You are on page 1of 12

pubs.acs.

org/journal/ascecg Research Article

Locating Shunt Currents in a Multistack System of All-Vanadium


Redox Flow Batteries
Han-Wen Chou, Feng-Zhi Chang, Hwa-Jou Wei, Bhupendra Singh, Amornchai Arpornwichanop,
Prathak Jienkulsawad, Yi-Sin Chou, and Yong-Song Chen*

Cite This: ACS Sustainable Chem. Eng. 2021, 9, 4648−4659 Read Online

ACCESS
See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

Metrics & More Article Recommendations


Downloaded via UNIV OF NEW SOUTH WALES on January 29, 2024 at 13:53:23 (UTC).

ABSTRACT: An all-vanadium redox flow battery (VRFB) system,


with multiple stacks, is typically used for large-scale electrical energy
storage applications. In a VRFB system, pumps deliver positive and
negative electrolytes, through a piping system, to each stack.
Because the electrolytes are electrically conductive, shunt currents
can occur within a multicell stack and within the piping system,
connecting the stacks due to the voltage differences between cells
and between stacks. Shunt currents cause energy loss and are
affected by the number of cells in a single stack, the number of
stacks, and the piping system dimensions. In this study, we develop
a mathematical model, based on Kirchhoff’s law, to locate shunt
currents in a multistack system. Using this model, we estimate the
charge efficiency with various numbers of stacks. The results show
that the shunt currents in the central stacks are larger than the currents in other stacks. In addition, the piping system dominates the
distribution of the electrolytes, and the shunt currents gradually shift from inside the stack to the piping system.
KEYWORDS: All-vanadium redox flow battery, Multistack, Shunt current, Piping system, System efficiency

■ INTRODUCTION
With the rapid development of green energy technologies, such
currents can occur in the manifolds and distribution channels
within the stacks as well as in the piping system. The existence of
as solar and wind power, large-scale energy storage systems are shunt currents decreases the system’s capacity and energy
required to integrate with the power distribution system and efficiency.6 Xiong et al.7 developed a hydraulic model to estimate
the pump power consumption for optimal battery efficiency.
smooth out peaks and troughs in power generation.1 Among the
They found that the design of the distribution channels and the
proposed energy storage technologies, all-vanadium redox flow
manifold is a trade-off between shunt current losses and pump
batteries (VRFBs) have attracted much attention and are now at
power losses due to hydraulic resistance.8
the demonstration and commercialization stage. The VRFB,
Kaminski et al.9 proposed a technique for calculating shunt
employing V2+/V3+ and VO2+/VO+2 as redox couples in the
current in a multicell stack. Xing et al.10 developed a model to
negative and positive electrolytes, respectively, shows com-
study shunt current losses in a VRFB stack based on the
petitive advantages, such as a long cycle life, low self-discharge,
electrical circuit analog method. Their results suggested that
low maintenance cost, and flexible design in terms of stack
increasing the power of individual cells, reducing the number of
power, as well as being environmentally friendly.2−4 In a VRFB
cells, and increasing the electrolyte resistance in the manifolds
stack, multiple cells, connected in series, provide a high voltage.
and channels can help reduce shunt current losses. Moro et al.11
During VRFB stack operation, pumps deliver electrolytes into
applied a three-dimensional, finite-element method to compute
the stack. The electrolytes flow through manifolds connecting
shunt currents in the piping system and manifolds, using an
the cells and are distributed to each cell through flow channels
equivalent circuit, and solved the model using Matlab. The shunt
that connect the manifold and the active areas.5
Connecting multiple VRFB stacks, in series or parallel, can
meet the voltage and power requirements of a large-scale Received: January 14, 2021
renewable energy storage system. Although there are multiple Revised: March 2, 2021
stacks in such a system, two recirculation pumps, one on each Published: March 18, 2021
side, supply the electrolytes through the piping system to all the
stacks. The electrolytes flow through the manifold to each stack.
Because the electrolytes are electrically conductive, shunt

© 2021 American Chemical Society https://doi.org/10.1021/acssuschemeng.1c00287


4648 ACS Sustainable Chem. Eng. 2021, 9, 4648−4659
ACS Sustainable Chemistry & Engineering pubs.acs.org/journal/ascecg Research Article

Figure 1. (a) Electrolyte flow paths in a multistack VRFB system; (b) equivalent electrical circuit of a system with multiple stacks in series.

