IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 6, NO.

3, JULY 2015 729

Optimal Sizing of a Vanadium Redox Battery

System for Microgrid Systems
Tu A. Nguyen, Mariesa L. Crow, Fellow, IEEE, and Andrew Curtis Elmore

Abstract—The vanadium redox battery (VRB) has proven to
be a reliable and highly efficient energy storage system (ESS)
for microgrid applications. However, one challenge in designing
a microgrid system is specifying the size of the ESS. This selec-
tion is made more complex due to the independent power and
energy ratings inherent in VRB systems. Sizing a VRB for both
required power output and energy storage capacity requires an in-
depth analysis to produce both optimal scheduling capabilities and
minimum capital costs. This paper presents an analytical method
to determine the optimal ratings of VRB energy storage based
on an optimal scheduling analysis and cost-benefit analysis for
microgrid applications. A dynamic programming (DP) algorithm
is used to solve the optimal scheduling problem considering the
efficiency and operating characteristics of the VRBs. The proposed
method has been applied to determine the optimal VRB power and
energy ratings for both isolated and grid-connected microgrids,
which contain PV arrays and fossil-fuel-based generation. We first
consider the case in which a grid-tie is not available and diesel gen-
eration is the backup source of power. The method is then extended
to consider the case in which a utility grid tie is available.
Index Terms—Energy storage, optimal scheduling, optimal
sizing, microgrids, renewable energy, vanadium redox battery
(VRB).
I. I NTRODUCTION
T HE INTEGRATION of renewable energy resources in
microgrids has been increasing in the recent decades.
Higher penetration of renewable energy is an environmental-
friendly solution for improving the reliability and decreasing
costs of microgrid systems. Renewable power sources are typi-
cally highly variable depending on weather conditions, thereby
necessitating the use of highly efficient and rapid response
energy storage systems (ESSs) to store the surplus renew-
able energy and redispatch that energy when needed. One
of the most recent ESS technologies commercially available
is the vanadium redox battery (VRB). The VRB exhibits
many advantages over many traditional battery storage sys-
tems; it has independent power and energy ratings, quick charge
and discharge response, high efficiency, long life cycle, low
self-discharge, and an easily estimated state of charge (SOC)
[1]–[5]. By proper control and scheduling of the VRB sys-
tem, the microgrid operating costs (OCs) can be significantly
Manuscript received September 02, 2014; revised January 08, 2015; accepted
February 11, 2015. Date of publication April 03, 2015; date of current version
June 17, 2015. This work was supported in part by the Department of Energy
SunShot program under DE-0006341 and in part by the National Science
Foundation FREEDM ERC program. Paper no. TSTE-00450-2014.
T. A. Nguyen and M. L. Crow are with the Department of Electrical and
Computer Engineering, Missouri University of Science and Technology, Rolla,
MO 65401 USA (e-mail: tan2b4@mst.edu; crow@mst.edu).
A. C. Elmore is with the Department of Geological Engineering, Missouri
University of Science and Technology, Rolla, MO 65401 USA (e-mail:
acelmore@mst.edu).
Digital Object Identifier 10.1109/TSTE.2015.2404780
1949-3029 © 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
reduced. However, the initial capital and maintenance costs for
the VRB are still relatively high in comparison to other ESS
(such as lead-acid batteries) [6]. Therefore, it is imperative to
size and operate the VRB, such that the reduction in OCs can
justify the increase in the capital investment. The fundamen-
tal contribution of this paper is an approach to optimally select
the (independent) power and energy ratings of the VRB for the
particular microgrid in which it will operate.
The problem of sizing ESS has been addressed in a number
of studies, but the majority of previous studies have consid-
ered only lead-acid or Li-ion batteries [7]–[10]. The power
and energy ratings of these battery types are coupled; there-
fore, these approaches are not directly applicable in sizing a
VRB. However, these studies do offer several pertinent analy-
sis approaches that can be exploited for the VRB. In [6] and
[11], a microgrid-focused unit commitment approach is used
to determine the relationship between the ESS capacity and
daily-OC. A similar cost analysis to determine the optimal
size of a compressed-air storage system for large-scale wind
farms was proposed in [12]. The cost sensitivity of varying
ESS sizes and technologies for wind-diesel systems was ana-
lyzed in [13]. However, in these studies, the ESS charge and
discharge efficiencies were often neglected or considered to be
constant over the entire operating range. We expand on these
methods to explicitly include the impact of variable efficiencies
as a function of charge (discharge) rates and SOC. We extend
these earlier works to include these oversights.
As opposed to other ESS, the VRB’s charge/discharge effi-
ciencies can be characterized as explicit nonlinear functions of
charge/discharge power and the SOC [3], [4]. Furthermore, the
VRB’s power and energy ratings are independently scalable,
which allows better flexibility in sizing the VRB for microgrid
systems. Therefore, this paper focuses on the optimal sizing
of power and energy ratings for a VRB system in isolated and
grid-connected microgrids. An analytical method is developed
to solve the problem based on a per-day cost model, in which
the OC is obtained from optimal scheduling. The charge, dis-
charge efficiencies, and operating characteristics of the VRB
are considered in the problem. Case studies are performed
under different conditions of load and solar insolation.
II. F ORMULATION OF THE O PTIMAL S IZING P ROBLEM
FOR VRB M ICROGRIDS
A. Problem Definition
In this paper, the optimal size of the VRB is defined as the
independent power and the energy ratings required to minimize
the total cost (TC) per day. The proposed sizing method is
730 IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 6, NO. 3, JULY 2015

