchers in the past decade.
dissatisfied, current particle is uninvolved and next one is evaluated. Reviewing the current state of the research in this field can potentially
This procedure is repeated to calculate objective function for all ease future works by classifications of the problem components and
particles in the population. Finally, the convergence criterion is offering research gaps. This is done in the current work by collecting
checked. If the convergence condition is met, the algorithm is over related high quality research papers. Then, the works are classified
and final solution is achieved. Otherwise, the algorithm iterates until based on the used energy storage technologies and models, considered
finding optimal result. applications for the storage systems and associated objective functions,
Various Meta-heuristic methods are used to solve the problem of network modeling, solution methods, and uncertainty management of
the ESS planning in distribution networks including genetic algorithms the problem. Each section is equipped with relevant future works for
(GA) [22,24,26–28,30,43,47,49,51], particle swarm optimization those who are interested in the field. Considering broad scope of the
(PSO) [23,29,32,33,36,42,50], tabu search (TS) [33,42], simulated applications and benefits of the ESSs in distribution networks, it is
annealing (SA) [24,43], differential evolution (DE) [41], and intelligent expected that in the future, new and more research works will be
single particle optimizer (ISPO) [48]. The characteristics and details of published where this work can constitute a roadmap.
each Meta-heuristic method in the optimization problems especially
distribution planning are out of the scope of this paper and well- References
reviewed in the literature previously [90–93].
One of the important inputs of the planning problems is forecast of [1] Granado Del, Crespo Pedro, Pang Zhan, Wallace Stein W. "Synergy of smart grids
the system parameters for the planning horizon. These forecasts are and hybrid distributed generation on the value of energy storage. Appl Energy
2016;170:476–88.
usually obtained by current values of the parameters and a certain [2] Wade NS, Taylor PC, Lang PD, Jones PR. Evaluating the benefits of an electrical
growth factor. Sensitivity analysis demonstrates that quality of the energy storage system in a future smart grid. Energy Policy 2010;38(11):7180–8.
planning process is quit sensitive to the future value of the system [3] Luo X, Wang J, Dooner M, Clarke J. Overview of current development in electrical
energy storage technologies and the application potential in power system opera-
parameters obtained by foresting. Considering this fact that any tion. Appl Energy 2015;137:511–36.
forecasting method involves some degrees of inherent error, planning [4] Aneke Mathew, Wang Meihong. Energy storage technologies and real life applica-
outcomes may no longer work properly in the real-life systems. In other tions–a state of the art review. Appl Energy 2016;179:350–77.
[5] Vazquez S, Lukic SM, Galvan E, Franquelo LG, Carrasco JM. Energy storage
word, inherent uncertainties in the system parameters in the future like
systems for transport and grid applications. IEEE Trans Ind Electron
load demand, renewable resources generation, electric vehicles pene- 2010;57(12):3881–95.
tration, and so on may influence planning quality significantly. [6] Hadjipaschalis I, Poullikkas A, Efthimiou V. Overview of current and future energy
storage technologies for electric power applications. Renew Sustain Energy Rev
Uncertainty management methods are useful tools to address this
2009;13(6):1513–22.
issue. Therefore, uncertainty management methods are exercised by [7] Koohi-Kamali S, Tyagi VV, Rahim NA, Panwar NL, Mokhlis H. Emergence of energy
the authors in the ESS planning problem in the distribution networks. storage technologies as the solution for reliable operation of smart power systems: a
In this context various methods are used to handle inherent uncer- review. Renew Sustain Energy Rev 2013;25:135–65.
[8] Kloess M, Zach K. Bulk electricity storage technologies for load-leveling operation–
tainty in the system parameters including load, renewable generation An economic assessment for the Austrian and German power market. Int J Electr
(wind and PV), PEV, and characteristics of the ESS systems Power Energy Syst 2014;59:111–22.
[27,28,30,32–35,39–45]. The most used methods are probabilistic [9] Zafirakis Dimitrios, et al. The value of arbitrage for energy storage: evidence from
European electricity markets.". Appl Energy 2016.
power flow and optimal power flow [27,28,30,32,33,42,43], Monte- [10] Zhao H, Wu Q, Hu S, Xu H, Rasmussen CN. Review of energy storage system for
Carlo methods [27,28,30,32,41], point estimate methods [33,42], wind power integration support. Appl Energy 2015;137:545–53.
stochastic programming [34,35,44,45], chance-constrained program- [11] Go Roderick S, Francisco D Munoz, Watson Jean-Paul. Assessing the economic
value of co-optimized grid-scale energy storage investments in supporting high
ming [40,41], fuzzy logic [43], and factor analysis method [43]. renewable portfolio standards. Appl Energy 2016;183:902–13.