4649 https://doi.org/10.1021/acssuschemeng.1c00287
ACS Sustainable Chem. Eng. 2021, 9, 4648−4659
ACS Sustainable Chemistry & Engineering pubs.acs.org/journal/ascecg Research Article

currents within a VRFB result in Joule heating, which causes an investigated the effects of combining different numbers of stacks
increase in temperature. Trovò et al.12 proposed a thermal and cells in the system, the pipe dimensions, and the electrolyte
model to study the dynamic cell temperature distribution in a flow rate on the system efficiency. This study helps optimize cell
VRFB due to the heat generated by shunt currents and self- and stack numbers in a VRFB system and reduce the
discharge effects. development cost.
Tang et al.13 investigated the shunt current’s effect on
efficiency and temperature variation within a 40-cell stack during
the standby condition by combining the shunt current model
■ MATHEMATICAL MODEL
Shunt Current Model. Designing a large-scale VRFB system and
and the thermal model. Their results showed that the stack its piping is complicated, and the design parameters may vary case by
temperature increased significantly because of shunt currents case, depending on different applications.20 To simplify the design and
within the VRFB stack. modeling parameters, the total number of cells in the model was fixed at
Only a few studies have reported experiments to measure 120 for a common voltage requirement,21 and the number of VRFB
stacks was varied from 2 to 24. Figure 1(a) shows the electrolyte flow
shunt currents in a VRFB because it is difficult to install current paths within the stacks and the piping system. Individual pumps supply
sensors in the manifolds and distribution channels inside a positive and negative electrolytes to each stack through the piping
VRFB stack. Fink et al.14 measured the shunt currents of a stack system. The manifolds are the pipes that connect to the pumps, and the
in which the flow paths between cells were connected externally. channels are the branches connecting the manifolds and the stacks. In
Their experiment measured shunt currents of the external the stack, electrolytes flow through the manifolds within the stack to
hydraulic system of a five-cell stack and developed a each cell and through the channels connecting the manifolds and the
mathematical model to investigate how the shunt currents active areas. Figure 1(b) shows the equivalent electrical circuit diagram
affect Coulombic efficiency. The results showed that inner cells of the shunt currents in the system. The equivalent resistance of the
electrolytes in the flow paths can be predetermined from the
within a stack discharge faster than outer cells. Yin et al.15 also
dimensions of the channels and the manifolds. Rpc,P, Rnc,P, Rpm,P, and
developed a three-dimensional model to investigate the shunt Rnm,P represent the electrolytes’ resistance within the piping system’s
current distribution’s effects on Coulombic efficiency. They positive channels, negative channels, positive manifolds, and negative
validated the model by experimental data, and the results manifolds, respectively. Rpc,S, Rnc,S, Rpm,S, and Rnm,S represent the
showed that a short channel design caused a Coulombic equivalent resistances of the electrolyte within the positive channels,
efficiency loss of approximately 23%. negative channels, positive manifolds, and negative manifolds in the
The shunt current distribution becomes more complicated in stack, respectively.
a multiple stack system because of inside-stack flow paths and We simplified the model based on the following assumptions.
interstack piping. The overall shunt current distribution depends (1) The distributions of temperature and electrolyte concentration
within the stack and the piping system are uniform.
on the design of the stacks and the piping system, such as the (2) Each cell’s electrical potential is uniform throughout the active
number of cells in a single stack, the number of stacks in a series, area.
and the dimensions of the piping system. To date, there are only (3) The inlet and outlet flow paths are symmetrical, so shunt currents
a few published works regarding shunt currents in a multistack in the inlet and outlet paths are the same in a single cell and in the piping
VRFB system. system.
Wandschneider et al.16 developed a mathematical model to In this system, there are Z stacks connected in series, and each stack
investigate how the piping layout affected a three-stack system. consists of K cells, resulting in M = KZ cells in total. Appling Kirchhoff’s
The results showed an uneven distribution of shunt currents in law to the ith cell of the equivalent electrical circuit for the stack results
the pipes. The piping system connecting the stacks also in the following linear equations.
increased the internal shunt currents within the stacks. König ISj , i − ISj , i − 1 − 2Ipc,Sj , i − 2Inc,Sj , i = 0 (1)
et al.17 developed a model to investigate the pipe diameter’s
effect on system efficiency for six stacks connected in series. Ipc,Sj , i + Ipm,Sj , i − Ipm,Sj , i − 1 = 0 (2)
Their results showed that the optimal pipe diameter depended Inc,Sj , i + Inm,Sj , i − Inm,Sj , i − 1 = 0 (3)
on the method used to connect the stacks. Ye et al.18
systematically studied the trade-off between shunt current losses VSj , i − R pc,SIpc,Sj , i + R pm,SIpm,Sj , i + R pc,SIpc,Sj , i + 1 = 0 (4)
and pumping losses in a multistack VRFB system considering
numerous parameters, including the number of stacks and the VSj , i − R nc,SInc,Sj , i + R nm,SInm,Sj , i + R nc,SInc,Sj , i + 1 = 0 (5)
resistances of the intrastack channels and interstack pipes. where ISj,i and VSj,i represent the current and voltage of the ith cell in the
Although optimal design parameters will vary case by case, their jth stack of the system, respectively.
study suggested guidelines for designing a large-scale VRFB The inlets and outlets of the electrolytes are placed at the end plate,
system. Chen et al.19 studied the transport delay of the close to the first cell of each stack; as a result, the current and voltage
electrolyte flow in the piping systems of a multistack VRFB relations for the positive electrolyte of the first cell in each stack are
module using a dynamic model. They found they could modified such that
minimize the effect of the transport delay on system perform- ISj ,1 − ITj − 2Ipc,Sj ,1 = 0 (6)
ance and energy capacity by optimizing the flow rate, pipe
design, and feed mode of the electrolyte. Ipc,Sj ,1 + Ipm,Sj ,1 − Ipc,P, j = 0 (7)
Most of the aforementioned studies focused on the trade-off
between shunt current losses and pumping power losses. VSj ,1 − R pcIpc,Sj ,1 + R pmIpm,Sj ,1 + R pcIpc,Sj ,2 = 0 (8)
However, there are no in-depth studies of shunt current where ITj is the overall current of a single stack. Similarly, the negative
distribution in the piping system that connects the stacks or the electrolyte is connected with the negative piping system:
effects of combining different numbers of stacks and cells. In this
Inc,Sj ,2 + Inm,Sj ,2 − Inc,P, j = 0 (9)
study, we developed a mathematical model, based on Kirchhoff’s
law, to investigate the shunt currents’ locations and distribution The last cell in each stack does not have a complete positive loop, so
within each stack and within the piping system. We also the related equations are modified such that