applicable to ESS such as the VRB that have independent power

and energy ratings. The approach is less suitable to storage
systems, in which the power and energy ratings are coupled,
such as lead-acid or lithium-ion batteries. The cost per day is
determined by

TC = TCPD + OC − TB (1)

where TCPD is the total capital cost amortized per day for the
VRB, OC is the total OC of the system in that day, and TB is
the total benefit (TB) achieved by selling the extra renewable
energy to the grid. If isolated microgrids are considered, then
TB is zero. Fig. 1. Total capital cost per day for a 10-kW scale system.
The following data are given as inputs to the formulation:
1) maximum and minimum load power;
2) power rating of the PV arrays and the diesel generators TI = cp PVRB
rated
+ ce EVRB
rated
+ co (3)
(DGs); where
3) historical hourly insolation profile for the site;
AC annualized capital ($/year);
4) buying and selling prices of the utility’s electricity.
TI one-time investment ($);
The system constraints are as follows. i(1+i)l
R1: The primary system load will always be met. This ensures ci = (1+i)l −1
annual carrying charge factor (equal to
that the high priority (primary) loads of the system are 0.162 with typical 10-year life cycle and a
always served. The primary loads have been identified 10% interest rate);
prior to microgrid deployment. OM fixed operation and maintenance cost
R2: Sufficient energy capacity will always be held in reserve to ($/kW year);
meet essential load demand if fuel becomes unavailable. cp power related cost ($/kWh);
This constraint ensures that if the fuel supply is compro- ce energy related cost ($/kWh);
rated
mised (through disaster, inclement weather, etc.), there PVRB power rating of the VRB (kW);
rated
is enough energy reserve to serve the primary load for a EVRB fixed cost (accounts for structure, housing,
prespecified length of time (typically between 1 and 36 h). and installation costs) ($).
R3: The VRB is not charged (or discharged) beyond the
maximum (or minimum) recommended efficiency and/or The total capital cost of the VRB system per day is, therefore
operational SOC limits. To achieve optimal performance,
AC
the VRB is not allowed to operate if its efficiency drops TCPD =
below a predetermined threshold. 365 (4)
R4: The charging rate is limited by the absorb power. The (ci cp + OM)PVRB + ci ce EVRB + ci co
= .
charging rate of the VRB is limited by electrochemical 365
processes which depend on both the VRB SOC and the In [15], the cost information is given for VRB ESSs
voltage. This constraint ensures that the charging rate with power ratings ranging from 200 kW to 10 MW
conforms to physical limitations. and with durations from 2 to 16 h. These costs are:
R5: The VRB is charged only by the PV array. One of the OM = 58.4 $/kW year, cp = 2300 $/kW, ce = 300 $/kWh
primary objectives of a microgrid is to minimize the and co = $250 000. For smaller scale VRB systems, cp and
amount of fossil fuel used; therefore, the VRB may only ce are in the same range, but co is smaller due to the smaller
be charged via renewable resources and not by the DG or structure, housing, and installation costs. In this paper, co is
the utility grid tie. estimated to be $25 000 for a 10-kW scale system (Fig. 1).
R6: The DG is not operated at light load. The DGs are most 2) OC: For an isolated PV-diesel microgrid, the OC is the
efficient when heavily loaded; therefore, it is preferred to daily fuel cost of the DGs. For a grid-connected microgrid,
run fewer generators at higher output. the OC is the sum of the diesel fuel costs and the cost of
R7: Once a generator is started (or shut down) it remains the grid electricity. The expected OC can be determined as
online (or offline) a minimum time before changing its sta- follows.
tus. This constraint minimizes “chopping” (discontinuous For an isolated microgrid:
operation) around breakpoints. m 
N 
OC = Cdgi Hdgi (Pdgi,k )T . (5)
B. Per-Day Cost Model
k=1 i=1
1) Capital Cost: The total capital of an ESS includes a one-
For a grid-connected microgrid:
time investment and the annual maintenance cost [11]. The
m 
annual financial requirement of the VRB is calculated as [14] N 
OC = Cdgi Hdgi (Pdgi,k )T + Cbuy,k Pgrid,k T (6)
AC = ci TI + PVRB
rated
OM (2) k=1 i=1
NGUYEN et al.: OPTIMAL SIZING OF A VRB SYSTEM FOR MICROGRID SYSTEMS 731

