Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

2482 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 37, NO.

4, AUGUST 2022

Co-Optimizing the Siting and Sizing of Batteries


and the Siting of Isolation Devices in
Distribution Systems
Kexing Lai , Member, IEEE, Xuan Wu , Senior Member, IEEE, and Antonio J. Conejo , Fellow, IEEE

Abstract—This paper presents an optimization model to help bi Binary variable indicating whether a BESS is in-
distribution system planners to site and size battery energy stor- stalled at bus i (1) or not (0)
age systems (BESSs) and to site isolation devices. The purpose is ĒiB Energy capacity of the BESS installed at bus i (kWh)
twofold: (i) to enhance system reliability and (ii) to generate revenue
P̄iB Power capacity of the BESS installed at bus i (kW)
by taking advantage of energy arbitrage. Our work fills a research B
gap regarding the optimal coordination of the siting of BESSs Ei,t,o Energy available at the BESS installed at bus i at the
and isolation devices. Input data for the model includes utility beginning of time t during normal operating scenario
historical load and fault data. This data is used to generate multiple o (kWh)
B∗
scenarios pertaining to normal and faulted operating conditions. Ei,t,s Energy available at the BESS installed at bus i at the
The proposed optimization model is recast as a mixed-integer linear beginning of time t during fault scenario s (kWh)
programming (MILP) problem and solved using a commercial B,D
Pi,t,o Discharging power from the BESS installed at bus i
solver. It is successfully applied in a real-world case study. at time t during normal operating scenario o (kW)
B,D∗
Index Terms—Distribution reliability, distribution operations, Pi,t,s Discharging power from the BESS installed at bus i
isolation device, battery energy storage, optimization. at time t during fault scenario s (kW)
B,C
Pi,t,o Charging power to the BESS installed at bus i at time
t during normal operating scenario o (kW)
NOMENCLATURE B,C∗
Pi,t,s Charging power to the BESS installed at bus i at time
t during fault scenario s (kW)
Indices and sets UE
Pi,t,s Unserved energy at bus i at time t during fault sce-
S Set of fault scenarios, indexed by s nario s (kW)
O Set of operating scenarios, indexed by o F
Pl,t,o Power flow through distribution line l at time t during
I Set of buses, indexed by i normal operating scenario o (kW)
L Set of distribution lines, indexed by l F∗
Pl,t,s Power flow through distribution line l at time t during
T Set of time periods, indexed by t fault scenario s (kW)
I Sub Set of buses connected to the substation xLl,s Status of distribution line l during fault scenario s; 1
LUi Set of distribution lines upstream of bus i if disabled and 0 otherwise
LDi Set of distribution lines downstream of bus i xIi,s Status of bus i during fault scenario s; 1 if disabled
Il Set of buses connected to distribution line l and 0 otherwise
Os Set of operating scenarios for fault scenario s rlU /rlD Binary variable indicating whether a breaker is in-
stalled at the upstream or downstream side of line l
Variables (1) or not (0)
Sub
Pi,t,o Power exchange with the substation at time t at bus xU D
l,s /xl,s Binary variable indicating whether the upstream or
i during normal operating scenario o (kW) downstream side of the distribution line l during fault
Sub∗
Pi,t,s Power exchange with the substation at time t at bus scenario s is compromised (1) or not (0)
i during fault scenario s (kW)
Constants
ω Weighting factor regarding reliability en-
Manuscript received April 11, 2021; revised June 8, 2021 and August 6, 2021;
accepted September 3, 2021. Date of publication September 10, 2021; date of
hancement
B,O&M
current version July 25, 2022. Paper no. TPWRD-00575-2021. (Corresponding Ct,o Charging/discharging cost of any BESS
author: Xuan Wu.) ($/kW)
PV
Kexing Lai is with Hitachi ABB Power Grids, San Jose, CA 95134 USA Pi,t,o Local generation during time t at bus i
(e-mail: klaikstate@gmail.com).
Xuan Wu is with American Electric Power, New Albany, OH 43054 USA
during normal operating scenario o (kW)
(e-mail: wu.3018@osu.edu). SoC max /SoC min Maximum/minimum energy level of any
Antonio J. Conejo is with the Integrated Systems Engineering and Electrical BESS
LMP
and Computer Engineering Department, Ohio State University, Columbus, OH Ct,o Electric energy price at time t during nor-
43210 USA (e-mail: conejo.1@osu.edu).
Color versions of one or more figures in this article are available at
mal operating scenario o ($/kWh)
https://doi.org/10.1109/TPWRD.2021.3111643. P robo Probability of normal operating scenario
Digital Object Identifier 10.1109/TPWRD.2021.3111643 o

0885-8977 © 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY WARANGAL. Downloaded on July 27,2022 at 11:26:36 UTC from IEEE Xplore. Restrictions apply.
LAI et al.: CO-OPTIMIZING THE SITING AND SIZING OF BATTERIES AND THE SITING OF ISOLATION DEVICES 2483