4650 https://doi.org/10.1021/acssuschemeng.1c00287
ACS Sustainable Chem. Eng. 2021, 9, 4648−4659
ACS Sustainable Chemistry & Engineering pubs.acs.org/journal/ascecg Research Article

Figure 2. Comparison of the experimental data and simulated charge−discharge curves.

ISj , M − ISj , M − 1 − 2Ipc,Sj , M − 2Inc,Sj , M = 0 (10) V = AI (24)


where V is the vector that mainly consists of cell voltages, A is the matrix
ISj ,pc, M − ISj ,pm, M − 1 = 0 (11) with precalculated resistances, and I is the current vector to be solved.
Inc, M + 1 − Inm , M = 0 Solution Process. In this proposed model, all cell currents and cell
(12) voltages are coupled together. To solve eq 24, each cell voltage needs to
In the piping system, similar equations for the nodes and circuit loops be determined; however, the cell current is required for calculating the
can be obtained by the same method: cell voltage according to eq 23. As noted, it is possible to solve these
coupled equations by an iterative procedure. In the beginning, all the
Ipc,P, j + Ipm,P, j − Ipm, P , j − 1 = 0 (13) cell currents were initially set to be equal to the operating current, and
the cell voltages were then obtained using eq 23. The calculated cell
Inc, P , j + 1 + Inm,P, j + 1 − Inm,P, j = 0 (14) voltage can then be placed in the voltage vector of eq 24 to calculate all
currents, including cell and shunt currents. The cell currents were
VSj − R pc,SIpc,Sj ,1 − R pc,PIpc,P, j + R pm,PIpm,P, j + R pc,PIpc,P, j + 1 plugged into eq 23 to check the convergence.
+ R pc,SIpc,Sj + 1,1 = 0