where SOCVRB,k SOC of VRB in period k;

N number of time periods in a 1-day cycle; EVRB capacity rating of the VRB;
T duration of each time period; Eresv energy reserve requirement for
m number of DGs; storage;
Hdgi (gal/h) fuel consumption of DG i (which is a VRB and SOCVRB
SOCmin max
highest and lowest allowable
function of Pdgi,k ); SOC, respectively;
Cdgi ($/gal) fuel price for DG i; Pab,k and Vab,k absorb power and absorb voltage
Cbuy,k ($/gal) electricity buying price in time period k; of the VRB in period k,
Pdgi,k dispatched power for DG i in time period respectively;
min min
k; Pdgi and Pdgi power output limits of DG i;
Cdgi ($/gal) fuel price for DG i; up/dw
Tdgi,k up/down time of DG i until period
Pgrid,k power supplied by the grid in time period up/dw,min
k, where Tdgi is the
k.
minimum up/down time for DG i.

3) Benefit: For a grid-connected microgrid, the unused

renewable power can be sold to the grid as a benefit. The TB III. M ICROGRID C OMPONENT C HARACTERIZATION
per day is as follows.
For an isolated microgrid: To operate the system at greatest efficiency, each of the sys-
tem components must be characterized with respect to their
TB = 0 (7) power and energy consumption versus efficiencies. These char-
acteristics will be integrated into the objective function (9) and
For a grid-connected microgrid: constraints R1–R7.

N
TB = Csell,k Pext,k T (8)
k=1 A. PV Array
in which Assuming that maximum power point tracking is used, the
Csell,k ($/kWh) electricity selling price in time period k; output power of the PV array can be calculated
Pextra,k power sold to the grid in time period k.
PPV = Pmax,STC Is + 0.001(Tc − 25)Kp (10)

C. Problem Formulation where

Pmax,STC (in kW) maximum power at the standard
The objective function follows from (1) in which the TC, operating condition (1 kW/m2 , 25◦ C);
defined as the TC per day (TCPD) plus the daily OC less the Is (in kW/m )
2
solar insolation;
TB, is minimized Tc (in◦ C) panel temperature;
min(TC) = min(TCPD + OC − TB) (9) Kp (in W/◦ C) sensitivity of the output power to
temperature.
where the constraints are expressed as
m
R1: Pdgi,k + PVRB,k = Pload,k − PPV,k The panel temperature can be found from the ambient tem-
i=1 perature by
R2: SOCVRB,k EVRB ≥ Eresv
R3: SOCmin
VRB ≤ SOCVRB,k ≤ SOCVRB
max
NOCT − 25
R4: PVRB,k ≤ Pab,k (SOCVRB,k , Vab,k ) Tc = I s + Ta (11)
0.8
R5: PVRB,k ≥ 0 if Pdg,k > 0
R6: P min ≤ Pdgi,k ≤ Pdgimax where NOCT is the normal operating cell temperature mea-
dgi sured at 0.8 kW/m2 and 20◦ C.
up up,min
Tdgi,k ≥ Tdgi , if generator i is online
R7: dw,min
All of the coefficients and parameters can be obtained from
Tdgi,k ≥ Tdgi
dw
, if generator i is offline manufacturers’ data or by experimental characterization.

in which
Pload,k demand in period k; B. DG
PPV,k power output of PV arrays in The fuel consumption of a DG can be characterized as a
period k; linear function of output power [16]
PVRB,k dispatched power for VRB in
period k, where PVRB,k ≥ 0 if the Hdg (Pdg ) = Bdg Pdg
rated
+ Adg Pdg (12)
VRB is discharging and
PVRB,k < 0 if the VRB is rated
where Hdg is the fuel consumption in gal/h, Pdg and Pdg are
charging; the rated power and output power of the DG in kW, respectively,
732 IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 6, NO. 3, JULY 2015

and Adg and Bdg are the coefficients of the consumption curve TABLE I
in gal/kWh. VRB P ER -U NIT M ODEL C OEFFICIENTS
The efficiency ηdg in % of DG output is calculated based on
the fuel consumption function [17]

100Pdg
ηdg = (13)
Hdg LHVgas

where LHVgas is the lower heating value of the fuel in kWh/gal,

which for diesel fuel is LHVdg = 43.75 kWh/gal.

