Journal of Power Sources 196 (2011) 10737–10747

Contents lists available at SciVerse ScienceDirect

Journal of Power Sources

journal homepage: www.elsevier.com/locate/jpowsour

Dynamic modelling of the effects of ion diffusion and side reactions on the
capacity loss for vanadium redox flow battery
Ao Tang, Jie Bao, Maria Skyllas-Kazacos ∗
School of Chemical Engineering, University of New South Wales, Sydney, NSW 2052, Australia

a r t i c l e i n f o a b s t r a c t

Article history: The diffusion of vanadium ions across the membrane along with side reactions can have a significant
Received 27 May 2011 impact on the capacity of the vanadium redox flow battery (VFB) over long-term charge–discharge
Received in revised form 30 August 2011 cycling. Differential rates of diffusion of the vanadium ions from one half-cell into the other will facilitate
Accepted 1 September 2011
self-discharge reactions, leading to an imbalance between the state-of-charge of the two half-cell elec-
Available online 7 September 2011
trolytes and a subsequent drop in capacity. Meanwhile side reactions as a result of evolution of hydrogen
or air oxidation of V2+ can further affect the capacity of the VFB. In this paper, a dynamic model is devel-
Keywords:
oped based on mass balances for each of the four vanadium ions in the VFB electrolytes in conjunction
Vanadium redox flow battery
Dynamic modelling
with the Nernst Equation. This model can predict the capacity as a function of time and thus can be used
Capacity loss to determine when periodic electrolyte remixing or rebalancing should take place to restore cell capac-
Ion diffusion ity. Furthermore, the dynamic model can be potentially incorporated in the control system of the VFB to
Side reactions achieve long term optimal operation. The performance of three different types of membranes is studied
on the basis of the above model and the simulation results together with potential operational issues are
analysed and discussed.
© 2011 Elsevier B.V. All rights reserved.

1. Introduction At positive electrode:

charge
The vanadium redox flow battery has been considered to be VO2+ + H2 O  VO+ + −
2 + 2H + e , E0 = 1 V (1)
one of the most promising large scale energy storage systems that discharge
can be combined with renewable energy sources such as solar and At negative electrode:
wind energy for electrical energy storage and distribution [1–4].
Compared with conventional rechargeable batteries, the VFB stores charge
V3+ + e−  V2+ , E 0 = −0.26 V (2)
energy in the form of two vanadium solutions contained in two sep- discharge
arate electrolyte reservoirs, thereby enabling the battery capacity
to be readily increased by simply adding more solution. In addi- One of the essential parts of the VFB is the membrane through
tion, the power output is determined by adjusting the number of which the charge is balanced by the movement of hydrogen ions
cells and the electrode area. As vanadium ions are employed in during charging and discharging. The ideal membrane should have
both half-cells, the issue of cross-contamination during long term low permeability to vanadium ions, high proton conductivity, good
use is eliminated, so that the electrolyte has an indefinite life. With chemical stability in acid solution, good mechanical integrity as
its excellent features and flexibility, a number of applications have well as low cost. To date many types of membrane have been suc-
been successfully field tested in different parts of the world in areas cessfully evaluated for use in the VFB [5–12] and some of these
such as peak shaving, load levelling, emergency back-up, as well as have been applied in both demonstrations and commercial sys-
renewable energy output smoothing. During operation of the VFB, tems [13–15], while further investigation and development of new
the electrolyte is pumped through pipes from the tanks into the membranes for cost reduction have been also reported [16–18].
cells where the following reactions occur. Nevertheless, as with all electrochemical energy storage systems,
self-discharge and side reactions will occur and reduce capacity
during long-term cycling.
In the case of the VFB, self-discharge reactions are mainly asso-
ciated with the diffusion of vanadium ions from one half-cell to
the other due to the concentration gradients of each of the ions
∗ Corresponding author. Tel.: +61 2 9385 4335; fax: +61 2 9385 5966. across the membrane. This will give rise to a reduction in the
E-mail address: [email protected] (M. Skyllas-Kazacos). coulombic efficiency, the magnitude of which will depend on the