P rob∗s Probability of fault scenario s BESSs planning in distribution networks has been widely
TsS /TsE Starting/ending time of fault scenario s discussed in the prior literatures with various objectives. For
P̄lF Capacity of distribution line l (kW) instance, reference [1] presents an optimization model to de-
D
Pi,t,o Demand at time t at bus i during normal termine the optimal BESS sizing and siting in a distribution
operating scenario o (kW) network for cost reduction and voltage regulation. Reference
C UE Cost of the unserved energy ($/kWh) [2] proposes a technique for selecting the locations and ratings
fl,s Binary variable indicating the status of of BESSs to maximize revenues of end users in local energy
distribution line l at fault scenario s; 1 if trading. Mitigating operational risk of a distribution grid due to
disabled and 0 otherwise electricity price volatility becomes the goal for BESS sizing and
Rmax Maximum number of isolation devices to siting in [3]. References [4] and [5] discuss the effects of BESS
be installed integration on reliability improvement of distribution networks.
C ID Investment cost of any single isolation References [6] and [7] develop optimization models for BESS
device ($) planning with the objectives of cost minimization and reliability
C B,E Investment cost of any BESS per energy enhancement.
unit ($/kWh) On the other hand, isolation device placement and investment
C B,P Investment cost of any BESS per power optimization problems have been performed in several research
unit ($/kW) works. For instance, reliability enhancement and investment
σB Energy to power ratio of a BESS cost saving are pursued simultaneously by developing a multi-
Eimax Maximum energy capacity of the BESS objective optimization model in [8]. In [9], load uncertainties
installed at an individual bus (kWh) are represented to determine isolation device placements in dis-
B max Maximum number of buses available for tribution networks. The effects of distributed energy resources
installing BESSs (DERs) on isolation device placements are discussed in [10].
r Interest rate A common drawback of references [8] to [10] is the use of
LT Lifetime of any BESS considered (yrs) heuristic solution algorithms, which do not allow quantifying
η Efficiency of BESSs the quality of the solutions attained. Therefore, a few research
P̄ Sub Maximum power exchange with the sub- works have been conducted to develop MILP [11], [12] or
station (kW) mixed integer nonlinear programming (MINLP) [13], [14] opti-
mization models to determine isolation device placements for
reliability enhancement. In addition, combining demand re-
sponse programs and BESS operation is beneficial for reliability
I. INTRODUCTION improvement [15].
INCE a frequent occurrence of power outages is undesir- The aforementioned references do not consider the combined
S able, reliability is a critical concern of distribution system
operators. Reliability issues can be tackled by jointly installing
values and synergic effects of BESSs and isolation devices.
Moreover, models to analyze the functioning logic of isolation
battery energy storage systems (BESS) and isolation devices. devices along with BESS operation (in normal and fault con-
A BESS provides an alternative supply source other than the ditions) are not reported in the literature. On the other hand,
substation to a radial distribution system, which has an im- several research works have demonstrated that the coordination
pact on reliability. Isolation devices, typically composed of a between BESSs and isolation devices is critical for service
pole-mounted recloser/breaker and protective relays, are able to restoration during fault conditions [16], [17]. Therefore, in this
isolate a fault while minimizing the outage impacts to customers. paper, we propose a novel methodology to co-optimize the
BESSs and isolation devices complement each other to enhance siting and sizing of BESSs and the siting of isolation devices,
reliability. A BESS alone cannot create an island if there is no based on a number of representative operating conditions and
isolation device between it and the fault. On the other hand, an fault scenarios. This optimization model targets minimizing
isolation device alone enables supplying the customers upstream the annualized investment cost, annualized operation cost and
of a fault by isolating it. However, the customers downstream this annualized energy not supplied (ENS).
fault are in blackout since there is no backup source. Although The contributions can be summarized as:
the reconfiguration of a distribution network may improve its 1) To propose a methodology to co-optimize the siting and
reliability, reconfiguration is rarely available in suburban/rural sizing of BESSs and the siting of isolation devices to
areas (this is not the case, though, in urban ones). Additionally, minimize total cost.
weather events typically spread across a large footprint, which 2) To validate the proposed methodology in a real-world case
render reconfiguration generally ineffective in suburban/rural study, which proves its benefits.
networks. The rest of the paper is organized as follows. Section II
In addition of enhancing reliability, BESSs can be used for describes the proposed model for BESS siting and sizing and
energy arbitrage during normal operating conditions. A BESS for isolation device siting. Section III reports numerical results
can charge and discharge energy when electricity prices are pertaining to an illustrative example and a real-world case study.
low and high, respectively. This may reduce electricity bills Section IV provides conclusions and recommendations.
for customers and relieve feeder/transformer congestions for
utilities.
However, investment budgets and physical conditions limit II. PROBLEM FORMULATION
the deployments of BESSs and isolation devices. In this paper,
A. Assumptions
we propose an optimization model to co-optimize the siting and
sizing of BESSs and the siting of isolation devices from both the Our paper lays novel work on the co-optimization of
reliability and operational (energy arbitrage) perspectives. (i) BESS siting and sizing, and (ii) isolation device siting,

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY WARANGAL. Downloaded on July 27,2022 at 11:26:36 UTC from IEEE Xplore. Restrictions apply.
2484 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 37, NO. 4, AUGUST 2022