(VSj − VSj ,1) − R nc,SInc,Sj ,2 − R nc,PInc,P, j + 1 + R nm,PInm,P, j + 1


(15)
■ RESULTS AND DISCUSSION
Model Validation. The model was validated based on
experimental data published in a previous study.22 Briefly, the
+ R nc,PInc,P, j + 2 + R nc,SInc,Sj + 2,2 + VSj + 1,1 = 0 (16) data were obtained by running a VRFB stack comprising 19 cells
The currents in the flow channels and manifolds of the piping system
with an active area of 50 × 18 cm2, sulfuric acid-treated graphite
for the first and last stacks are special cases: felt electrodes, and a Nafion 117 (DuPont, United States)
membrane separator. The electrolyte comprised vanadium and
Ipc,P,1 + Ipm,P,1 = 0 (17) sulfate ion concentrations of 1.68 and 4.8355 M, respectively,
Inc, P , j + 1 + Inm,P, j + 1 = 0 which were recirculated between the stack and the electrolyte
(18)
tanks using pumps (NF1.600, KNF, Germany) at a flow rate of
Ipc,P, z + Ipm,P, z − 1 = 0 (19) 2.5 L min−1. The stack was operated at two different operating
current levels of 36 and 54 A (40 and 60 mA cm−2) with the
Inc, P , z + 1 − Inm,P, z = 0 (20) voltage limits of 19 and 29 V, respectively, during discharging
The system voltage is the summation of the stack voltages and charging processes.
z
Figure 2 shows voltages of simulated charge−discharge curves
VSYS = ∑ VS, j versus SOC, for the 36 and 54 A operating current levels, with
j=1 (21) the corresponding experimental data. As can be seen, for both
current levels, the simulated voltages are in good agreement with
and the system current is the same as the stack current the experimental voltages. The slight divergence at the
ISYS = IT, j (22) extremities of the curves may be caused by the nonuniform
performance distribution of each cell within the stack during the
Assuming the cell voltage is a function of operating current: charge−discharge processes.
0
Vi = f (Ii) = V cell − R cellIi (23) Model Convergence and Uncertainty. Since the cell and
flow channel design of the stack are the same as they were in a
where the parameter Rcell is the equivalent resistance of each cell, previously published study,22 the electrolyte conductivity and
including the active, ohmic, and concentration resistances.
We can calculate the resistances of the electrolytes in the flow paths dimensions of the flow channel and manifold in the stack can
in advance if the flow paths’ dimensions and geometry are known. therefore be obtained, and the equivalent resistances Rcell, Rpc,S,
Shunt currents and cell currents are unknown and are to be determined Rnc,S, Rpm,S, and Rnm,S at the SOC of 0.5 are 4.4 × 10−3, 89.5, 89.5,
by this model. The overall number of unknowns in the model is 5M − 2. 0.376, and 0.376 Ω, respectively. In the piping system, the length
All linear equations can be expressed in a matrix form: of the connected pipes varies with the number of cells in the
4651 https://doi.org/10.1021/acssuschemeng.1c00287
ACS Sustainable Chem. Eng. 2021, 9, 4648−4659
ACS Sustainable Chemistry & Engineering pubs.acs.org/journal/ascecg Research Article

stack. The channel distance from the manifold of the piping


system to each stack is approximately 15 cm. The thickness of
each cell was 1 cm, including the membrane, gaskets, and bipolar
plates. Table 1 lists the piping system dimensions and the
corresponding equivalent electrolyte resistances.

Table 1. Pipe Length and Resistance for Various Numbers of


Stacks
cell manifold length of piping manifold resistance in
stack number in between stacks, the piping Rpm,P, Rnm,P
number, Z a stack, K Lm = 40 + K (cm) (Ω)
2 60 100 42.16
3 40 80 33.72
4 30 70 29.51
5 24 64 26.98
6 20 60 25.29
8 15 55 23.19
12 10 50 21.08
24 5 45 18.97

The model uncertainty can be caused by inconsistent cell


resistances due to nonuniform flow distribution or assembling
uncertainty. Moreover, slight discrepancies in the cell voltage
may arise due to the ion crossover and the side reactions that are
not considered in numerical simulation. As a result, 5% and 10%
variation in cell resistance were assumed among the 120 cells in
eq 23 for the analysis of model uncertainty.
The convergence of the model was tested by comparing
distributed cell voltages determined by the iteration process
mentioned in the solution process. When the cell resistances are
all the same, the distributed cell voltages in the system can be
obtained from the initial setting, first time calculation, and
Figure 3. Convergence test of the model with six twenty-cell stacks. (a)
second time calculation and are shown in Figure 3(a). The cell All cell resistances are the same; (b) 5% variation in cell resistance; and
voltages obtained from the first and second calculations were (c) 10% variation in cell resistance.
very close. Figure 3(b) and 3(c) show the results for 5% and 10%
variation in cell resistance, respectively. It can be seen that
convergent results can be obtained after the second time The minimum cell current increases with an increasing
calculation. number of stacks because of the decrease in the overall stack
The effect of variation in cell resistance on voltage distribution voltage. When the number of stacks was three or more, the cell
is shown in Figure 4(a). The 5% variation in cell resistance current gradually approached the applied charge current,
resulted in a voltage uncertainty of less than ±0.6%, whereas indicating a reduction in shunt current loss. The cell current
10% variation resulted in a voltage uncertainty of ±1.6%. The differences within a single stack and between stacks also
effect of variation in cell resistance on the current distribution is decreased with an increasing number of stacks. When 24 stacks
not meaningful, as shown in Figure 4(b). In summary, the model were connected in series, the charging current for the VRFB was
is robust and reliable for further analysis. In the following approximately 52 A for most of the stacks.
section, the modeling results obtained from the second The cell current distribution during the discharging process
calculation with the same cell resistance are used for analysis. showed a similar distribution to that during charging, as shown
Shunt Current Distribution. We simulated the shunt in Figure 5(b). Most central cells within a single stack generated
current distribution for a 120-cell system with various numbers a current higher than the load current to compensate for shunt
of stacks, specifically 2, 3, 4, 5, 6, 8, and 12. The effect of the current losses. The cell current decreased with an increasing
number of stacks on the cell current distribution with an number of stacks, and the cell current distribution for most of
operating current of 54 A (60 mA cm−2) is shown in Figure 5. the stacks became uniform. This is because of a reduced voltage
Figure 5(a) shows that, during the discharge process, the cell difference between the cells with a decreasing number of cells in
current within a single stack had a U-shaped distribution, and the a single stack.
minimum current value was located in the central cells. When Figure 6 presents the shunt currents in the flow channels and
two stacks were connected in series, the minimum cell current manifolds of each stack for various numbers of stacks in series
values in each stack were −36 (−40 mA cm−2) and −41.5 A when the system was discharged at 54 A. In each stack, the shunt
(−46 mA cm−2). In addition, the U-shaped distribution within currents in the manifolds were larger than the currents in the
each stack was asymmetrical and inclined toward the left. This is flow channels because the manifolds had the smallest electrolyte
because the inlet and exit pipes were connected to the left-hand resistance, which is a result of the short distance between
side of each stack (cells 1 and 61), resulting in a large shunt adjacent cells. Although increasing the thickness of the bipolar
current loss for the left stack. plates or the porous electrodes can increase the resistance, it will
4652 https://doi.org/10.1021/acssuschemeng.1c00287
ACS Sustainable Chem. Eng. 2021, 9, 4648−4659
ACS Sustainable Chemistry & Engineering pubs.acs.org/journal/ascecg Research Article

