Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

IET Renewable Power Generation

Research Article

Optimal allocation of battery energy storage ISSN 1752-1416


Received on 24th November 2015

systems in distribution networks with high


Revised 4th February 2016
Accepted on 13th March 2016
E-First on 20th May 2016
wind power penetration doi: 10.1049/iet-rpg.2015.0542
www.ietdl.org

Yongxi Zhang1, Zhao Yang Dong1, Fengji Luo2 , Yu Zheng3, Ke Meng1, Kit Po Wong4
1School of Electrical and Information Engineering, University of Sydney, NSW 2006, Australia
2Centre for Intelligent Electrical Networks, University of Newcastle, Newcastle, NSW 2384, Australia
3Electric Power Research Institute, China Southern Power Grid Company, Guangzhou 510080, People's Republic of China
4School of Electrical, Electronic, and Computer Engineering, University of Western Australia, Perth, WA, Australia

E-mail: fengji.luo@newcastle.edu.au

Abstract: In recent years, the battery energy storage system (BESS) has been considered as a promising solution for mitigating
renewable power generation intermittencies. This study proposes a stochastic planning framework for the BESS in distribution
networks with high wind power penetrations, aiming to maximise wind power utilisation while minimise the investment and
operation costs. In the proposed framework, the uncertainties in wind power output and system load are modelled by the
Monte–Carlo simulation, and a chance-constrained stochastic optimisation model is formulated to optimally determine the
location and capacity of BESS while ensuring wind power utilisation level. Then, the Monte–Carlo simulation embedded
differential evolution algorithm is used to solve the problem. Simulation studies performed on a 15-bus radial distribution system
prove the efficiency of the proposed method.

Nomenclature ��emi emission cost for micro-turbine units ($)


Φ set of all the nodes ��grid cost for purchasing power from substation ($)
Lbat potential location of the BESS ��loss system power loss cost ($)
t, T index and total number of time intervals BESS life cycle depression cost ($)
Δt time duration of each time interval ��depre
S, ξ index and total numbers of scenarios ai, bi cost coefficient for ith micro turbines units ($/MW)
N total numbers of wind turbines Ccur wind power curtailment cost per unit ($/MW)
M total numbers of micro-turbine units Cbuy cost for buying power from substation per unit
D total numbers of buses ($/MW)
vin, vr, vout cut-in, cut-out and rated wind speed (m/s) δ cost coefficient of the BESS lifetime depression
fl(L) load demand forecast error distribution ($/MW)
qm micro turbine units outage rate loss cost for power loss per unit ($/MW)
Pwi wind power output at node i (MW) Nr maximum life time of BESS (Year)
PDj load demand at node i (MW) d interest rate for BESS project
Prated,i rated wind power output of ith wind turbine (MW) ch, dis storage charging/discharging efficiency
Iij current output of line ij (A) CAP capital investment budget for BESS installation ($)
Pbat,max rated power rating of BESS (MW)
vij voltage of line ij
Cbat,max rated energy capacity of BESS (MWh)
PGi,min minimum active power output of ith MT units (MW) �
�ch charging power of BESS at time t for scenario ξ
PGi,max maximum active power output of ith MT units (MW) ,�
(MW)
QGi,min minimum reactive power output of ith MT units discharging power of BESS at time t for scenario ξ
��dis, �
(MW) (MW)
QGi,max maximum reactive power output of ith MT units � BESS energy capacity at time t for scenario ξ (MWh)
�bat
(MW) ,�

Imin minimum current output of line ij (A)



�cur ,�
wind power curtailment at time t for scenario ξ (MW)
Imax maximum current output of line ij (A) �
�buy ,�
power supplied by the substation at time t for scenario
vmin minimum voltage limitation (v) ξ (MW)
vmax maximum voltage limitation (v)

���,�
power output of ith micro turbines units at time t for
maximum power flow constraint scenario ξ (MW)
�max