without representing voltage constraints. This way, the problem


can be formulated as a mixed-integer linear programming one,
whose global optimal solution can be obtained by commercial
solvers. Otherwise, representing voltage magnitudes and AC
power flow constraints, the problem would become a mixed-
integer non-linear programming one, which is generally com-
putationally intractable.
Based on our best knowledge of the industry, (i) distribution
Fig. 1. Equivalent circuit for reliability analysis.
asset siting/sizing optimization problem and (ii) distribution
power flow analysis are considered in two consecutive stages
to ensure an adequate performance of the required optimization
tools. After siting and sizing BESSs and siting isolation de-
vices, the distribution planner performs distribution power flow
analysis on normal and fault scenarios to determine the distri-
bution capacitor and regulator locations for potential voltage
constraint violations. In other words, voltage regulation issues
are addressed in a subsequent stage.
Distribution systems, unlike transmission systems, are typ-
ically unbalanced because the majority of customers are con-
nected between a phase and neutral. Based on well-stablished
industry practice, we assume that the distribution circuit is Fig. 2. Decision making process of the logic.
supplied from a substation power transformer with its secondary
side connected in solidly grounded wye. We also assume that
there are few or no three-phase or phase-to-phase loads in B. Reliability Considerations Regarding the Proposed
a circuit dominated by residential or commercial customers. Formulation
Additionally, we note that the coupling effects between phase An illustrative circuit is shown in Fig. 1. Two power sources
conductors are rather insignificant due to relatively low distribu- are available, namely the substation and a BESS tapped on bus
tion voltage levels. As a result, the three phases of a distribution IV. Note that each source is equipped with an isolation device.
circuit have very low impacts among each other and are treated Six candidate isolation device locations on the main feeder are
independently for the sake of simplicity and clarity. Moreover, considered and denoted I1 though IV3. For example, potential
three-phase and phase-to-phase loads can be split into three and isolation device I1 is located on the main-feeder section 1, close
two single-phase loads, respectively. to bus I.
We assume that a fuse always exists on a branch, which is in- If only two isolation devices can be installed and a fault at
tended to isolate any faults on the branch preventing the tripping section 2 of the main feeder is the only considered scenario,
of any main-feeder isolation devices. Therefore, fault conditions then II2 and III2 are the best locations for the two isolation
within a branch have no impact on decisions pertaining to the devices. If they both operate to isolate the single considered
placement of main-feeder isolation devices. Additionally, we fault, customers A and B are supplied by the substation, while
assume that a relay with a directional protection scheme is customers C and D are supplied by the BESS.
used along with the isolation device, which is common industry In order to formulate the relationship between a fault and the
practice [18]. Such protection scheme only detects the fault impact of the corresponding isolation device, the logic detailed
current flowing from the bus to the line side of the isolation in [19] is used here. The decision steps are shown in Fig. 2.
device. From Fig. 2, we note that a fault on section 2 of the main
Note that we use fault scenario probabilities that are based feeder results in that the line between buses II and III is unable to
on utility historical fault data, which are different from the transfer any power. Otherwise, this fault would cause a cascading
failure rates used for System Average Interruption Frequency failure. Assuming that an isolation device is installed on the
Index (SAIFI) or System Average Interruption Duration Index upstream side of line 2 (II2), the fault can be isolated from the
(SAIDI) calculations. For example, if historical data shows that substation and line 1 by tripping such isolation device. Further,
the scenario “line 1 is in fault condition from 12:00-3:00 pm if no isolation device is installed at the downstream side of line 2
during a winter weekday” occurs once in 100 winter weekdays, (III2), then this side is compromised under the considered fault
the associated probability of such fault scenario is calculated condition. Consequently, bus III should be taken out of service.
as 1% × 17.9%. Note that 17.9% is the probability of a winter Specifically, if any line connected to a bus is compromised, this
weekday in a year, which is computed as 100 × 5 (weekdays of bus needs to be disabled. Additionally, if a bus is disabled, all
a week) / 7 (days of a week) / 4 (seasons of a year). In contrast lines and power sources that are connected to it are disabled as
to considering failure rates, the durations of the fault scenarios well. In the example considered, line 3 is affected and should be
are incorporated in our analysis. This is because we consider disabled. Note as well that an isolation device installed at III3
operational cost reductions during both normal operation and would not trip because the fault current it detects flows from
fault conditions, along with reliability enhancement during fault line 3 to bus III, which opposes the relay’s setting direction.
condition. Therefore, we obtain and assign the probabilities of This way, the logic covers line 2 and bus III and loops back
diverse representative scenarios to be used within a stochastic to line 3. If an isolation device is installed at the downstream
programming framework. In a nutshell, the probability we use side of line 3 (IV-3), then it trips and bus IV and its downstream
is related to the failure rate, but incorporates the fault duration circuits are protected. Note that the above logic is represented by
as well, so that it makes sense from a stochastic programming constraints (25)-(28) in the proposed optimization model below
viewpoint. and illustrated in Fig. 2.

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY WARANGAL. Downloaded on July 27,2022 at 11:26:36 UTC from IEEE Xplore. Restrictions apply.
LAI et al.: CO-OPTIMIZING THE SITING AND SIZING OF BATTERIES AND THE SITING OF ISOLATION DEVICES 2485

 
C. Formulation B B B,C B,D
Ei,t+1,o = Ei,t,o + Pi,t,o − Pi,t,o ,
(1) shown at the bottom of this page. subject to:
∀i ∈ I, t ∈ [t1, t23] , o ∈ O (12)
ĒiB = P̄iB σ B , ∀i ∈ I (2)  
B B B,C B,D
Ei,t1,o = Ei,t24,o + Pi,t24,o − Pi,t24,o , ∀i ∈ I, o ∈ O
ĒiB ≤ Eimax bi , ∀i ∈ I (3)
 (13)
bi ≤ B max (4) B
Ei,t1,o = ĒiB SoC min , ∀i ∈ I, o ∈ O (14)
i∈I
 −P̄ Sub ≤ Pi,t,o
Sub
≤ P̄ Sub , ∀i ∈ I Sub , t ∈ T, o ∈ O (15)
rlU + rlD ≤ Rmax (5)
l∈L Constraints (6)-(15) pertain to the operation of the system
The objective function (1) to be minimized consists of three under normal operating conditions. Constraints (6) and (7) are
terms, i.e., investment cost pertaining to BESSs and isolation nodal power balance equations for the substation and main-
devices, operation cost during normal operating conditions, and feeder buses, respectively. Constraint (8) enforces power flow
loss of load cost during contingencies. The investment cost limits for distribution lines. Discharging from and charging to
consists of the investment cost in BESSs that is determined by BESSs are limited by their power capacities, as enforced by
energy and power capacities, along with the investment cost constraints (9) and (10), respectively. Constraint (11) bounds
in isolation devices. The weighting factor ω is used to realize the stored energy in each BESS. Constraints (12) and (13) are
the tradeoff between operation cost reduction and reliability energy balances for BESSs in time periods between the first
enhancement. and 23rd hour and between the 24th and the first hour of next
Constraints (2)-(5) are constraints for the planning stage. day, respectively. Constraint (14) sets the initial stored energy
Equation (2) enforces the relationship between power and energy of any BESS to a minimum. Finally, power exchange with the
capacities of BESSs. Constraint (3) limits the installed BESS substation is limited by P̄ Sub , as enforced by (15).
capacity at each bus. The number of available buses for BESS B,C∗
deployment is set by (4), and (5) sets a budget cap for isolation    B,D∗ Pi,t,s
Sub PV
Pi,t,s + Pi,t,o 1 − xIi,s + Pi,t,s η−
devices deployment. η
o∈Os
Sub
Pi,t,o + PV
Pi,t,o   
F∗ F∗ D UE
+ Pl,t,s − Pl,t,s = Pi,t,o − Pi,t,s , (16)
B,C   o∈Os
B,D Pi,t,o F F D
l∈LU
i l∈LD
i
+ Pi,t,o η− + Pl,t,o − Pl,t,o = Pi,t,o , (6)  
η ∀i ∈ I Sub , t ∈ TsS , TsE , s ∈ S
l∈LU
i l∈LD
i