Figure 4. Effect of variation in cell resistance on the distributions of (a) cell voltage and (b) cell current.

Figure 5. Effect of the number of stacks on cell current distribution during (a) charging and (b) discharging processes.

4653 https://doi.org/10.1021/acssuschemeng.1c00287
ACS Sustainable Chem. Eng. 2021, 9, 4648−4659
ACS Sustainable Chemistry & Engineering pubs.acs.org/journal/ascecg Research Article

Figure 6. Shunt current distributions within the stacks during the charge process for (a) two 60-cell stacks, (b) three 40-cell stacks, (c) four 30-cell
stacks, (d) five 24-cell stacks, (e) six 20-cell stacks, (f) eight 15-cell stacks, (g) twelve 10-cell stacks, and (h) twenty-four 5-cell stacks.

also significantly increase the stack volume, causing additional in shunt currents in the manifolds to balance those in the piping
material costs and reducing the specific power density. Further, system.
the maximum shunt current in the manifolds is located in the Figure 6 also shows that the shunt current magnitudes in the
central cells of each stack regardless the number of stacks in the flow channel or manifold decreased with an increasing number
series. This is because most of the cell pairs contribute to the of stacks due to the decrease in the voltage difference between
shunt currents in the central manifolds. For example, the shunt the cells and the stacks. During the discharge process, the shunt
current distribution showed a similar trend, except for a slight
current in the manifold that connects cells a and b may be due to
difference in the magnitude of the shunt currents, as shown in
the voltage difference between cells q and k, where q ≤ a and k ≥ Figure 7.
b. The combination of the contributions of q and k for the central Shunt Currents in the Piping System. The trends in the
cells is greater than that of the other cells. variation of the shunt current distribution in the piping system
The U-shaped shunt current distribution in the manifold also with respect to the number of stacks for the charging and
presented asymmetrically, and the magnitude on the left was discharging processes were similar, as shown in Figures 8 and 9,
larger than that on the right. This is because the manifolds on the respectively, except for a slight magnitude difference. When the
left of each stack were connected to the piping system, resulting multistack system was charged at 54 A, the maximum shunt
4654 https://doi.org/10.1021/acssuschemeng.1c00287
ACS Sustainable Chem. Eng. 2021, 9, 4648−4659
ACS Sustainable Chemistry & Engineering pubs.acs.org/journal/ascecg Research Article

Figure 7. Shunt current distributions within the stacks during the discharge process for (a) two 60-cell stacks, (b) three 40-cell stacks, (c) four 30-cell
stacks, (d) five 24-cell stacks, (e) six 20-cell stacks, (f) eight 15-cell stacks, (g) twelve 10-cell stacks, and (h) twenty-four 5-cell stacks.