SOC��, � state of charge of ith BESS units at time t for scenario
Cinv capital cost of BESS energy capacity per day ($) ξ
Copm operation and maintenance cost per day for BESS ($) �
�loss power loss at time t for scenario ξ (MW)
,�
Qop expected daily operation cost ($) αi, βi confidence level of chance constraint
CO fixed operation and maintenance cost per unit ($/
MWh)
CE capital cost of BESS per unit ($/MWh) 1 Introduction
��gen micro turbines operation cost ($) 1.1 Background
��pen penalty cost of the wind curtailment ($) Wind power penetration level has been growing rapidly in modern
power systems. Due to its inherent intermittent and stochastic
IET Renew. Power Gener., 2016, Vol. 10 Iss. 8, pp. 1105-1113 1105
© The Institution of Engineering and Technology 2016
energy as possible. Therefore, integration BESS to reduce wind
power curtailment worth further study. In [9], the improvement in
wind power utilisation after installing ESS was investigated.
Although different wind penetration levels are discussed, the
seasonal characteristics of wind power as well as the load variation
are neglected. In [11], the BESS capacity was determined on the
basis of worst-case analysis aiming to absorb all the excess wind
energy. Since the worst case scenario is low probability high
consequence event, this approach would lead to the uneconomical
and inefficiency installation of the BESS. The improvements of
renewable power injection on to a distribution network after ESS
integration was discussed in [15] based on deterministic model.
Regarding the uncertainties in wind power, electric demand, the
latest research shows a clear transition from the traditional
deterministic approaches to the advanced stochastic methods. It is
Fig. 1 Schematic of the distribution network considered in this paper natural for us adopting stochastic approach in BESS planning.
Moreover, the impacts of BESS integration on DG operations are
natures, the integration of wind power could lead to supply– ignored in aforementioned works. As a result, in this paper, the DG
demand imbalance, and subsequently brings significant challenges units operational cost and power loss cost are taken into
to the secure and reliable operations of power systems [1]. One consideration as well.
promising solution to alleviate the renewable energy
intermittencies is to use the energy storage systems (ESSs). The 1.3 Major contributions of this paper
applications of ESS could enhance power system reliability and
stability while providing other benefits, including better loading The major contribution of this paper is to propose a stochastic
management, transmission congestion relief, T&D upgrade optimisation model to determine the locations and capacities of
deferral, carbon emission reduction etc. [2]. A number of ESS BESSs in the distribution networks (DN). The wind power output,
projects have been launched around the world in recent years [3]. availability of micro turbine units, and load demands variation are
Among the different kinds of ESSs, the battery energy storage considered as stochastic variables and are sampled by the scenario
system (BESS) has been drawn widespread concerns due to its based Monte–Carlo simulations. In this work, the owner of DN is
distinct advantages such as rapid response, easy to deploy etc. [4, the utility that invests on BESS units. The objective of the
5]. While much progress has been made in promoting BESS proposed model is to minimise the BESS investment costs, as well
utilisation, successful implementation relies heavily on planning. as daily operation costs under distribution system constraints.
The optimal siting, sizing, and operation rules of BESS can help to Moreover, the chance-constrained programming technique is
adjust the power flow, increase the renewable penetration, and employed in the constraint to ensure the wind power utilisation
reduce the power loss. level. A Monte–Carlo simulation involved differential evolution
(DE) algorithm is developed to solve the proposed model. The
1.2 Related work daily operation costs are calculated based on multi-period AC OPF
with hourly basis under different scenarios. Finally, a modified
The BESS planning in distribution system has been studied IEEE 15-bus system is used to validate the effectiveness and
extensively in recent years. In [6], the authors proposed a two-stage accuracy of the proposed methodology. Conclusions are drawn in
stochastic programming method to optimally size the BESS while the last section.
considering the correlations between wind power penetration level
and daily load profile. In [7], a BESS capacity optimisation model
in a micro-grid was proposed by considering system reliability 2 Uncertainty modelling
index, BESS investment cost, and system operation cost. In [8], a The DN considered in this paper consists of wind generators,
novel genetic algorithm was proposed to optimally size the BESS BESSs, micro turbines (MT) units, and load demand. The system is
in a micro-grid with photovoltaic under different weather connected to the main grid through a connection line with limited
conditions. The above works are based on the DC model and capacity. The conceptual structure of the distribution network is
ignore some important factors such as active/reactive power flow depicted in Fig. 1. Various scenarios are generated using Monte–
change, voltage stability etc. The AC models are also adopted in Carlo simulation method. Wind speed, load forecast errors, micro
the literatures [9–14]. In [9], the optimal BESS placement problem turbine operation states are incorporated in each scenario. Each
was studied to maximise wind power utilisation level and minimise scenario and their probabilities are represented by a discrete
the hourly social cost based on probabilistic optimal power flow distribution function.
(OPF) method. In [10], a two-stage framework for ESS capacity
determination was proposed for system with high wind penetration 2.1 Wind power generation model
via scenario analysis. In [11], a BESS planning model was
proposed in the distribution system with high wind penetrations, The variations of wind speed have been statistically proven to
where the BESS capacity is determined according to the follow the Weibull distribution [9]. Its probability density function
historically recorded maximum value of wind power deficit/excess. is given as,
The analytical method and sequential Monte–Carlo simulation
�− �
approach were employed to model system status and find optimal � � �
� � = exp − (1)
ESS allocation in [12, 13]. Although the computation complexity � � �
was greatly reduced under their method, the diurnal variation of
wind speed and daily load variation are ignored. In our previous where v, k, and c are wind speed, shape factor, and scale factor,
work [14], a BESS planning strategy was proposed to mitigate the respectively. The power output of the wind turbine strongly
operation risk of the distribution system, where the capacities and depends on the wind speed as
allocations of the BESSs are optimised through the cost-benefit
analysis. In [15], the ESS is adopted for active distribution network � ≤ �in, � ≥ �out
management with abundant wind resources, different control
schemes are compared. � − �in
�w = � �� ≤ � ≤ �r (2)
Due to the operational constraints, the wind power curtailment �r − �in rated
events occur frequently especially in the area with high wind �rated �r ≤ � ≤ �out
penetration. Considering the environmental and economic factors,
system operators are encouraged to utilise as much renewable
1106 IET Renew. Power Gener., 2016, Vol. 10 Iss. 8, pp. 1105-1113
© The Institution of Engineering and Technology 2016
3 Formulation for the proposed BESS allocation
problem
In this work, the objective is to find optimal allocation scheme of
BESS to minimise total investment and operation cost while satisfy
wind utilisation constraint. Uncertainties such as random wind
power, load variation and MT unit's availability are generated using
Monte–Carlo simulation. In particular, a scenario ξ is defined as a
possible realisation of all the random variables which represents 24
hour interval. At first, a set of planning solution representing where
and ‘how much’ BESS to be installed is generated, subsequently
the operation costs of the DN are calculated under all the different
scenarios. The generalised representation of proposed model is
shown in Fig. 3.
The objective problem includes three items, investment cost,
operation &maintenance cost and daily operation cost of DN. The
objective function can be formulated as,
Fig. 2 Markov chain modelling generator units
min �inv + �opm + �op (4)