0378-7753/$ – see front matter © 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.jpowsour.2011.09.003
10738 A. Tang et al. / Journal of Power Sources 196 (2011) 10737–10747

diffusion rates of the vanadium ions across the particular mem-

brane. Many researchers have already focused on the self-discharge
behaviour for different types of membranes in the VFB. Early stud-
ies by Skyllas-Kazacos and co-workers reported the properties of a
wide range of anion and cation exchange membranes in small VFB
single cells [5–15] and found that the magnitude and direction of
water and vanadium ion transfer depends on the state-of-charge
of the solutions and the properties of the membrane (including
ion exchange capacity and type of ion exchange groups–anionic vs.
cationic). The group proposed a number of membrane treatment
processes to reduce the magnitude of water and electrolyte trans-
fer in order to minimise operational problems and capacity loss
during long-term operation of the VFB. No membrane or treatment
process can, however, totally eliminate the transfer of vanadium
ions and water from one half-cell to the other, so some capacity
loss will always be observed during long-term cycling. Apart from
diffusion effects, side reactions such as evolution of hydrogen at
the negative electrode during charging and air oxidation of V2+ in
the negative electrolyte solution will also influence capacity loss
by disturbing the molar ratio of the charged to discharged species
in the two half-cells. All these factors exist in the VFB systems and
cause imbalance among the vanadium ions of different states of
oxidation, thereby leading to a deteriorative capacity performance
during long term use. A dynamic model that can be used to simu-
Fig. 1. Performance of a small static cell employing perfluorinated cation exchange
late these phenomena and predict their effect on battery capacity membrane. (a) Voltage and current vs. time during charge–discharge cycling. (b)
would be invaluable in the development of sophisticated control Capacity vs. cycle number.
systems for the VFB.
In recent years, mathematical models of the VFB have been however, the crossover of vanadium ions through the membrane
developed to predict and investigate the performance of the VFB cannot be entirely avoided due to the fact that diffusion always
systems [19–25]. Walsh and co-workers [19–22] developed a tran- occurs when there is the concentration difference of vanadium ions
sient two-dimensional model based on conservation principles existing between the positive and negative half-cell electrolytes.
combined with a global kinetic model for reactions to simulate the This gives rise to self-discharge reactions in each electrolyte and
performance of the VFB with respect to variations in parameters. a subsequent loss of capacity. With regard to ion diffusion, all the
It was further extended to include effects of temperature variation vanadium ions in four different states of oxidation undergo dif-
and several side reactions. Li and Hikihara [23] proposed a simple fusion across the membrane and react with each other causing
dynamic model suitable for control purpose on the basis of physical reactions shown as follows.
phenomena and chemical kinetics. To date, no dynamic models of In the positive half-cell, diffusion of V2+ and V3+ ions from the
the VFB, however, have been developed to describe capacity loss negative half-cell will lead to the following self-discharge reac-
caused by vanadium ion diffusion and side reactions. In this paper, tions:
a dynamic model based on the conservation of mass is developed
to study the charge–discharge cycling capacity loss caused by the V2+ + 2V5+ → 3V4+ (3)
diffusion of vanadium ions across the membrane as well as side 3+ 5+ 4+
V +V → 2V (4)
reactions. Such a model will be useful for developing battery control
2+ 4+ 3+
systems that can automatically restore the capacity of the VFB by V +V → 2V (5)
rebalancing the electrolytes after long term operation. Moreover,
the Nernst Equation is used in conjunction with the mass balance while in the negative half-cell, V5+ and V4+ ions from the positive
model to produce a dynamic model of the cell voltage that can pre- side will react as follows:
dict the capacity of the cell for long term charge–discharge cycling V5+ + 2V2+ → 3V3+ (6)
under specific operating limits of cell voltage. The prediction of
4+ 2+ 3+
capacity can help with making decisions for the implementation of V +V → 2V (7)
periodic electrolyte rebalancing to equalize the vanadium ion con- 5+ 3+ 4+
V +V → 2V (8)
centrations in both electrolyte tanks and thereby restore capacity
during operation. Simulations have been performed to investigate According to the above self-discharge reactions, an imbalance
the capacity loss for different types of membranes and different in vanadium ions will occur in the two half-cells if the diffusion
levels of hydrogen evolution and air oxidation during long term rates of the different vanadium ions are not equal. This would lead
charge–discharge cycling operation. to a build up of vanadium ions in one half-cell and a correspond-
ing decrease in the other, thus giving rise to loss of capacity. This
phenomenon is demonstrated in Fig. 1a that shows the laboratory
2. Effects of diffusion and side reactions cycling test charging–discharging current and cell voltage curves
generated from a small sealed static cell employing a perfluorinated
The membrane in the VFB plays an important role in separating cation exchange membrane. A small current density of 20 mA cm−2
the positive and negative half-cell electrolytes containing pentava- was employed in order to minimise the evolution of hydrogen dur-
lent/tetravalent and bivalent/trivalent vanadium ions respectively ing charging, while the sealed cell prevented any air ingress that
within the cell. It allows hydrogen ions to go through the mem- would cause oxidation of V2+ . A gradual decrease of capacity can be
brane in order to balance the charge while preventing the transfer observed in 40 charge–discharge cycles in Fig. 1b, which is mainly
of vanadium ions from one half-cell to the other. In reality, attributed to the differential rates of diffusion of the vanadium ions
 A. Tang et al. / Journal of Power Sources 196 (2011) 10737–10747 10739