C. VRB
SOCVRB,k is the SOC of VRB in period k, such that
 PVRB,k T
SOCk−1 − d , if discharging
SOCk = ηVRB EVRB
c
PVRB,k ηVRB T (14)
SOCk−1 − EVRB , if charging.

Note that the VRB SOC is a function of power, energy, and

d d
efficiency, where ηVRB and ηVRB are the discharge and charge
efficiencies, respectively.
The per-unit discharge and charge models are developed in
[3]. During discharge, the discharge efficiency is
Fig. 2. VRB charge and discharge efficiencies.
d
ηVRB = ηvd ηpd (15)

where IV. A NALYTICAL A PPROACH

In (4), the TCPD of the VRB has been characterized as
adv PVRB (p.u.) + bdv SOC + cdv an explicit function of power (PVRB rated
) and energy (EVRBrated
)
ηvd = (16)
aov Te SOC + bov Te + cov rated rated
ratings. At specific values of PVRB and EVRB , the TCPD
PVRB (p.u.) is constant and independent of the OC and the TB of the
ηpd = d (17)
ap PVRB (p.u.) + bdp SOC(SOC − 1) + cdp microgrid. Therefore, the process of minimizing TC becomes
the minimization of (OC − TB) by finding an optimum dis-
and ηv and ηpd are, respectively, the voltage and stack efficien- patch pattern for the generators, the grid-tie, and the VRB as
cies and Te is the electrolyte temperature in K. a function of PV power output. However, assuming constant
During charge, the VRB efficiency is power and energy ratings provides only a local optimal solution.
To obtain the global solution for the optimization, the optimiza-
rated
c
ηVRB = ηvc ηpc (18) tion should be performed over the whole domain of PVRB and
rated
EVRB . The existence of a solution for an optimal VRB size
where is guaranteed in this problem due to the fact that the TCPD is
rated
linearly increasing, whereas the OC is decreasing when PVRB
aov Te SOC + bov Te + cov rated
and EVRB are increasing, resulting in a convex optimization
ηvc = (19)
(acv SOC + bcv )PVRB (p.u.) + ccv SOC + dcv surface.
(acp SOC + bcp )PVRB (p.u.) + ccp SOC + dcp Fig. 3 shows a flowchart of the algorithm to search for opti-
ηpc = . (20) mal size of the VRB. The algorithm contains the following main
PVRB (p.u.)
steps.
The absorb power is the maximum power that the VRB can
absorb during charging when the voltage is regulated to be con-
stant. It is characterized as a function of absorb voltage and the A. Domain Definition
SOC As previously mentioned, PVRBrated rated
and EVRB are indepen-
dent, which means that the search domain is a multidimensional
Pab (p.u.) = (aab Vab (p.u.) + bab )SOC + cab Vab (p.u.) + dab . space. To narrow the search space, upper and lower limits of
(21) rated
PVRB rated
and EVRB are defined

All model coefficients are given in Table I and the charge  rated,min rated,max
PVRB ≤ PVRB
rated
≤ PVRB
and discharge efficiencies as a function of power are shown in rated,min rated,max
EVRB ≤ EVRB ≤ EVRB
rated
.
Fig. 2.
NGUYEN et al.: OPTIMAL SIZING OF A VRB SYSTEM FOR MICROGRID SYSTEMS 733

Fig. 4. PV-diesel microgrid oneline diagram.