current in the piping system was located in the central manifold. shunt current in the negative channel connected to the second
This is because the manifold diameter is larger than the channel stack reduced to almost zero, as shown in Figure 8(c). More
diameter, resulting in a lower electrolyte resistance in the shunt currents in the channels of the central stacks reduced to
manifold between adjacent stacks. Also, the magnitudes of the zero as the number of stacks gradually increased. These zero
shunt currents in the positive flow channel that connects to the shunt currents in both the positive and negative channels of the
first stack and in the manifold that connects the first and the piping system became more significant when connecting more
second stacks were the same, except for an opposite sign, due to than five stacks in series, as shown in Figure 8(e−h). The shunt
the definition of the current flow direction and the special current distribution gradually transferred from the inside of the
position of the first flow channel. There were similar results in stack to the piping system as the number of stacks increased or as
the negative piping system that connects to the last stack. the number of cells in a single stack decreased. Because the
As the number of stacks connected in series increased, the voltage difference within a single stack gradually decreases with a
magnitudes of the shunt currents in the either the channel or the decreasing number of cells, there is a reduction in internal shunt
manifold decreased on both the positive and negative sides. currents. As a result, the electrolyte in the piping system
With the four stacks connected in series, it can be seen that the dominates the shunt current distribution. The extreme case is
4655 https://doi.org/10.1021/acssuschemeng.1c00287
ACS Sustainable Chem. Eng. 2021, 9, 4648−4659
ACS Sustainable Chemistry & Engineering pubs.acs.org/journal/ascecg Research Article

Figure 8. Shunt current distributions in the piping system during the charge process for (a) two 60-cell stacks, (b) three 40-cell stacks, (c) four 30-cell
stacks, (d) five 24-cell stacks, (e) six 20-cell stacks, (f) eight 15-cell stacks, (g) twelve 10-cell stacks, and (h) twenty-four 5-cell stacks.

only one cell in a single stack; in this case, the piping system ∑ (VcellIcell)
εcharge =
serves as the external flow path for the electrolytes, resulting in a VsysIsys (25)
shunt current in the piping system but not within each stack.
However, this extreme case would result in high material costs whereas the discharge efficiency is the ratio of the power
for the stack components and the piping system. A full cost generated from the battery system to the sum of the power
generated by each cell.
analysis is beyond the scope of this study.
Efficiency Analysis. As previously noted, system efficiency VsysIsys
εdischarge =
is a trade-off between shunt currents and pump power ∑ (VcellIcell) (26)
consumption. The system design is also highly customized on
The overall efficiency, determined by the multiplication of the
a case-by-case basis. As a result, we analyzed only the effect of the
charge efficiency and the discharge efficiency, indicates the
number of stacks on the charging efficiency in this study. The round-trip efficiency of the VRFB system. Table 2 shows the
charging efficiency is the ratio of the sum of the power charged overall efficiency of systems with various numbers of stacks. The
into each cell to the power charged into the battery system efficiency should increase with an increasing number of stacks, as
4656 https://doi.org/10.1021/acssuschemeng.1c00287
ACS Sustainable Chem. Eng. 2021, 9, 4648−4659
ACS Sustainable Chemistry & Engineering pubs.acs.org/journal/ascecg Research Article

Figure 9. Shunt current distributions in the piping system during the discharge process for (a) two 60-cell stacks, (b) three 40-cell stacks, (c) four 30-
cell stacks, (d) five 24-cell stacks, (e) six 20-cell stacks, (f) eight 15-cell stacks, (g) twelve 10-cell stacks, and (h) twenty-four 5-cell stacks.

Table 2. Comparison of Efficiencies of Systems with Various voltage decreases because of the reduced number of cells and the
Numbers of Stacks piping system dominates the distribution of electrolytes. As a
number of stacks charge efficiency discharge efficiency overall efficiency result, the shunt current loss decreases as the shunt currents
2 0.804 0.87 0.699 gradually shift from the inside of the stack to the piping system.
3 0.831 0.885 0.736 Although the charge and discharge efficiencies increase with
4 0.855 0.9 0.77 an increasing number of stacks, the optimal number of stacks in a
5 0.875 0.912 0.798 system cannot be identified until the specific practical
6 0.891 0.922 0.821 requirements of the system are determined. The cost of energy
8 0.914 0.938 0.857
storage during the whole service life of an energy storage system
12 0.942 0.957 0.901
24 0.976 0.982 0.959
is always the major factor for commercialization. The costs of
battery materials and auxiliary components depends on the
discussed in previous sections. If there is a constant number of volume and scale of energy storage systems; we will discuss this
cells in a system, when the number of stacks increases, the stack in a future study.
4657 https://doi.org/10.1021/acssuschemeng.1c00287
ACS Sustainable Chem. Eng. 2021, 9, 4648−4659
ACS Sustainable Chemistry & Engineering


pubs.acs.org/journal/ascecg Research Article

CONCLUSIONS Notes
The authors declare no competing financial interest.