Fig. 3 Flowchart of the proposed stochastic planning framework

2.2 Load forecast error modelling The first item indicates BESS investment cost which is decided by
the energy capacity, shown as (5). The second item denotes the
The load forecast errors are assumed to follow the Gaussian daily operation and maintenance cost which is proportional to the
cumulative distribution shown as (3), where, σL and μL are energy capacity of BESS, shown as (6). The capital factor A is
statistical mean and standard deviation of load demand; lc and uc introduced to convert the initial payments into annual terms. For
are the minimum and maximum value of load levels, respectively. consistency, the annual BESS energy capacity cost is also
normalised in daily basis.
��
� − ��
�� � = exp − (3) �
��� ��� �inv = ���bat max (5)
��

The curve fitting, maximum likelihood estimation method is �opm = ∑ �bat max��
, (6)
applied to estimate the distribution parameters for simulating the
system wind power output and load demand. ��
� +�
�= �� + (7)
2.3 Micro turbine units availability +� −

The unit availabilities of the MT units are taken into consideration. where d denotes the real interest rate and Nr is the time span of the
Random outage rates of generation units are often presented using BESS project. The BESS allocation problem is constraint by BESS
the binomial distribution [16]. Outage rate and availability rate of investment budget constraint and the location constraint,
the mth MT unit are represents as qm and 1 − qm, respectively. A
two state continuous Markov model is used in this paper to ���bat max ≤ CAP (8)
simulate the transitions between the operative and outage states
during the daily scheduling period [17]. The scheme diagram is �bat ∈ Φ (9)
shown in Fig. 2, where L states are represented and the transition
intensity from state i to j is denoted as Aij.
The third item Qop shown in (10) is a daily scheduling problem
aiming to minimise the expected daily operation cost under

IET Renew. Power Gener., 2016, Vol. 10 Iss. 8, pp. 1105-1113 1107
© The Institution of Engineering and Technology 2016
different scenario. Once the energy capacity and location of BESS BESS integration will reduce the power generation from MT units
is determined, the charging/discharging power of BESS, MT units hence decrease the emission. The emission cost function can be
output power, power bought from substation and wind power expressed as (17), where CO2, NOx and SO2 are the equivalent
curtailment are optimised at hourly time interval. The system emission cost functions.
operational costs are consists of six terms: the MT generation cost, The daily scheduling problem is subjected to following
system power loss cost, wind curtailment penalty cost, energy constraints:
purchase cost from the substation, BESS life cycle depression cost,
and emission cost of the MT units, the above cost functions can be i. System operation constraints
formulated by (11)–(17):
(a) Load balance equations
�op = Exp ∑ ��gen + ��loss + ��pen + ��grid + ��depre + ��emi (10)
�∈Ω The total electrical energy generation should satisfy load
consumption and power loss at each time period t:
• MT generation costs
� � �
The MT generation cost is expressed with cost coefficients ai and ∑ ����, � + �loss

,� = ∑ �
��� �
, � − �cur, � + ∑ ��� �,
�= �= �=
bi: (18)

+ �ch / dis, �
� �
��gen = ∑ ∑ ������, � + �� (11)
�= �= (b) Power flow equations:

• Power loss costs � � �


∑ �
��� �
, � − �cur, � + ∑ ���� � + ��ch dis � − ∑
, / ,

��� , � = ��
The total amount of network loss costs can be calculated as: �= �= �=
(19)

∑ � � ���cos ��� + ���sin ���
��loss = ∑ �loss��loss � , (12)
�=
� �

• Wind curtailment costs


∑ ���� � − ∑ ,

��� , � = �� ∑ � � ���cos ��� + ���sin ��� (20)
�= �=

For clarity, we assume the wind power has the priority to serve where Gij and Bij are the real and imaginary parts of the bus
system load. When the total system generation exceeds load admittance matrix, and ij is angle difference between the ith and
demand, the excess wind power will be absorbed by the BESS. the jth buses. For calculation simplicity, in this work, the reactive
Once BESS reached its maximum capacity, the excess wind power power injection/consumption of BESS units and wind turbine units
has to be curtailed. are ignored.
� (c) Line current constraint
��pen = ∑ ��cur ��cur , (13)
�= The current magnitude of each branch must lie within their
allowable range to maintain system stability:
• Energy purchase cost from the substation
�min ≤ ����, � ≤ �max (21)

��grid = ∑ �
�buy , ��buy (14)
(d) Bus voltage constraint
�=
• BESS life cycle cost
The voltage magnitude of each node must lie within their allowable
The life cycle is defined as the number of total charge/discharge range:
cycles of BESS. The life cycle depression cost is in this work is
calculated based on the total energy usage, which is the total �min ≤ ����, � ≤ �max (22)
charging/discharging power of BESS. According to the impacts of
discharge rate on battery life in [18], the total energy charging/ (e) Reverse power flow constraint
discharging capability of battery is remained stable within the
reason depth of discharge, e.g. 70%. The cost coefficient factor δ of Reverse power flow would increase system loss while leading
the battery life cycle depression is calculated as overheat of feeders, hence there is strict restrict on the reverse
power flow export limitation. This constraint guarantee that no
CAP reverse power flow through the substation transformer:
�= (15)
�bat max ⋅ Lifecycle
≤ ��� ≤ ��max (23)

��depre = ∑ �ch �, �dis �
, , ⋅� (16)
ii. Micro turbine constraints
�=

• Pollution emission cost The active/reactive power output of ith MT units is constrained by
the generation capacity, expressed as:
� �
�emi = ∑ ∑ �
CO ��� �
, � + NO� ���, � ���, min ≤ ����, � ≤ ���, max (24)
�= �=
(17)

+ SO ����, � ���, min ≤ ���, � ≤ ���, max (25)

iii. Wind turbine constraints


Power generation from burning diesel may result in significant
environmental effect through emissions of CO2, NOx and SO2.
1108 IET Renew. Power Gener., 2016, Vol. 10 Iss. 8, pp. 1105-1113
© The Institution of Engineering and Technology 2016
(a) Wind power output constraint (a) Set the Counter NF = 0.
(b) Generate S scenarios based on the known probability
The wind power generation of ith wind turbine units should not distribution function, each scenario is defined as a possible
exceeding the rated wind power output capacity: realisation of all the random variables which represents 24 hour
interval.
≤ ����, � ≤ �rated, � (26) (c) Check the feasibility of chance constrained in (32) with the S
scenarios.
(b) Wind Power curtailment constraint (d) Set NF = NF+1 if chance constraint hold under this scenario.
Otherwise, go to next step.
≤ ��cur, � ≤ ���

,� (27) (e) Repeat steps (b)–(e) for S times.
(f) Calculate the value of NF/S. If NF/S is larger or equals to the
iv. BESS constraints predefined confidence level, then the chance constraint is satisfied;
otherwise, terminate the Monte–Carlo simulation.
The BESS can store the excess energy generated by the wind
turbines when the load is low, and release energy when the load 4.2 DE algorithm based solving methods
demand is high. For clarity, we assume the BESS is only charged
by the wind power. The main operation constraints of batter storage To date, many algorithms have been applied to solve the BESS
are as following: planning problem, such as non-linear programming NIP [6, 7, 11],
Bender's decomposition [10], genetic algorithms (GA) [8, 9, 12,
(a) BESS charging/discharging power constraint: 13], and particle swarm optimisation (PSO) [14]. In this work,
power flow equations are involved in the objective functions with
The BESS charging/discharging power must not violate the BESS respective constraints. The OPF should be executed completely to
power rating constraints: calculate charge/discharge power in each BESS, MT units output
power and wind power etc., in each time period under various
≤ ��ch, �, ��dis, � ≤ �bat max (28) scenarios. Considering the complexity of the optimisation problem,
we choose the heuristic based algorithm to solve the model.
(b) BESS SOC constraint In this study, we use the Monte–Carlo embedded DE algorithm
to solve the BESS capacity and location determination problem.
The state of charge of BESS is calculated as: We choose DE due to the following considerations: (i) it is simple
and easy to implement; (ii) many exhaustive experiments showed
that DE has very strong global search capability and superior
SOC��, � = �bat