Table 6 presents a summary of the abovementioned explanations [12] Díaz-González F, Sumper A, Gomis-Bellmunt O, Villafáfila-Robles R. A review of
about power flow models, model types, solution methods, and un- energy storage technologies for wind power applications. Renew Sustain Energy
Rev 2012;16:2154–71.
certainty management methods for the problem in hand. In this table [13] Abdelrazek SA, Kamalasadan S. Integrated PV capacity firming and energy time
and in column “model”, B,U, LP, and NLP refer to balanced load flow, shift battery energy storage management using energy-oriented optimization. IEEE
unbalance load flow, linear programming model, and nonlinear Trans Ind Appl 2016;52:2607–17.
[14] Rosso ADD, Eckroad SW. Energy storage for relief of transmission congestion.
programming model, respectively.
IEEE Trans Smart Grid 2014;5:1138–46.
[15] Vargas LS, Bustos-Turu G, Larra F. Ed. Wind power curtailment and energy storage
4.3. Future works in transmission congestion management considering power plants ramp rates.
IEEE Trans Power Syst, 30; 2015. p. 2498–506.
[16] de Sisternes , Fernando J, Jenkins Jesse D, Botterud Audun. The value of energy
In the context of the network models, solution methods, and storage in decarbonizing the electricity sector. Appl Energy 2016;175:368–79.
uncertainty management of the problem the following recommenda- [17] Wang Y, Tan KT, Peng XY, So PL. Coordinated control of distributed energy-
tion can be targeted as future works. storage systems for voltage regulation in distribution networks. IEEE Trans Power
Deliv 2016;31:1132–41.
[18] Shen YW, Ke DP, Sun YZ, Kirschen DS, Qiao W, Deng XT. Advanced auxiliary
– Developing new detailed linear network models, i.e. augmented DC control of an energy storage device for transient voltage support of a doubly fed
models or approximated AC models, can help to find optimal induction generator. IEEE Trans Sustain Energy 2016;7:63–76.
[19] Shankar Ravi, KalyanChatterjee , Bhushan Ravi. Impact of energy storage system
solution for realistic and real-life networks with many buses without on load frequency control for diverse sources of interconnected power system in
divergence concerns. deregulated power environment. Int J Electr Power Energy Syst 2016;79:11–26.
– Unbalance models of the network for single-phase ESS planning are [20] Saboori H, Abdi H. Application of a grid scale energy storage system to reduce
distribution network losses. In: Proceedings of the 18th conference on Electrical
not addressed yet. Addressing this issue is the key factor to exploit Power Distribution Networks (EPDC). IEEE; 2013. p. 1–5.
the application of the single-phase ESS to reduce network unbalance [21] Atwa Yasser, Moustafa , El-Saadany EF. Optimal allocation of ESS in distribution
[94] systems with a high penetration of wind energy. IEEE Trans Power Syst
2010;25.4:1815–22.
– New methods of uncertainty management like risk aversion [95] or
[22] Carpinelli Guido, et al. Optimal integration of distributed energy storage devices in
robust optimization [96] can be utilized to better capture the smart grids. IEEE Trans Smart Grid 2013;4.2:985–95.
inherent uncertainties in the problem. [23] Zheng Yu, et al. Optimal allocation of energy storage system for risk mitigation of
DISCOs with high renewable penetrations. IEEE Trans Power Syst
1119
H. Saboori et al. Renewable and Sustainable Energy Reviews 79 (2017) 1108–1121
1120
H. Saboori et al. Renewable and Sustainable Energy Reviews 79 (2017) 1108–1121
[90] Gamarra Carlos, Josep M Guerrero. Computational optimization techniques [94] Chua KH, et al. Energy storage system for mitigating voltage unbalance on low-
applied to microgrids planning: a review. Renew Sustain Energy Rev voltage networks with photovoltaic systems. IEEE Trans Power Deliv
2015;48:413–24. 2012;27.4:1783–90.
[91] Sedghi Mahdi, Ahmadian Ali, Aliakbar-Golkar Masoud. Assessment of optimization [95] Hemmati Reza, Saboori Hedayat, Saboori Saeid. Stochastic risk-averse coordinated
algorithms capability in distribution network planning: review, comparison and scheduling of grid integrated energy storage units in transmission constrained
modification techniques. Renew Sustain Energy Rev 2016;66:415–34. wind-thermal systems within a conditional value-at-risk framework. Energy
[92] Jordehi A Rezaee. Optimisation of electric distribution systems: a review. Renew 2016;113:762–75.
Sustain Energy Rev 2015;51:1088–100. [96] Dehghan Shahab, Amjady Nima. Robust transmission and energy storage expan-
[93] Banos Raul, et al. Optimization methods applied to renewable and sustainable sion planning in wind farm-integrated power systems considering transmission
energy: a review. Renew Sustain Energy Rev 2011;15.4:1753–66. switching. IEEE Trans Sustain Energy 2016;7.2:765–74.
1121