B,C∗
∀i ∈ I Sub , t ∈ T, o ∈ O 
PV
 I
 B,D∗ Pi,t,s
Pi,t,o 1 − xi,s + Pi,t,s η −
B,C
Pi,t,o   η
PV B,D F F D o∈Os
Pi,t,o + Pi,t,o η− + Pl,t,o − Pl,t,o = Pi,t,o ,   
η F∗ F∗ D UE
U l∈Li D l∈Li + Pl,t,s − Pl,t,s = Pi,t,o − Pi,t,s , (17)
(7) l∈LU
i l∈LD
i
o∈Os
/ I Sub , t ∈ T, o ∈ O
∀i ∈ I ∈  
/ I Sub , t ∈ TsS , TsE , s ∈ S
∀i ∈ I ∈
−P̄lF ≤ F
Pl,t,o ≤ P̄lF , ∀l ∈ L, t ∈ T, o ∈ O (8)
   
B,D
0 ≤ Pi,t,o ≤ P̄iB , ∀i ∈ I, t ∈ T, o ∈ O (9) −P̄lF 1 − xLl,s ≤ Pl,t,sF∗
≤ P̄lF 1 − xLl,s , (18)
 
B,C
0 ≤ Pi,t,o ≤ P̄iB , ∀i ∈ I, t ∈ T, o ∈ O (10) ∀l ∈ L, t ∈ TsS , TsE , s ∈ S
B,D∗    
ĒiB SoC min ≤ Ei,t,o
B
≤ ĒiB SoC max , ∀i ∈ I, t ∈ T, o ∈ O 0 ≤ Pi,t,s ≤ P̄iB 1 − xIi,s , ∀i ∈ I, t ∈ TsS , TsE , s ∈ S
(11) (19)

min r(r + 1)
LT    ID  U 
B,E B B,P B D
C Ēi + C P̄i + C rl + rl
Ξ1 (r + 1)LT − 1
i∈I l∈L
⎧    LMP Sub   ⎫

⎨ P robo ω Ct,o Pi,t,o + Ct,o B,O&M B,D
Pi,t,o B,C
+ Pi,t,o ⎪

+ 365 o∈O t∈T  
i∈I

⎩+ P rob∗s (1 − ω) C UE Pi,t,s
UE ⎪

s∈S t∈T i∈I

Sub Sub∗ B,D B,D∗ B,C B,C∗


Ξ1 = {Pi,t,o , Pi,t,s , ĒiB , P̄iB , Ei,t,o
B B∗
, Ei,t,s , Pi,t,o , Pi,t,s , Pi,t,o , Pi,t,s ,
UE F F∗
Pi,t,s , Pl,t,o , Pl,t,s , xLl,s , xIi,s , rlU , rlD , xU D
l,s , xl,s , bi } (1)

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY WARANGAL. Downloaded on July 27,2022 at 11:26:36 UTC from IEEE Xplore. Restrictions apply.
2486 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 37, NO. 4, AUGUST 2022

B,C∗    
0 ≤ Pi,t,s ≤ P̄iB 1 − xIi,s , ∀i ∈ I, t ∈ TsS , TsE , s ∈ S
(20)
ĒiB · SoC min ≤ Ei,t,sB∗
≤ ĒiB SoC max , (21)
 
∀i ∈ I, t ∈ TsS , TsE , s ∈ S
 
B∗ B∗ B,C∗ B,D∗
Ei,t+1,s = Ei,t,s + Pi,t,s − Pi,t,s , (22)
 
∀i ∈ I, t ∈ TsS , TsE − 1 , s ∈ S
  
B∗ B B,C∗ B,D∗
Ei,T S ,s =
s
E i,T S −1,o +
s
P i,T S ,s − Pi,T S ,s , (23) Fig. 3. 4-bus system layout.
s s
o∈Os

∀i ∈ I, s ∈ S 
    downstream side of bus i, is compromised ( l∈LD xU l,s = 1) or
i
−P̄ Sub 1 − xIi,s ≤ Pi,t,s Sub∗
≤ P̄ Sub 1 − xIi,s , (24) the downstream side of line l, located on the upstream side of
  bus i, is compromised ( l∈LU xD l,s = 1), then bus i is disabled
∀i ∈ I Sub , t ∈ TsS , TsE , s ∈ S i
(xi,s = 1). Accordingly, the binary variable xIi,s is computed
I
  L  
xUl,s = 1 − rl
U
xl,s , ∀l ∈ L, s ∈ S (25) as min( l∈LD xU l,s + l∈LU xDl,s , 1). Finally, constraint (28)
i i
  defines the status of line l. As shown in Fig. 2, line l is disabled
xDl,s = 1 − rl
D
xLl,s , ∀l ∈ L, s ∈ S (26)
⎛ ⎞ (xLl,s = 1) if affected by the fault (fl,s = 1) or either of its end

  buses is isolated ( i∈Il xIi,s ≥ 1). We note that the min functions
xIi,s = min ⎝ xU l,s + xD ⎠
l,s , 1 , ∀i ∈ I, s ∈ S (27) in constrains (27) and (28) and the products between binary
l∈LD
i l∈LU
i variables can be easily linearized [20]. As a result, the proposed
  model is recast into a MILP problem, whose global optimality