, � /�bat, max (29) performance to other evolutionary algorithms; (iii) DE has been
proven to be efficient in solving many engineering problems [24].
(c) BESS energy transition constraint The detailed principles of DE can be found in [25]. It should be
pointed out that: (i) the chance constraints is modelled as a penalty
� � � �
�bat , � = �bat, � − + �ch, ��ch − �dis, � /�dis Δ� (30) function; (ii) the fitness function is performed by the objective
(d) BESS energy capacity constraint function also with the penalty constraints. The whole optimisation
procedures are as follows:

�bat, min ≤ �bat , � ≤ �bat, max (31)
i. Input the original data: Input the parameters of the DE
algorithm including: scaling factor F, population size NP, cross
Cbat,min and Cbat,max are lower and upper bound of energy allowed over rate CR, mutation vector μ and MAXGEN (maximum
to be stored in battery storage units at the end of every charge/ generation).

discharge. The charge operation ends when �bat , � increases up to ii. Initialisation of population: randomly generate M individuals,

Cbat,max while discharge ends up when �bat and each one present a potential BESS allocation solution.
, � decreases down to
Each individual is a two-dimensional vector, representing the
Cbat,min.
installed BESS's capacity and location, respectively.
iii. Formation of new populations: through mutation, crossover
v. Wind power utilisation level constraint
and selection procedure, the new generation are formulated.
Based on the fixed BESS allocation solution, the generation
To maintain the wind power utilisation level in the distribution
scheduling problem is formulated to minimise the daily
network, a chance-constrained constraint is applied to ensure that
operation cost under all the scenarios. The daily scheduling
for the entire planning horizon (one day), the total wind power
problem in (10)–(32) is essentially a multi-period AC-OPF
curtailment must be smaller than or equal to αi per cent of total
problem, which can be solved by MATPOWER 4.0b3 toolbox
wind power generation with at least 1 − βi probability. [26]. Decision variables including the charging/discharging
power of BESS, diesel units output power, power bought from
� � �
substation and wind power curtailment are optimised on hourly
Pr ∑ �cur � − �� ∑ ∑ ��� � ≤
, , ≥ − �� (32)
basis. If a constraint is violated, a penalty function is added to
�= �= �=
the objective function in (4).
iv. Evolve the population until the number of MAXGEN is met.
4 Solution procedure
v. Display the final decision of the BESS allocation plan.
4.1 Chance constraint checking
The details of the application for Monte–Carlo simulation
It would be difficult to convert the chance constraint presented in
embedded DE algorithm for solving the objective problem are
(32) into its determined equivalent with given confidence level.
summarised in Fig. 4.
Since the decision variables are multi-dimensional and the
constraint in (32) does not form an analytical distribution function
[19, 20]. Therefore, the Monte–Carlo simulation is adopted in this 5 Case studies
work to check if the chance constraint is hold. The application of The modified IEEE 11kV, 15-bus radial system is used to validate
Monte–Carlo simulation on the chance constrained stochastic the proposed method. The topology of the system is shown in
model has been extensively studied in [21–23], and its convergence Fig. 5. The parameters of the system are obtained from [14]. The
analysis has been strictly proved. The detailed procedures are as main substation at bus 1 is used to feed a rural area. The
follows:
IET Renew. Power Gener., 2016, Vol. 10 Iss. 8, pp. 1105-1113 1109
© The Institution of Engineering and Technology 2016
from [27, 28]. The emission cost is defined as $0.125/kg for CO2,
$0.25/kg for SO2, and $0.00125/kg for NOx. The wind farm
consists of three identical Vestas V52-850kW wind turbines, where
the cut-in, cut-out, and rated speeds are 4, 17, and 25 m/s,
respectively. The price of wind curtailment cost is set as $100/MW.
The minimum allowable power bought from substation is assumed
to be 0.05 MWh with the price of $500/MW. The operational
scheduling horizon in one scenario and the dispatch interval are set
to be 24 hours and 1 hour, respectively. The wind speed is obtained
through the National Renewable Energy Laboratory (NREL) 2006
eastern wind data set [29].
In this paper, the Monte–Carlo embedded DE parameters
setting as CR = 0.6, NP = 100, μ = 0.7, and MAXGEN as 200,
respectively. The numbers of scenarios S are assumed as 1000. The
load data are taken from [30]. The chance constraint confidence
levels are set as αi = 0.2, βi = 0.2. In this study, two types of battery
technology are considered. Type I is the lead-acid battery and type
II is the zinc-bromine battery. Since the power rating and energy
capacity of BESS are coupled, once a candidate solution of BESS
capacity is determined, the BESS power rating is set to be 20% of
its total energy capacity. Technical and economical parameters of
the batteries are shown in Table 3 [18]. All the experiments were
executed on a dell PC by Matlab R 2014a with 3.40 GHz quad-core
Intel Core i7 processor and 8 GB of RAM.