Shunt current losses, arising from the potential differences
between cells within a single stack and between stacks in a series, ACKNOWLEDGMENTS
causes energy loss in a multistack VRFB system. We developed a The authors would like to acknowledge the support from the
mathematical model, based on Kirchhoff’s law, to investigate the Bureau of Energy (107-D0112), Ministry of Economic Affairs,
effect of the number of stacks on the shunt current distribution and the Advanced Institute of Manufacturing with High-tech
with a constant overall number of cells. We reached the Innovations (AIM-HI) from The Featured Areas Research
following conclusions. Center Program within the framework of the Higher Education
(1) In a VRFB system, the shunt currents in the central stacks Sprout Project by the Ministry of Education (MOE) in Taiwan
are larger than the currents in the other stacks. In addition, the for this study.
magnitudes of the shunt currents in the flow channels or
manifold decrease with an increasing number of stacks because
of the decreased voltage differences between cells and stacks.
■ NOMENCLATURE
A, matrix with precalculated resistances
(2) The shunt current distribution in the manifold of a single I, current
stack is asymmetrical because the shunt currents in the I, vector that consists of cell currents and shut currents
manifolds need to balance those in the piping system. K, cell number in a stack
(3) In this study, connecting more than five stacks in series L, manifold length of piping between stacks
reduced the magnitudes of the shunt currents in the channels of M, total cell number in the system
R, resistance
the piping system to almost zero.
SOC, state of charge
(4) Charge and discharge efficiencies increase with an
SYS, system
increasing number of stacks because of the reduced number of V, voltage
cells in a single stack. As a result, the piping system dominates V, vector that mainly consists of cell voltages
the distribution of the electrolytes, and the shunt currents Z, stack number
gradually shift from inside the stack to the piping system.

■ AUTHOR INFORMATION
■ SUBSCRIPTS AND SUPERSCRIPTS
cell, cell
Corresponding Author
i, the ith cell
j, the jth stack
Yong-Song Chen − Department of Mechanical Engineering and m, manifold
Advanced Institute of Manufacturing with High-tech nc, negative channel
Innovations, National Chung Cheng University, Minhsiung nm, negative manifold
Township, Chiayi County 62102, Taiwan; orcid.org/ P, piping
0000-0002-6182-6418; Email: imeysc@ccu.edu.tw pc, positive channel
Authors pm, positive manifold
S, stack
Han-Wen Chou − Institute of Nuclear Energy Research, Atomic
T, total


Energy Council, Longtan District, Taoyuan 32546, Taiwan
Feng-Zhi Chang − Department of Mechanical Engineering and REFERENCES
Advanced Institute of Manufacturing with High-tech
(1) Poullikkas, A. A comparative overview of large-scale battery
Innovations, National Chung Cheng University, Minhsiung
systems for electricity storage. Renewable Sustainable Energy Rev. 2013,
Township, Chiayi County 62102, Taiwan 27, 778−788.
Hwa-Jou Wei − Institute of Nuclear Energy Research, Atomic (2) Fabjan, C.; Garche, J.; Harrer, B.; Jörissen, L.; Kolbeck, C.;
Energy Council, Longtan District, Taoyuan 32546, Taiwan Philippi, F.; Tomazic, G.; Wagner, F. The vanadium redox-battery: an
Bhupendra Singh − Department of Mechanical Engineering efficient storage unit for photovoltaic systems. Electrochim. Acta 2001,
and Advanced Institute of Manufacturing with High-tech 47, 825−831.
Innovations, National Chung Cheng University, Minhsiung (3) Skyllas-Kazacos, M.; Kazacos, G.; Poon, G.; Verseema, H. Recent
Township, Chiayi County 62102, Taiwan; orcid.org/ advances with UNSW vanadium-based redox flow batteries. Int. J.
0000-0002-0773-5312 Energy Res. 2010, 34, 182−189.
Amornchai Arpornwichanop − Center of Excellence in Process (4) Díaz-González, F.; Sumper, A.; Gomis-Bellmunt, O.; Villafáfila-
Robles, R. A review of energy storage technologies for wind power
and Energy Systems Engineering, Department of Chemical applications. Renewable Sustainable Energy Rev. 2012, 16, 2154−2171.
Engineering, Faculty of Engineering, Chulalongkorn University, (5) Chang, C.-H.; Chou, H.-W.; Hsu, N.-Y.; Chen, Y.-S. Development
Pathum Wan District, Bangkok 10330, Thailand; of integrally molded bipolar plates for all-vanadium redox flow batteries.
orcid.org/0000-0001-9259-5010 Energies 2016, 9, 350.
Prathak Jienkulsawad − Center of Excellence in Process and (6) Skyllas-Kazacos, M.; McCann, J.; Li, Y.; Bao, J.; Tang, A. The
Energy Systems Engineering, Department of Chemical mechanism and modelling of shunt current in the vanadium redox flow
Engineering, Faculty of Engineering, Chulalongkorn University, battery. Chem. Select 2016, 1, 2249−2256.
Pathum Wan District, Bangkok 10330, Thailand (7) Xiong, B. Y.; Zhao, J. Y.; Tseng, K. J.; Skyllas-Kazacos, M.; Lim, T.
Yi-Sin Chou − Institute of Nuclear Energy Research, Atomic M.; Zhang, Y. Thermal hydraulic behavior and efficiency analysis of an
Energy Council, Longtan District, Taoyuan 32546, Taiwan all-vanadium redox flow battery. J. Power Sources 2013, 242, 314−324.
(8) Tang, A.; Bao, J.; Skyllas-Kazacos, M. Studies on pressure losses
Complete contact information is available at: and flow rate optimization in vanadium redox flow battery. J. Power
https://pubs.acs.org/10.1021/acssuschemeng.1c00287 Sources 2014, 248, 154−162.