xLl,s = min fl,s + xIi,s , 1 , ∀l ∈ L, s ∈ S (28) can be ensured by commercial solvers, such as CPLEX [21].
i∈Il
III. NUMERICAL RESULTS
Constraints (16)-(28) pertain to system operation during con-
tingencies. Note that these constraints are only enforced during In this section, we validate the performance of the proposed
contingencies, i.e., within the time range [TsS , TsE ]. The nodal optimization model using two case studies, an illustrative IEEE
power balance equations are (16) and (17) for the substation 4-bus system and a real-world distribution network located in
and main-feeder buses, respectively. In contrast to constraints Oklahoma, USA. Note that both test systems are modified for
UE the purpose of the studies carried out. Both case studies have
(6) and (7), ENS (Pi,t,s ) is incorporated in (16) and (17) and
the power output of the local generation at bus i depends on the been implemented on a laptop with an Intel Core i5 CPU and
status of such bus (xIi,s ). Constraint (18) calculates the power 8GB of RAM. CPLEX [21] is used as the optimization solver.
flow through line l during fault scenario s, which depends on the Energy and power investment costs (C B,E and C B,P ) of
status of distribution line l (xLl,s ). Discharging/charging of each BESSs are set to $150/kWh and $150/kW, respectively. Besides,
BESS during fault conditions is stated by constraints (19) and the investment cost on a single isolation device is considered to
(20), respectively, which depends on the BESS power capacity be $10000. Note that the cost data for BESSs and isolation de-
and the status of the corresponding bus. Energy stored in each vices are selected based on [22], and [11] and [23], respectively,
BESS during a fault condition is limited by (21). Constraint (22) and should be determined based on utility actual data. Moreover,
is the energy balance equation for the BESS at bus i during fault the lifetime of a BESS and the interest rate are set as 15 years
scenario s. By constraint (23), the stored energy at the beginning and 10%, respectively.
of a fault condition period is determined by the stored energy at The outcomes of the model are first analyzed and evaluated.
the hour prior to the fault condition in the corresponding normal Then, we examine the impact of various parameters on the
B
operating condition and is denoted as Ei,T S . Constraint (23) results.
s −1,o
links normal operating and fault conditions. Constraint (24)
limits the power exchange with the substation during a fault A. Illustrative 4-Bus System
condition.
Constraints (25)-(28) enforce the logic illustrated in Fig. 2. The schematic of a modified IEEE 4-bus system, used as an
Constraint (25) defines the status of the upstream side of a line, illustrative example, is shown in Fig. 3. This is an unbalanced
which is determined by the line status and whether or not an 3-phase distribution system with two PV panels installed on
isolation device is installed at the corresponding side. If line phases A and C of bus I3. Additionally, loads are located at
l is disabled (xLl,s = 1) and no isolation device is available the three phases of bus I4 and phase B of buses I1 and I2.
We assume that for each distribution line, the side closer to the
at the upstream side (rlU = 0), the upstream side of line l is substation is defined as the upstream side, being the other side
compromised (xU l,s = 1). A similar logic is applied to determine the downstream one. We also assume that the substation can
the status of the downstream side of line l with xD l,s , as enforced supply all the demand.
by constraint (26). Constraint (27) is used to identify the status The 24-hour correlated profiles of electricity prices, total local
of bus i (xIi,s ). If the upstream side of line l, located on the (PV) generations and load demands are depicted in Figs. 4, 5 and

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY WARANGAL. Downloaded on July 27,2022 at 11:26:36 UTC from IEEE Xplore. Restrictions apply.
LAI et al.: CO-OPTIMIZING THE SITING AND SIZING OF BATTERIES AND THE SITING OF ISOLATION DEVICES 2487

Fig. 4. 24-hour profile of electricity price.

Fig. 5. 24-hour profile of total PV generation.

Fig. 6. 24-hour profile of total load demand.

TABLE I
OPERATION CONDITIONS AND TIME PERIODS OF FAULT SCENARIOS

Fig. 7. Optimal siting of BESSs and isolation devices with various fault
locations.

We note that the proposed model can represent different


fault scenarios involving any fault locations, time periods, or
operating conditions, based on historical fault data.
6, respectively. We consider four operating conditions (days) 1) Impact of Fault Locations: In this section, outcomes re-
representing the four seasons of the year. Additionally, three garding BESS and isolation device placements with various
fault scenarios are considered per operating condition, which faulted lines are discussed. We assume that the weighting fac-
leads to 12 fault scenarios in total. Finally, the ENS cost (C UE ) tor of operation cost reduction versus ENS reduction (ω), the
is set to $30/kWh and the probability of each fault scenario maximum energy capacity of each BESS installed on each
(P rob∗s ) to 1%. individual bus (Eimax ), the budget for isolation devices (Rmax )
Table I summarizes the operating conditions and time periods and the available number of buses where a BESS can be installed
related to the fault scenarios considered in this example. Fault (B max ) are set to 0.5, 500 kWh, 3 and 3, respectively. The
scenarios have different time periods, operating conditions, and optimal siting (not sizing) of BESSs and isolation devices based
fault locations. For the sake of illustration, we assume that the on three sets of fault locations are illustrated in Fig. 7. The
fault locations for the 12 fault scenarios are the same, while the optimization outcomes shown in Fig. 7(a) through Fig. 7(c) are
time periods and operating conditions vary based on Table I. results from three separate cases. We consider one fault scenario

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY WARANGAL. Downloaded on July 27,2022 at 11:26:36 UTC from IEEE Xplore. Restrictions apply.
2488 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 37, NO. 4, AUGUST 2022

Fig. 8. Total number of deployed isolation devices (IDs) and total capacity of
BESSs with different values of ω.
Fig. 9. Operation cost and ENS with different Eimax .