5.1 Planning results with different BESS technology


First, we test our method on a base case (without BESS) and two
different BESS technologies involved. Case I and Case II
representing different wind power utilisation levels: no less than 80
and 85% of total wind power generation must be utilised in one
day horizon, respectively, while the confidence level must be held
at least 1−βi = 80% chance. The planning results are shown in
Table 4. The larger the value of α, the more capacity of BESS
needs to be installed. The increased capacity of BESS indicates that
Fig. 4 Flow chart of the proposed Monte–Carlo based DE algorithm
larger BESS has to be installed to accommodate the surplus wind
energy. Consequently, the total costs are increased. We also found
lead/acid batteries would be a better choice due to the relatively
low investment cost. The MT generator outputs and the power
bought from substation in one typical day with/without L/A BESS
installation are shown in Fig. 6. The results clearly show that the
#1 MT unit's output power largely reduced (hour 8–9) and the
power bought from external substation reduced to zero (hour 10–
12) with the installation of BESS, which therefore reduces the
system operation cost. To illustrate the efficiency of the L/A BESS,
the SOC profiles of each BESS units under a specific day are
shown in Fig. 7.
We also compare the hourly wind power output differences
before and after the L/A BESS installation in one typical day in
Fig. 8. We easily found that the wind power curtailment is greatly
reduced especially in the heavy wind periods (hour 1–5, hour 22–
24). The average wind utilisation level increases from 65.7%
(without BESS) to 83.96% (L/A installed). This indicates the BESS
plays an important role in enhancing the wind power utilisation
Fig. 5 Modified IEEE 1η-bus distribution radial system level.
Fig. 9 clearly shows the comparison of daily operation cost
distribution system consists of two micro-turbine units located at before and after different BESS technology (L/A and Zn/Br)
nodes 4 and 6, respectively; and one wind farm connected at bus installation. The sum of total operation cost per day reduced from
11. Parameters of the micro-turbine units and emission cost factors $10018 (without BESS) to $9027 (L/A installed), thus result in
are given in Tables 1 and 2. The emission cost factors are obtained 10.98% saving. In detail, every items of system operation cost

Table 1 Micro-turbine generator data


Gen # ai, $ bi, $/MW PGi,max, MW PGi,min, MW QGi,max, var QGi,min, var Outage rate, %
1 30 130 0.6 0.2 1 0 5
2 50 350 0.9 0.2 0.8 0 8

Table 2 Emission factors for micro-turbine generation units decreased to some extent.
Fuel consumption coefficients for MT units, g/kWh
SO2 NOx CO2
0.000928 0.6188 184.0829

1110 IET Renew. Power Gener., 2016, Vol. 10 Iss. 8, pp. 1105-1113
© The Institution of Engineering and Technology 2016
Fig. 6 Comparison of generator output power with/without BESS (Case I)

level βi decreases, BESS capacity goes up due to the fact that the
larger BESS has to be installed to accommodate the excess wind
energy. Consequently, the total costs are increase. Generally
speaking, there is no known to all optimal allocation plan scheme
in an uncertain situation. The investors need to make a trade-off
between wind power utilisation level and investment cost. The
different planning schemes are provided with different confidence
levels for the investors to make the decision (Table 5).

5.3 Convergence analysis


The convergence analysis of DE is performed and the results are
shown in Fig. 10. We employ the same system settings of Case I:
wind power utilisation level is set not less than 80% with 80%
confidence. We set the population size and maximum iteration time
to be 100 and 200, respectively, and compare DE with GA, and

Table 3 Capital and maintenance factors of different bess


technologies
Fig. 7 Storage activity of BESS in one typical day (Case I) BESS Type LA Zn/Br
energy related cost, kWh 325 467
5.2 Confidence level analysis operation and maintenance cost, Kw year 15 20
life cycle 1500 times 1000 times
To compare the effects of confidence level βi on the chance
round trip efficiency 75% 77%
constraint (31), the BESS planning is performed under multiple
cases with different confidence levels while the wind power SOC limits 20–80%
utilisation level is fixed as 80% (20% of wind power can be project period, years 5
curtailed). In this case, we assume that the battery location is fixed discount rate, per cent 5%
at bus 11 and 9. From the results we can find that when confidence