TABLE II
DG DATA

while searching for the optimum sequence. The primary draw-

back to DP is that it can be computationally burdensome. For
Fig. 3. VRB optimal size search flowchart. example, in an N -unit system, there are 2N − 1 combinations
at each period and for M periods, the total number of com-
Due to the constraint that requires the energy storage to serve binations is (2N − 1)M [19]. For a large-scale system, the
a critical load during a minimum period of time when the fuel computation required to traverse this space can be overwhelm-
rated,min ing. However, in microgrid applications, the small number of
is not available, EVRB is defined as the minimum required
reserve. Pv rb rated,min
is chosen as the maximum of the priority units and large number of constraints significantly decrease the
loads. In this problem, it is assumed that renewable power is search space; therefore, DP can be an appropriate choice of
available to charge the VRB only if the PV power is greater algorithm [3], [20].
rated,max rated,max The solution of the optimal scheduling problem provides the
than the load power. Therefore, PVRB and EVRB can
be specified minimum value of (OC − TB) and the loading pattern for the
microgrid. The TCPD is also found from (1). At a constant
rated,max
PVRB = PPV
rated
− Pload
min
(22) rated
value, PVRB characterizing the relationship between TCPD
rated,max
EVRB = rated
PPV Tpeak − min
N Pload (23) and the energy rating provides a suboptimal solution. The pro-
rated
cess is repeated with increasing incremental values of PVRB
rated min until the upper limit is reached. The optimal solution is then
where PPV and Pload are the rated PV power and the daily
minimum load, respectively, and Tpeak is the time (in hours) obtained as the minimum value of the suboptimal incremental
at which the daily insolation is maximum. N is the number solutions.
of time increments per day over which the optimization occurs Both short-term and long-term optimization approaches
(typically 24 h). should be considered to size the battery. The proposed method
in this paper first finds the optimal ratings for the battery in a
1-day scenario (short term) and then considers different typical
B. Optimization Algorithm scenarios for the entire year (long term) to choose the solution
From the solar insolation, the available PV output power that is most cost-effective.
rate rated
is calculated from (10). At a specific PVRB and EVRB , the
optimal operation scheduling problem is set up with objec-
tive function (9) and the constraints as determined by R1–R8. V. C ASE S TUDIES R ESULTS
This UC problem can be categorized as a sequential decision- The case studies are performed to illustrate the optimization
making problem for which dynamic programming (DP) is well method. These cases are based on an actual microgrid system
known. DP is a method to find the shortest route to the desti- located at the Fort Leonard Wood U.S. Army base in Missouri,
nation by breaking it down to a sequence of steps over time; USA (Fig. 4), which can be either isolated or grid-connected.
at each step, it finds the possible optimum sequences (routes) The ac load ranges between 2 and 5 kW. The PV arrays are
based on the possible optimum subsequences in the previ- rated 15 kW. Hourly data of the solar insolation and ambient
ous steps and finally finds the optimum sequence at the last temperature at the site were collected from December 2012 to
step. The main advantage of DP is that it can maintain solu- September 2013. The field results of the Fort Leonard Wood
tion feasibility by its ability to find the optimum subsequence U.S. Army base system upon which this optimization process
734 IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 6, NO. 3, JULY 2015

Fig. 5. PV output power data.

Fig. 6. Dispatch on 1 day of June 2013 at minimum load and PVRB
rated = 5 kW

and EVRB
rated = 65 kWh.

is based are given in [3]–[5] and are not reproduced here. The
DG maximum output is 8 kW. The initial state and fuel con-
sumption of the DG are given in Table II. Fuel usage at forward
operating military bases will vary with size, location, and mis-
sion. By the time fuel reaches in-theater staging, it can cost 10
or more times the domestic price [21]. Similarly, fuel costs in
extreme rural areas experience significant inflation due to trans-
portation costs. For example, delivering fuel to rural areas of
Alaska is complicated and expensive, since fuel must typically
be delivered by either barge or plane [22]. Electricity price will
also vary widely with locations. The electricity in the remote
areas can cost double or more the price in urban cities due to
the high cost of transmission and generation. For example, the Fig. 7. Costs for operation in June 2013 at minimum and maximum load for a
average electricity price for residential customers in Hawaii in power rating PVRB
rated of 5 and 10 kW.