(with three simultaneous faults) per each case to illustrate how


fault locations impact the siting of BESSs and isolation devices.
The results indicate that the budgets of BESSs and isolation
devices are fully used since their investments can be fully
justified by the reduction in operation and ENS costs. Moreover,
the interdependence between BESSs and isolation devices can
be observed, which highlights the merit of a joint planning of
both resources. Specifically, with the deployments of isolation
devices to isolate faults, BESSs are placed to ensure that they
are available to supply isolated loads during faults. For instance,
in Fig. 7(a), with isolation devices installed at phases B and C to Fig. 10. Operation cost and ENS with different B max .
isolate the faults at L3 and L5, BESSs are also placed at phases
B and C to support loads D2 and D3, since the installed BESSs
can be islanded by the isolation devices. In Fig. 7(b), an isolation and Rmax are analyzed. We assume ω = 0.5, being the faulted
device is installed at the upstream side of bus I3 of phase A to lines are L2, L8 and L9.
island load D1, the local generation and an installed BESS at bus When B max = 3 and Rmax = 3 , the operation cost and
I3 from the fault at L4. Another isolation device is installed at L2 ENS for various values of Eimax are plotted in Fig. 9. The results
to protect D4. Since no fault affects phase C, neither isolation indicate that both operation cost and ENS decrease as the value
device nor BESS is installed in such phase. In Fig. 7(c), D1, of Eimax increases. This is because a higher value of Eimax
D3 and D4 are protected by the corresponding isolation devices, allows a greater total capacity of BESSs to be installed, which is
while D2 and D5 are not. This is because D3 and D4 are the two helpful from both energy arbitrage and reliability enhancement
largest loads. In addition, both an isolation device and a BESS perspectives.
need to be installed downstream of the fault at L2 to supply D2 Fig. 10 illustrates the effects of B max on operation cost
and D5. This results in that a part of D1, D3 or D4 cannot be and ENS, assuming Eimax = 150 kWh and Rmax = 3. The
supplied under fault condition. operation cost and ENS decrease for B max ≤ 2 and remain con-
We note that, in contrast to the conventional approach in which stant for B max > 2. This is because with B max ≤ 2 , installing
two isolation devices are expected to be placed at both sides of BESSs at more buses helps providing additional backup power to
a line to fully isolate a line fault, using the proposed model, support loads under faults. However, with B max > 2, installing
isolation devices are deployed strategically to make the best use BESSs at additional buses does not result in further ENS reduc-
of the budget available. tion due to the limited deployments of isolation devices. Note
2) Results of BESS Sizing and Isolation Device Planning: that a BESS is useless if no isolation device is planned between
In this section, the impact of the weighting factor ω (operation it and a fault. In other words, without coordinating between
vs. reliability) on BESS sizing and siting and on isolation device isolation devices and BESSs, it is not guaranteed that additional
siting is discussed. We assume Eimax = 500 kWh , B max = 3 BESSs enhance reliability. On the other hand, reduced operation
and Rmax = 5. The resulting planning strategies for various cost by installing additional BESSs is generally not sufficient to
values of ω are plotted in Fig. 8. From this figure, we note that a justify the corresponding increase in investment cost. There-
higher value of ω leads to fewer isolation devices (IDs) deployed fore, no additional BESSs is deployed when B max > 2. As
and a lower total capacity of BESSs. This is because a lower a result, operation cost and ENS are not further reduced. This
value of ω indicates a higher weight on reliability enhancement, demonstrates the necessity of a coordinated expansion planning
instead of energy arbitrage. As a result, a larger number of of isolation devices and BESSs.
isolation devices are installed for reliability improvement. On Finally, with Eimax = 150 kWh and B max = 3, Fig. 11
the other hand, under the considered system setting, reduced shows the impact of Rmax (maximum number of isolation
operation cost due to energy arbitrage of BESSs cannot justify devices) on operation cost and ENS. The results indicate that, in
investing on them. However, the corresponding investment cost general, ENS decreases with a higher value of Rmax . However, if
can be justified by ENS reduction. Therefore, imposing a higher Rmax = 3 and 4, or Rmax ≥ 5, ENS remains constant . Clearly,
weight on reliability enhancement (i.e., lower value of ω) leads ENS can be reduced with additional isolation devices, since ad-
to an increased size of BESSs. ditional faults can be isolated. This explains the general decline
3) Impacts of Eimax , B max , and Rmax : In this section, the of ENS. However, with Rmax = 3 or 4 , even though additional
impacts of various system parameters, including Eimax , B max faults can be isolated, the budget of BESS deployments limits the

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY WARANGAL. Downloaded on July 27,2022 at 11:26:36 UTC from IEEE Xplore. Restrictions apply.
LAI et al.: CO-OPTIMIZING THE SITING AND SIZING OF BATTERIES AND THE SITING OF ISOLATION DEVICES 2489

Fig. 11. Operation cost and ENS with different Rmax .

system’s capability to reduce ENS. Moreover, when the number


of isolation devices reaches a certain threshold and all faults
can be isolated, deploying additional isolation devices is not
helpful in further decreasing ENS. This explains the unchanged
ENS when Rmax ≥ 5. Furthermore, Fig. 11 also reveals that the
operation cost remains unchanged if Rmax = 2 and 3, but ENS
decreases. This is because increasing the number of isolation
devices from 2 to 3 can isolate additional faults, rendering
reduced ENS. However, the additional isolation device cannot
isolate the bus that is available for BESS deployment from faults.
Under such circumstance, no BESS needs to be installed, as the
investment cost of BESSs hinders the deployments of BESSs,
thus the operation cost cannot be reduced by energy arbitrage. Fig. 12. The real-world distribution system layout.
With other values of Rmax , the operation cost reduces along
with ENS. This is because additional BESSs can be installed
for ENS mitigation, which in turn leads to an operation cost
reduction. If Rmax ≥ 5 , since deploying additional BESSs
does not enhance reliability (as discussed earlier), the operation
cost remains unchanged. This analysis further reiterates the
interdependence between expansion planning decisions regard-
ing BESSs and isolation devices, and highlights the need of
co-optimizing their planning decisions.