Table 4 Bess allocation optimisation results (confidence level: 80%, scenarios: 1000)
BESS Type LA Zn/Br
elapse time 16 Hour
case I II I II
battery location 11 9 11 4 11 9 11 9
energy capacity, MWh 1.68 0.68 1.78 0.91 1.68 0.68 1.78 0.89
power loss, MW 0.91 1.15 0.89 1.06
battery cost, $/day 1444.9 1645.91 2054.1 2323.8
emission cost, $/day 210.35 208.35 210.35 208.35
thermal power generation cost, $/day 6596.83 6444.6 6583.15 6428.5
power purchase cost, $day 2089 2043 2016 1972
wind power curtailment cost ($/day) 45.897 41.35 45.35 41.37
battery life cycle depression cost, $/day 10.9 10.9 10.7 10.7
total cost, $/day 10489 10509 11008 11091
wind power utilisation level 83.96% 87.96% 82.65% 86.75%

IET Renew. Power Gener., 2016, Vol. 10 Iss. 8, pp. 1105-1113 1111
© The Institution of Engineering and Technology 2016
Table 5 Computational results with different confidence
level (wind utilisation level: 80%, BESS type: L/A)
Confidence Investment Expected Energy
level βi, % costs per day, operation capacity, MWh
$ costs per day,
$
0 1832.5 11,026 1.78 0.78
10 1644.7 10,626 0.99 1.39
20 1444.9 10,485 0.73 0.91
30 1104.5 9945 0.38 0.78

to enhance the wind power utilisation. Uncertainties such as wind


power fluctuation, load demand variation, MT units availability are
taken into consideration. The system operation costs are calculated
based on the scenario analysis. The chance constraint is introduced
to ensure the utilisation level of wind power and the Monte–Carlo
Fig. 8 Hourly wind power output in one typical day (Case I) imbedded DE approach is applied to solve the proposed problem.
The daily scheduling problem is calculated under all scenarios
using multi-period AC-OPF method. A series of simulation studies
are performed on the modified IEEE 15-bus benchmark system.
The simulation results demonstrate the performance of the
proposed model and the method.

7 Acknowledgments
This work was supported in part by the China Scholarship Council,
in part by the National Natural Science Foundation of China under
grant nos. 71331001, 71401017, and 71420107027, in part by
China Southern Power Grid Funding under grant WYKJ00000027,
and in part by State Key Laboratory of Alternate Electrical Power
System with Renewable Energy Sources (grant no. LAPS14002).

8 References
[1] Opathella, C., Venkatesh, B.: ‘Managing uncertainty of wind energy with
wind generators cooperative’, IEEE Trans. Power Syst., 2013, 28, (3), pp.
2918–2928
[2] Gyuk, I.P., Eckroad, S.: ‘Energy storage for grid connected wind generation
Fig. 9 Impacts of BESS integration on operation cost (Case I) applications’ (US Department of Energy, Washington, DC, 2004), EPRI-DOE
Handbook Supplement, vol. 1008703
[3] Clayton, M.: ‘How enormous batteries could safeguard the power grid’,
Christ. Sci. Monit. Mar, 2009, 22
[4] Luo, F., Meng, K., Dong, Z.Y., et al.: ‘Coordinated operational planning for
wind farm with battery energy storage system’, IEEE Trans. Sustain. Energy,
2015, 6, (1), pp. 253–262
[5] Chakraborty, S., Senjyu, T., Toyama, H., et al.: ‘Determination methodology
for optimising the energy storage size for power system’, IET Gener. Transm.
Distrib., 2009, 3, (11), pp. 987–999
[6] Abbey, C., Joos, G.: ‘A stochastic optimization approach to rating of energy
storage systems in wind-diesel isolated grids’, IEEE Trans. Power Syst., 2009,
24, (1), pp. 418–426
[7] Bahramirad, S., Reder, W., Khodaei, A.: ‘Reliability-constrained optimal
sizing of energy storage system in a microgrid’, IEEE Trans. Smart Grid,
2012, 3, (4), pp. 2056–2062
[8] Changsong, C., Shanxu, D., Tao, C., et al.: ‘Optimal allocation and economic
analysis of energy storage system in microgrids’, IEEE Trans. Power
Electron., 2011, 26, (10), pp. 2762–2773
[9] Ghofrani, M., Arabali, A., Etezadi-Amoli, M., et al.: ‘A framework for
optimal placement of energy storage units within a power system with high
wind penetration’, IEEE Trans. Sustain. Energy, 2013, 4, (2), pp. 434–442
[10] Xiong, P., Singh, C.: ‘Optimal planning of storage capacity in power systems
integrated with pind power gGeneration’, IEEE Trans. Sustain. Energy, 2016,
7, (1), pp. 232–240
[11] Atwa, Y.M., El-Saadany, E.: ‘Optimal allocation of ESS in distribution
systems with a high penetration of wind energy’, IEEE Trans. Sustain.
Fig. 10 Optimisation process of different algorithm Energy, 2010, 25, (4), pp. 1815–1822
[12] Awad, A.S.A., El-Fouly, T.H.M., Salama, M.M.A.: ‘Optimal ESS allocation
for load management application’, IEEE Trans. Power Syst., 2014, 30, (1), pp.
PSO. All above three algorithms are repeated by 10 trails and the 327–336
averaged fitness values are plotted. The maximum deviation [13] Awad, A.S., El-Fouly, T.H., Salama, M.M.: ‘Optimal ESS allocation and load
between the optimal solutions was founded to be less than 5%. The shedding for improving distribution system reliability’, IEEE Trans. Smart
Grid, 2014, 5, (5), pp. 2339–2449
results show that comparing with other algorithms, DE shows [14] Zheng, Y., Dong, Z.Y., Luo, F.J., et al.: ‘Optimal allocation of energy storage
superior performance in terms of rapid convergence rate and system for risk mitigation of DISCOs with high renewable penetrations’,
robustness. IEEE Trans. Power Syst., 2014, 29, (1), pp. 212–220
[15] Gill, S., Premier, G.C., Guwy, A.J., et al.: ‘Energy storage for active network
management on electricity distribution network with wind power’, IET
6 Conclusion Renew. Power Gener., 2014, 8, (3), pp. 249–259
[16] Leite, A.P., Borges, C.L., Falcao, D.M.: ‘Probabilistic wind farms generation
In this paper, a stochastic optimisation method was proposed to model for reliability studies applied to Brazilian sites’, IEEE Trans. Power
find the optimal siting and sizing of BESS in the DN, with the aim Syst., 2006, 21, (4), pp. 1493–1501