4658 https://doi.org/10.1021/acssuschemeng.1c00287
ACS Sustainable Chem. Eng. 2021, 9, 4648−4659
ACS Sustainable Chemistry & Engineering pubs.acs.org/journal/ascecg Research Article

(9) Kaminski, E. A.; Savinell, R. F. A technique for calculating shunt


leakage and cell currents in bipolar stacks having divided or undivided
cells. J. Electrochem. Soc. 1983, 130, 1103−1107.
(10) Xing, F.; Zhang, H.; Ma, X. Shunt current loss of the vanadium
redox flow battery. J. Power Sources 2011, 196, 10753−10757.
(11) Moro, F.; Trovò, A.; Bortolin, S.; Del Col, D.; Guarnieri, M. An
alternative low-loss stack topology for vanadium redox flow battery:
Comparative assessment. J. Power Sources 2017, 340, 229−241.
(12) Trovò, A.; Marini, G.; Sutto, A.; Alotto, P.; Giomo, M.; Moro, F.;
Guarnieri, M. Standby thermal model of a vanadium redox flow battery
stack with crossover and shunt-current effects. Appl. Energy 2019, 240,
893−906.
(13) Tang, A.; McCann, J.; Bao, J.; Skyllas-Kazacos, M. Investigation
of the effect of shunt current on battery efficiency and stack temperature
in vanadium redox flow battery. J. Power Sources 2013, 242, 349−356.
(14) Fink, H.; Remy, M. Shunt currents in vanadium flow batteries:
Measurement, modelling and implications for efficiency. J. Power
Sources 2015, 284, 547−553.
(15) Yin, C.; Guo, S.; Fang, H.; Liu, J.; Li, Y.; Tang, H. Numerical and
experimental studies of stack shunt current for vanadium redox flow
battery. Appl. Energy 2015, 151, 237−248.
(16) Wandschneider, F. T.; Röhm, S.; Fischer, P.; Pinkwart, K.;
Tübke, J.; Nirschl, H. A multi-stack simulation of shunt currents in
vanadium redox flow batteries. J. Power Sources 2014, 261, 64−74.
(17) König, S.; Suriyah, M. R.; Leibfried, T. Model based examination
on influence of stack series connection and pipe diameters on efficiency
of vanadium redox flow batteries under consideration of shunt currents.
J. Power Sources 2015, 281, 272−284.
(18) Ye, Q.; Hu, J.; Cheng, P.; Ma, Z. Design trade-offs among shunt
current, pumping loss and compactness in the piping system of a multi-
stack vanadium flow battery. J. Power Sources 2015, 296, 352−364.
(19) Chen, H.; Li, X.; Gao, H.; Liu, J.; Yan, C.; Tang, A. Numerical
modelling and in-depth analysis of multi-stack vanadium flow battery
module incorporating transport delay. Appl. Energy 2019, 247, 13−23.
(20) Schreiber, M.; Harrer, M.; Whitehead, A.; Bucsich, H.;
Dragschitz, M.; Seifert, E.; Tymciw, P. Practical and commercial issues
in the design and manufacture of vanadium flow batteries. J. Power
Sources 2012, 206, 483−489.
(21) Weng, G.-M.; Li, C.-Y. V.; Chan, K.-Y.; Lee, C.-W.; Zhong, J.
Investigations of high voltage vanadium-metal hydride flow battery
toward kWh scale storage with 100 cm2 electrodes. J. Electrochem. Soc.
2016, 163, A5180−A5187.
(22) Chen, Y.-S.; Ho, S.-Y.; Chou, H.-W.; Wei, H.-J. Modeling the
effect of shunt current on the charge transfer efficiency of an all-
vanadium redox flow battery. J. Power Sources 2018, 390, 168−175.

4659 https://doi.org/10.1021/acssuschemeng.1c00287
ACS Sustainable Chem. Eng. 2021, 9, 4648−4659

You might also like