B. A Real-World System
To further illustrate and to demonstrate the scalability of
the proposed optimization model, we consider in this section Fig. 13. 24-hour profile of electricity price.
a modified real-world distribution system located in Oklahoma,
USA. It is a three-phase unbalanced distribution network with 58
buses across the three single-phase circuits as shown in Fig. 12.
This figure also shows the peak load of each bus.
The 24-hour correlated profiles of electricity prices, total
local (PV) generations and total load demands are depicted in
Figs. 13, 14 and 15, respectively. We assume that the capacity
of a single PV panel is 10 kW. With 16 PV panels installed in
the system, the PV penetration level is 16.3%. Additionally, we
consider 8 operating conditions that represent the weekdays and
weekends of the four seasons. For each operating condition, 5
fault scenarios with different faulted lines and fault time periods
are considered, which result in a total of 40 fault scenarios. Note Fig. 14. 24-hour profile of total PV generation.
that each fault scenario involves two adjacent fault lines.
The ENS results for different values of Rmax and B max are
plotted in Fig. 16, assuming Eimax = 600 kWh and ω = 0.1. assuming that only one BESS can be installed (B max = 1).
The findings of this case study are consistent with those from From the results, we observe that an isolation device is always
the IEEE 4-bus example, which reveal that ENS decreases with installed at the downstream side of L45. This is due to the heavy
the increasing number of isolation devices and the increasing loading and high PV penetration at the area downstream of L45
number of available buses for BESS deployment. (enclosed by the dashed rectangle box in Fig. 12). Deploying an
The results regarding optimal BESS and isolation device isolation device at that location allows isolating that area if a fault
placements as a function of Rmax are provided in Table II, occurs on the upstream of L45. In such case, PV generation can

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY WARANGAL. Downloaded on July 27,2022 at 11:26:36 UTC from IEEE Xplore. Restrictions apply.
2490 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 37, NO. 4, AUGUST 2022

Fig. 15. 24-hour profile of total load demand.


Fig. 17. Expected ENS with different Rmax using two approaches.

reduced in Case 1, which implies that for deploying isolation


devices in a distribution system with a limited budget, it may
not be optimal to place two isolation devices per line. Instead, a
more flexible placement strategy allows maximizing the usage of
the available resources, thus enhancing reliability, and reducing
cost.

IV. CONCLUSION
A methodology to co-optimally site and size BESSs and site
isolation devices in a radial distribution system is proposed in
Fig. 16. Expected ENS with different Rmax and B max .
this paper. This methodology uses a MILP optimization model
with inputs comprising historical load and fault data. From the
TABLE II analysis carried out and the case studies, the conclusions below
PLACEMENT OF BESSS AND ISOLATION DEVICES are in order:
1. The sensitivity analysis carried out demonstrates the in-
terdependence between planning decisions of BESSs and
isolation devices, which highlights the necessity of a co-
optimization approach.
2. Imposing a higher weight on reliability enhancement pro-
motes BESS investments, since the corresponding invest-
ment cost can be further justified by ENS reduction.
3. The optimal number of BESSs to be deployed is impacted
by the placement of isolation devices and the trade-off
between investment and operation costs.
4. Expected ENS can be reduced by increasing the number
of isolation devices installed. However, when the number
of isolation devices reaches a certain threshold, installing
additional isolation devices reduces neither cost nor ENS.
For the sake of computational tractability, the considered
be utilized for load serving. Similarly, a BESS is always installed problem is typically addressed in two steps: first, the design
in that area to further reduce load shedding. of the active power components, including BESSs and isolation
Finally, to demonstrate the merit of the proposed approach, devices, and second, the required reactive power support. Our
two additional cases are considered. paper focuses on the first step. Nevertheless, an approach con-
Case 1: We use the proposed optimization model to site and size sidering simultaneously the active and reactive power dimen-
BESSs and site isolation devices. sions would certainly be more accurate, although potentially
Case 2: We use the proposed optimization model to site intractable. We will explore such approach in our future work.
and size BESSs and site isolation devices, but setting two
isolation devices per line. In other words, the constraint REFERENCES
rlU = rlD , ∀l ∈ L , is added to the optimization model. This
[1] Y. Zheng, Y. Song, A. Huang, and D. J. Hill, “Hierarchical optimal alloca-
approach reflects the most prevalent criterion for transmission tion of battery energy storage systems for multiple services in distribution
system protection. systems,” IEEE Trans. Sustain. Energy, vol. 11, no. 3, pp. 1911–1921,
Jul. 2020.
Fig. 17 provides comparison results with different values of [2] P. Tenti and T. Caldognetto, “A general approach to select location and
Rmax , assuming B max = 2, Eimax = 600 kWh, and ω = ratings of energy storage systems in local area energy networks,” IEEE
0.1. It can be observed that the expected ENS is comparatively Trans. Ind. Appl., vol. 55, no. 6, pp. 6203–6210, Nov./Dec. 2019.

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY WARANGAL. Downloaded on July 27,2022 at 11:26:36 UTC from IEEE Xplore. Restrictions apply.
LAI et al.: CO-OPTIMIZING THE SITING AND SIZING OF BATTERIES AND THE SITING OF ISOLATION DEVICES 2491