1112 IET Renew. Power Gener., 2016, Vol. 10 Iss. 8, pp. 1105-1113
© The Institution of Engineering and Technology 2016
[17] Billinton, R., Allan, R.N., Allan, R.N.: ‘Reliability evaluation of power [25] Wong, K.P., Dong, Z.Y.: ‘Differential evolution, an alternative approach to
systems’ (Plenum press, New York, USA, 1984) evolutionary algorithm, invited book chapter’, in Lee, K., EI-Sharkawi, M.
[18] ‘Battery and electric vehicle report’ July 2010. Available at http://bit.ly/ (Eds): ‘Modern heuristic optimization: theory and applications to power
fGaZPB systems’ (Wiley, New York, 2008)
[19] Charnes, A., Cooper, W.W.: ‘Chance constrained programming’, Manage. [26] MATPOWER, ed, 2006. Available at http://www.pserc.cornell.edu/matpower/
Sci., 1959, 6, (1), pp. 73–79 [27] Strachan, N.D., Farrell, A.E.: ‘Emissions from distributed generation’. CEIC
[20] Ahmed, S., Shapiro, A., Shapiro, E.: ‘The sample average approximation working paper, Carnegie Mellon, 2002
method for stochastic programs with integer recourse’. Technical Report, [28] Greene, N., Hammerschlag, R.: ‘Small and clean is beautiful: exploring the
ISyE Georgia Institute of Technology, 2002 emissions of distributed generation and pollution prevention policies’, Electr.
[21] Liu, Z.P., Wen, F.S., Ledwich, G.: ‘Optimal siting and sizing of distributed J., 2000, 13, (5), pp. 50–60
generators considering uncertainties’, IEEE Trans. Power Deliv., 2011, 26, [29] NREL. Wind integration dataset. Available at http://www.nrel.gov/electricity/
(4), pp. 2541–2551 transmission/eastern_wind_methodology.html
[22] Yang, N., Yu, C.W., Wen, F., et al.: ‘An investigation of reactive power [30] Grigg, C., Wong, P., Albrecht, P., et al.: ‘The IEEE reliability test
planning based on chance constrained programming’, Int. J. Electr. Power system-1996. A report prepared by the reliability test system task force of the
Energy Syst., 2007, 32, (6), pp. 650–656 application of probability methods subcommittee’, IEEE Trans. Power Syst.,
[23] Yang, N., Wen, F.: ‘A chance constrained programming approach to 1999, 14, (3), pp. 1010–1020
transmission system expansion planning’, Elect. Power Syst. Res., 2005, 75,
(2–3), pp. 171–177
[24] Sum-Im, T., Taylor, G.A., Irving, M.R., et al.: ‘Differential evolution
algorithm for static and multistage transmission expansion planning’, IET
Gener. Transm. Distrib., 2009, 3, (4), pp. 365–384

IET Renew. Power Gener., 2016, Vol. 10 Iss. 8, pp. 1105-1113 1113
© The Institution of Engineering and Technology 2016

You might also like