June 2014 is c 38.66/kWh [23]. Therefore, to better approxi-

mate the deployed performance of the microgrid, a diesel fuel
cost of 40/gal, electricity price of c 40/kWh, and buy back price
of c 20/kWh were used in this study.
The battery is required to have a reserve of 2 kWh, which is
sufficient to run the minimum critical load for 1 h. The VRB
may discharge to an SOC of 20%. The VRB is assumed to be
initialized at the minimum 20% SOC at the beginning of the
day (12 A . M .). The VRB rating range considered is
5 kW ≤ PVRB
rated
≤ 10 kW (24)
10 kWh ≤ rated
EVRB ≤ 100 kWh. (25)
The optimization is solved using PV data from 4 days in
December 2012, March 2013, June 2013, and September 2013
(Fig. 5), which represent typical scenarios of solar insolation
ranging from a low in December 2012 to a high in September
2013. From the given weather data, the PV power output is
calculated from (10) for both the minimum and the maximum
system loads. The optimization process illustrated in Fig. 3
is used to find the optimal size of the VRB for both maxi-
mum load and minimum load scenarios for both isolated and
grid-connected cases.
A. Case Study I—Isolated Microgrid
In this case study, the microgrid is isolated and the load must
be met through the combination of PV, diesel generation, and
VRB energy storage. The optimization process yields a series
of operating scenarios that have associated costs. As an example
of these, the costs for the minimum and maximum load cases
for June 2013 are shown in Fig. 7 for both power ratings. Note
that the cost axis for minimum load is on the right side of the
figure and the maximum load cost is on the left axis. There
is not a single combination of power and energy ratings that
provides a consistent minimum, but there is definitely a com-
monality between the traces. Note that the minimum cost occurs
at a 65 kWh VRB rating except in the 5 kW VRB minimum load
case, in which the minimum occurs at 60 kWh rating. However,
the minimum costs between the 65 and 60 kWh ratings are very
close.
Figs. 6 and 8 show the operating profiles of the (5 kW,
65 kWh) rated VRB for June 2013 under minimum and max-
imum loads, respectively. In the minimum load case, the DG
never engages which is why the OCs are much smaller than
for the maximum load case. In the maximum load case, the
DG runs at night, but note that it ramps down as the sunlight
increases. Also note that when the diesel hits the minimum
limit (35% loading) at just before 09:00, it is disengaged and
the VRB picks up the load in correspondence with constraint
R6. The VRB also picks up the load at 16:00 h to keep the
DG OFF during light load. Also note that the DG is OFF for the
majority of the sunlight hours during which the solar array can
simultaneously serve the load and charge the VRB according to
constraint R5. According to constraint R3, the VRB is always
charged or discharged when its efficiency is high (see Fig. 2).
This process is repeated for the remaining three operating
profiles of December 2012, March 2013, and September 2013.
For example, similar results for September 2013 are shown in
NGUYEN et al.: OPTIMAL SIZING OF A VRB SYSTEM FOR MICROGRID SYSTEMS
Fig. 8. Dispatch on 1 day of June 2013 at maximum load and PVRB
rated = 5 kW
and EVRB
rated = 65 kWh.
Fig. 9. Costs for operation in September 2013 at minimum and maximum load
for a power rating PVRB
rated of 5 and 10 kW.
Fig. 10. Dispatch on 1 day of September 2013 at minimum load and PVRB
rated =
5 kW and EVRB
rated = 65 kWh.
Figs. 9–11. In this month, the optimum energy rating shows
more variation than in June, but there is still consistency.
The minimum cost and energy rating across all months
and power ratings for minimum loading is given in Table III.
Similarly, the minimum cost and energy rating across all
months and power ratings for maximum loading is given in
Table IV.
From these results, it is possible to select the rating of the
VRB based on the optimum costs. As expected, the lowest costs
occur during the months in which there is the largest amount of
solar insolation. Note that the PV arrays are rated at 15 kW.
The PV panels seldom output their maximum rated power and
only do so for a short period of time during the day (for fixed
mounting panels) and for a few months of the year. Thus, only
735
Fig. 11. Dispatch on 1 day of September 2013 at maximum load and PVRB
rated =
5 kW and EVRB
rated = 65 kWh.
TABLE III
O PTIMUM C OST AND E NERGY R ATING AT M INIMUM L OAD
TABLE IV
O PTIMUM C OST AND E NERGY R ATING AT M AXIMUM L OAD
for a limited time is there any potential advantage for the extra
power capacity provided by the 10 kW VRB and as the opti-
mum costs indicate, the 10kW VRB is never cost effective. In
evaluating the energy ratings of the VRB, the optimal size can
be selected by noting the optimal rating sizes for the seasons in
which energy storage is effective. For both minimum and max-
imum loading profiles, a VRB rated (5 kW, 65 kWh) provides
the best cost performance over the longest period of time.
The disclaimer to these conclusions is that the VRB power
and energy ratings must be evaluated based on the typical inso-
lation profile and diesel fuel cost for the location in which the
system is to be located, but the evaluation procedure will remain
the same.
B. Case Study II—Grid-Connected Microgrid
Based on the operating strategies of a grid-connected micro-
grid, the utility grid can be treated as a backup or the main
source of power. The former operating strategy is often used
when the accessibility to the grid is limited, whereas the latter
is applied when the fuel supplies for the DGs are limited.
1) Case II.1: In this case, the microgrid’s DGs are the main
sources to balance the system’s load. The grid power is only
used to cover the peak load in a short period of time. As shown
in Tables V and VI, the results are only slightly different from
those of the isolated system from Case study I. It is because
the exchanged power between the microgrid and the grid is too
small to have significant impact on the TCPD. Therefore, the