[3] Y. Zheng, Z. Y. Dong, F. J. Luo, K. Meng, J. Qiu, and K. P. Wong, “Optimal [20] S. H. Gabriel, A. J. Conejo, J. D. Fuller, B. F. Hobbs, and C. Ruiz,
allocation of energy storage system for risk mitigation of DISCOs with Complementarity Modeling in Energy Markets. New York, NY, USA:
high renewable penetrations,” IEEE Trans. Power Syst., vol. 29, no. 1, Springer, 2013.
pp. 212–220, Jan. 2014. [21] “IBM CPLEX 20.1.0,” 2020. [Online]. Available: https://www.ibm.com/
[4] P. Gautam, R. Karki, and P. Piya, “Probabilistic modeling of energy storage analytics/cplex-optimizer
to quantify market constrained reliability value to active distribution [22] W. Cole, A. Frazier, and C. Augustine, “Cost projections for utility-scale
systems,” IEEE Trans. Sustain. Energy, vol. 11, no. 2, pp. 1043–1053, battery storage: 2021 update,” Jun. 2018. [Online]. Available: https://www.
Apr. 2020. nrel.gov/docs/fy21osti/79236.pdf
[5] X. Wu, A. J. Conejo, and S. Mathew, “Optimal siting of batteries in dis- [23] S. Abdi, K. Afshar, S. Ahmadi, N. Bigdeli, and M. Abdi, “Optimal
tribution systems to enhance reliability,” IEEE Trans. Power Del., vol. 36, recloser and autosectionalizer allocation in distribution networks using
no. 5, pp. 3118–3127, Oct. 2021, doi: 10.1109/TPWRD.2020.3034095. IPSO–Monte Carlo approach,” Int. J. Elect. Power Energy Syst., vol. 55,
[6] A. V. Pombo, J. Murta-Pina, and V. F. Pires, “Multiobjective formulation pp. 602–611, 2014.
of the integration of storage systems within distribution networks for
improving reliability,” Elect. Power Syst. Res., vol. 148, pp. 87–96, 2017.
[7] K. Masteri, B. Venkatesh, and W. Freitas, “A fuzzy optimization model
for distribution system asset planning with energy storage,” IEEE Trans. Kexing Lai (Member, IEEE) received the Ph.D. de-
Power Syst., vol. 33, no. 5, pp. 5114–5123, Sep. 2018. gree in electrical engineering from The Ohio State
[8] C. Chen, C. Lin, H. Chuang, C. Li, M. Huang, and C. Huang, “Optimal University, Columbus, OH, USA, in 2019. He is
placement of line switches for distribution automation systems using im- currently a Senior Application Engineer with Hitachi
mune algorithm,” IEEE Trans. Power Syst., vol. 21, no. 3, pp. 1209–1217, ABB Power Grids. His research interests include
Jul. 2006. power system operation and planning, power system
[9] R. Yari, M. Shakarami, F. Namdari, and H. CheshmehBeigi, “A novel protection, application of multi-agent and machine
practical method for simultaneous placement of switching and protective learning in power systems, state estimation, and re-
devices considering load uncertainty,” Int. Trans. Elect. Energy Syst., silience analysis of power systems.
vol. 29, no. 6, Jun. 2019, Art. no. e12025.
[10] M. Chehardeh and C. Hatziadoniu, “Optimal placement of remote-
controlled switches in distribution networks in the presence of distributed
generators,” Energies, vol. 12, no. 6, pp. 1–17, Mar. 2019.
[11] M. Izadi, A. Safdarian, M. Moeini-Aghtaie, and M. Lehtonen, “Optimal
placement of protective and controlling devices in electric power distri- Xuan Wu (Senior Member, IEEE) received the M.S.
bution systems: A MIP model,” IEEE Access, vol. 7, pp. 122827–122837, degree from Arizona State University, Tempe, AZ,
Aug. 2019. USA, in 2013, and the Ph.D. degree from The Ohio
[12] A. Heidari and V. Agelidis, “Considerations of sectionalizing switches in State University, Columbus, OH, USA, in 2018, both
distribution networks with distribution generation,” IEEE Trans. Power in electrical engineering. He is currently a Principal
Del., vol. 30, no. 3, pp. 1401–1409, Jun. 2015. Engineer with American Electric Power. His research
[13] A. Alam, V. Pant, and B. Das, “Switch and recloser placement in distri- interests include power system operations, planning,
bution system considering uncertainties in loads, failure rates and repair security and resilience, and power engineering and
rates,” Elect. Power Syst. Res., vol. 140, pp. 619–630, Nov. 2016. equipment. Dr. Wu is a Registered Professional En-
[14] A. Heidari, Z. Y. Dong, D. Zhang, P. Siano, and J. Aghaei, “Mixed-Integer gineer in Ohio of USA, the Editor of the IEEE
nonlinear programming formulation for distribution networks reliability TRANSACTIONS ON POWER DELIVERY, and an Edi-
optimization,” IEEE Trans. Ind. Informat., vol. 14, no. 5, pp. 1952–1961, torial Board Member of the International Transactions on Electrical Energy
May 2018. Systems.
[15] T. Khalili A. Jafari, M. Abapour, and B. Mohammadi-Ivatloo, “Opti-
mal battery technology selection and incentive-based demand response
program utilization for reliability improvement of an insular microgrid,”
Energy, vol. 169, pp. 92–104, 2019.
[16] M. Choobineh, D. Silva-Ortiz, and S. Mohagheghi, “An automation
scheme for emergency operation of a multi-microgrid industrial park,” Antonio J. Conejo (Fellow, IEEE) received the M.S.
IEEE Trans. Ind. Appl., vol. 54, no. 6, pp. 6450–6459. Nov./Dec. 2018. degree from MIT, Cambridge, MA, USA, in 1987,
[17] P. Prabawa and D. Choi, “Multi-agent framework for service restoration in and the Ph.D. degree from the Royal Institute of Tech-
distribution systems with distributed generators and static/mobile energy nology, Stockholm, Sweden, in 1990. He is currently
storage systems,” IEEE Access, vol. 8, pp. 51736–51752, Mar. 2020. a Full Professor with Integrated System Engineering
[18] IEEE Guide for Protective Relay Applications to Distribution Lines, IEEE and the Electrical and Computer Engineering Depart-
Standard C37.230-2020, pp. 1–106, 2021. ments, The Ohio State University, Columbus, OH,
[19] K. Lai, M. Illindala, and K. Subramaniam, “A tri-level optimization USA. His research interests include control, opera-
model to mitigate coordinated attacks on electric power systems in tions, planning, economics and regulation of electric
a cyber-physical environment,” Appl. Energy, vol. 235, pp. 204–218, energy systems, and also statistics and optimization
Feb. 2019. theory and its applications.

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY WARANGAL. Downloaded on July 27,2022 at 11:26:36 UTC from IEEE Xplore. Restrictions apply.

You might also like