from one half-cell to the other. Such capacity loss can be read-
ily restored by simply remixing the two electrolytes in order to
rebalance the vanadium concentrations in each half-cell.
The above effects of vanadium ion diffusion across the mem-
brane can be established by means of Fick’s first law shown below
which describes that the flux of diffusion is from region of high
concentration to region of low concentration with a magnitude
proportional to the concentration gradient.
dc
J = −k (9)
dx
where J is the flux of diffusion (mol L−1 s−2 ), k is the diffusion
coefficient (ms−1 ), dc is the concentration difference across the
membrane (mol L−1 ) and dx is the diffusion layer thickness which
is assumed to be equal to the thickness of membrane (m).
During the process of charging, however, gassing side reactions
can occur at both electrodes, giving rise to a loss in coulombic effi-
ciency. While oxygen evolution at the positive electrode has been
shown to be negligible up to relatively high states-of-charge (SOC),
hydrogen evolution at the negative electrode can be quite signifi-
cant above 90% SOC. For this reason, the practical operating range
of the VFB is usually set between 10% and 90% SOC. Any difference
between the rates of the two gassing reactions will, however, give
rise to an imbalance between the SOCs of the two electrolytes that
can in turn have a remarkable impact on cell capacity. For example,
if 1% of the charging current is consumed by hydrogen evolution at
the negative electrode, this will lead to a 1% loss of capacity at each
cycle and 100% capacity loss after only 100 cycles.
As mentioned above, a second side reaction that will lead to
further capacity loss is air oxidation of bivalent vanadium ion in
the negative half-cell electrolyte. This will also lead to an imbal-
ance between the V2+ and V5+ ions in the two half-cells thus reduce
capacity. The effect of these processes on cell capacity is illustrated
in Fig. 2, which shows the charge–discharge voltage profiles for
a small laboratory flow cell with open electrolyte reservoirs was
cycled at a relatively larger current density of 40 mA cm−2 . From
Fig. 2b it can be seen that the capacity decreased from 800 to
120 mAh over only 20 cycles, considerably more than the capac-
ity loss observed in the sealed static cell results of Fig. 1. This high Fig. 2. Performance of a flow cell with perfluorinated cation exchange membrane.
capacity loss can be mainly attributed to the high rate of air oxida- (a) Voltage and current curves. (b) Capacity vs. cycle number.
tion of the V2+ ions in the open reservoir with the large electrolyte
surface area to volume ratio used. In practice, this can be minimised
by employing an inert gas blanket in the negative electrolyte tank control. Steady-state models represent the static behaviour of the
to exclude oxygen and/or by designing the electrolyte reservoirs process. In most real situations, however, conditions change with
with a very small surface area to volume ratio to make the air oxi- respect to time. Under these circumstances, a dynamic model is
dation process diffusion limited. The other side reaction leading to required to characterize the transient behaviours of a process over
the high loss of capacity is hydrogen evolution during charging. This a wide range of conditions. Generally, dynamic modelling can be
lowers the conversion of vanadium ions at negative electrode sur- developed based on the fundamental mechanisms of processes,
face particularly at high states-of-charge and with large currents. such as conservations of mass and energy. Meanwhile, assump-
In practice, action can be taken to prevent evolution of hydrogen by tions need to be made in order to derive process models with
ceasing the charging process at a reasonable state-of-charge limit, suitable level of details for their purpose by omitting less important
although the usable capacity of the VFB would be restricted by such phenomena.
operations. As mentioned above, membrane diffusion and side reactions
Unlike the capacity loss caused by the different transfer of vana- lead to loss of capacity in charge–discharge cycling operation by
dium ions across the membrane, the above side reactions cause creating an imbalance in vanadium ions and vanadium ion ratios
an electrolyte imbalance that cannot be restored by simple elec- in the two half-cells. Thus, readjustment of the electrolyte chem-
trolyte remixing but requires either chemical or electrochemical istry is required to be implemented periodically so as to recover
rebalancing to restore capacity loss. An effective control system for capacity and achieve desirable performance over many thousands
cell rebalancing requires a dynamic model that is able to predict the of charge–discharge cycles. A dynamical model is able to help
capacity loss due to the above phenomena so that the controller can with capacity loss predictions in order to take any necessary elec-
determine when such rebalancing processes is to be scheduled. trolyte maintenance. In the development of this dynamic model for
the vanadium redox flow battery, the following assumptions were
made for the purpose of simplifying the model:
3. Dynamic model development