optimal VRB’s ratings are similarly chosen at (5 KW, 65 kWh).
736
TABLE V
O PTIMUM C OST AND E NERGY R ATING AT M INIMUM L OAD
TABLE VI
O PTIMUM C OST AND E NERGY R ATING AT M AXIMUM L OAD
TABLE VII
O PTIMUM C OST AND E NERGY R ATING AT M INIMUM L OAD
TABLE VIII
O PTIMUM C OST AND E NERGY R ATING AT M AXIMUM L OAD
2) Case II.2: In this case, the microgrid’s DGs are only
used as backup source of power. The system’s load is mainly
powered by the grid or the PV arrays. Results are shown in
Tables VII and VIII. The optimal TCPD for different months of
the year are observed at the minimum power and energy ratings
of the VRB—(5 kW, 5 kWh) in most scenarios. The VRB is
only a cost-effective serving minimum load in September when
the PV power is high. This is because the investment cost of the
VRB is high compared to the electricity cost of the utility grid;
therefore, the VRB is not cost effective in most scenarios when
a grid-tie is available.
VI. C ONCLUSION
In this paper, an analytical approach has been developed
for optimal sizing of VRB storage system for isolated micro-
grids. Nonlinear charge/discharge efficiencies and operating
characteristics of VRB system are considered in the system’s
constraints. The feasibility of the solution is ensured based on
per-day cost model. The method can also be valid for opti-
mal sizing of different types of ESSs if the cost and efficiency
model are known. Case studies have been performed to find
optimal VRB power and energy ratings for both isolated and
grid-connected microgrids considering different scenarios of
load and insolation. The results show that the optimal solution
occurs when the OC benefit by increasing VRB capacity can
justify the increment in investment capital.
IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 6, NO. 3, JULY 2015
R EFERENCES
[1] A. Poullikkas, “A comparative overview of large-scale battery systems
for electricity storage,” Renew. Sustain. Energy Rev., vol. 27, pp. 778–788,
2013.
[2] M. Mohamed, P. Leung, and M. Sulaiman, “Performance characterization
of a vanadium redox flow battery at different operating parameters under
a standardized test-bed system,” Appl. Energy, vol. 137, pp. 402–412,
2015.
[3] T. Nguyen, X. Qiu, J. Guggenberger, II, M. Crow, and A. Elmore,
“Performance characterization for photovoltaic-vanadium redox battery
microgrid systems,” IEEE Trans. Sustain. Energy, vol. 5, no. 4, pp. 1379–
1388, Oct. 2014.
[4] X. Qiu, T. Nguyen, J. Guggenberger, M. L. Crow, and A. C. Elmore, “A
field validated model of a vanadium redox flow battery for microgrids,”
IEEE Trans. Smart Grid, vol. 5, no. 4, pp. 1592–1601, Jul. 2014.
[5] J. Guggenberger, A. C. Elmore, J. Tichenor, and M. L. Crow,
“Performance prediction of a vanadium redox battery for use in portable,
scalable microgrids,” IEEE Trans. Smart Grid, vol. 3, no. 4, pp. 2109–
2116, Dec. 2012.
[6] S. Bahramirad, W. Reder, and A. Khodaei, “Reliability-constrained opti-
mal sizing of energy storage system in a microgrid,” IEEE Trans. Smart
Grid, vol. 3, no. 4, pp. 2056–2062, Dec. 2012.
[7] B. Borowy and Z. Salameh, “Methodology for optimally sizing the com-
bination of a battery bank and PV array in a wind/PV hybrid system,”
IEEE Trans. Energy Convers., vol. 11, no. 2, pp. 367–375, Jun. 1996.
[8] M. Elhadidy and S. Shaahid, “Optimal sizing of battery storage for hybrid
(wind+diesel) power systems,” Renew. Energy, vol. 18, no. 1, pp. 77–86,
1999.
[9] P. Arun, R. Banerjee, and S. Bandyopadhyay, “Optimum sizing of
battery-integrated diesel generator for remote electrification through
design-space approach,” Energy, vol. 33, no. 7, pp. 1155–1168, 2008.
[10] M. R. Aghamohammadi and H. Abdolahinia, “A new approach for
optimal sizing of battery energy storage system for primary frequency
control of islanded microgrid,” Int. J. Elect. Power Energy Syst., vol. 54,
pp. 325–333, 2014.
[11] S. Chen, H. Gooi, and M. Q. Wang, “Sizing of energy storage for
microgrids,” IEEE Trans. Smart Grid, vol. 3, no. 1, pp. 142–151, Mar.
2012.
[12] H. T. Le and T. Q. Nguyen, “Sizing energy storage systems for wind
power firming: An analytical approach and a cost-benefit analysis,” in
Proc. IEEE Power Energy Soc. Gen. Meeting, 2008, pp. 1–8.
[13] M. Ross, R. Hidalgo, C. Abbey, and G. Joos, “Analysis of energy storage
sizing and technologies,” in Proc. IEEE Elect. Power Energy Conf., 2010,
pp. 1–6.
[14] S. M. Schoenung and W. V. Hassenzahl, “Long- vs. short-term energy
storage technologies analysis—A life-cycle cost study,” Sandia Nat. Lab.,
Albuquerque, NM, Tech. Rep., SAND2003-2783, 2003.
[15] EPRI, “Vanadium redox flow batteries, an in-depth analysis,” EPRI, Palo
Alto, CA, USA, Tech. Rep. 1014836, 2007.
[16] O. Skarstein and K. Ulhen, “Design considerations with respect to long-
term diesel saving in wind/diesel plants,” Wind Eng., vol. 13, no. 2,
pp. 72–87, 1989.
[17] L. Rodolfo Dufo and A. Jose L. Bernal, “Multi-objective design of PV-
wind-diesel-hydrogen-battery systems,” Renew. Energy, vol. 33, no. 12,
pp. 2559–2572, 2008.
[18] T. Nguyen and M. Crow, “Optimization in energy and power management
for renewable-diesel microgrids using dynamic programming algorithm,”
in Proc. IEEE Int. Conf. Cyber Technol. Autom. Control Intell. Syst.,
2012, pp. 11–16.
[19] A. J. Wood and B. F. Wollenberg, Power Generation, Operation, and
Control, 2nd ed. Hoboken, NJ, USA: Wiley, 1996.
[20] L. M. Costa and G. Kariniotakis, “A stochastic dynamic programming
model for optimal use of local energy resources in a market environ-
ment,” in Proc. IEEE Lausanne Power Tech, Lausanne, Switzerland,
2007, pp. 449–454.
[21] Strategic Environmental Research and Development Program (SERDP),
“Sustainable forward operating bases,” Noblis, Annadale, VA, USA, Aug.
2010.
[22] AK Attorney General. (2009). Rural Fuel Pricing in Alaska [Online].
Available: www.law.state.ak.us/pdf/civil/021810RuralFuelPricinginAlas
ka.pdf
[23] U.S. Energy Information Administration. (2014). Electric Power Monthly
[Online]. Available: www.eia.gov/electricity/monthly/pdf/epm.pdf
NGUYEN et al.: OPTIMAL SIZING OF A VRB SYSTEM FOR MICROGRID SYSTEMS 737

Tu A. Nguyen received the B.S degree in power Andrew Curtis Elmore received the B.S. degree
systems from Hanoi University of Science and in geological engineering from the University of
Technology, Hanoi, Vietnam, in 2007, and the Ph.D. Missouri, Rolla, MO, USA, and the M.S. and Ph.D.
degree from Missouri University of Science and degrees in civil engineering from the University of
Technology, Rolla, MO, USA, in 2014. Arizona, Tucson, AZ, USA.
He worked as a Power Transformer Test Engineer He was employed as a Consulting Engineer with
with ABB High Voltage Test Department, Hanoi, URS Group, Overland Park, KS, USA, where he
Vietnam, from 2008 to 2009. His research inter- specialized in green and sustainable environmental
ests include microgrid system modeling/analysis and remediation. He is currently an Associate Professor
power electronics applications in microgrid systems. of Geological Engineering with the University of
Missouri.
Mariesa L. Crow (S’83–M’90–SM’94–F’10)
received the B.S.E. degree from the University
of Michigan, Ann Arbor, MI, USA, and the
Ph.D. degree from the University of Illinois at
Urbana-Champaign, Champaign, IL, USA.
She is currently the F. Finley Distinguished
Professor of Electrical Engineering with the Missouri
University of Science and Technology, Rolla, MO,
USA. Her research interests include computational
methods for dynamic security assessment and the
application of energy storage in bulk power systems.
Dr. Crow is a Registered Professional Engineer.