Mathematical modelling helps with understanding processes (1) Temperature is assumed constant at room temperature;
and plays an important role in process design, optimization and (2) Electrolyte flow rate is not taken into account;
10740 A. Tang et al. / Journal of Power Sources 196 (2011) 10737–10747

(3) Side reactions caused by diffusion of ions across the membrane Table 1
Values of diffusion coefficients for vanadium ions across three different membranes.
are instantaneous;
(4) The volumes of both electrolytes remain constant; Membrane k2 (dms−1 ) k3 (dms−1 ) k4 (dms−1 ) k5 (dms−1 )
(5) The rates of the cell charge–discharge reactions can be Selemion CMV 3.17 × 10 −7
0.716 × 10−7
2 × 10 −7
1.25 × 10−7
described by Faraday’s law of electrolysis; Selemion AMV 3.53 × 10−8 2.18 × 10−8 0.91 × 10−8 2.57 × 10−8
(6) Side reactions (5) and (8) are negligible; Nafion 115 6.9 × 10−7 2.54 × 10−7 5.37 × 10−7 4.64 × 10−7
(7) The electrolytes are perfectly mixed;
(8) The proton concentration in each half-cell electrolyte remains
constant during charge–discharge cycling. cell voltage balance given by Eq. (12) was therefore used in the
development of the dynamic cell voltage model in this study:
RT

c2 (t) × c5 (t)

Based on the above assumptions and molar mass balance,
E(t) = E  0 + ln (12)
dynamic equations associated with the diffusion terms for all vana- zF c3 (t) × c4 (t)
dium ions are developed as shown in Eq. (10). Here, diffusion terms
where E0 is the formal cell potential. For 2 M vanadium in 5 M
for side reactions (5) and (8) are neglected because these two reac-
total sulfate electrolyte, the formal potential was measured as 1.4 V
tions produce V3+ and V4+ respectively which would further react
at 50% state-of-charge [26].
with V5+ and V2+ respectively, thus giving the same net reactions as
given by reactions (3) and (6). Substituting the resultant concentra-
4. Simulation and results
tion terms into the Nernst Equation and adding the cell resistance
losses, the cell voltage balance in Eq. (11) can be obtained.
For the simulation carried out in this paper, the following
dc2 iA c2 c5 c4 parameters were used in the modelling:
Vs =± − k2 S − 2k5 S − k4 S
dt zF d d d
dc3 iA c3 c5 c4 (i) Upper and lower cell voltage limits were set at 1.7 V and 1.1 V
Vs =± − k3 S + 3k5 S + 2k4 S for the end of charge and end of discharge cycles respectively.
dt zF d d d
(10) (ii) Membrane area and electrode surface area were with the same
dc4 iA c4 c2 c3
Vs =± − k4 S + 3k2 S + 2k3 S value of 30 dm2 .
dt zF d d d
(iii) The volume of each half-cell electrolyte solution was 1 L.
dc5 iA c5 c2 c3
Vs =± − k5 S − 2k2 S − k3 S (iv) The cell resistance was set at 2  cm2 .
dt zF d d d
(v) Temperature was set to be 298.15 K.
Nernst Equation: (vi) The total amount of vanadium ions used in the simulation were
 2 (t)
 assumed to be 2 mol L−1 in both half-cells, while 0.01 mol L−1
0 RT c2 (t) × c5 (t) × cH + of V2+ and V5+ in conjunction with 1.99 mol L−1 of V3+ and V4+
E(t) = E + ln (11)
zF c3 (t) × c4 (t) were set for the initial values of the dynamic model to avoid a
zero logarithmic term in the Nernst Equation.
where
Furthermore, 1.4 V was employed as the formal potential which
c2 = concentration of V2+ ions, mol L−1 is the actual open-circuit potential measured at 50% SOC with 2 M
c3 = concentration of V3+ ions, mol L−1 vanadium in 5 M total sulfate [26]. The simulations were conducted
c4 = concentration of V4+ ions, mol L−1 by switching the charging and discharging processes between the
c5 = concentration of V5+ ions, mol L−1 upper and lower cell voltage limits. To investigate different types
cH + = concentration of hydrogen ions, mol L−1 of membrane with respect to the capacity loss, a current density
k2 = diffusion coefficient for V2+ across the membrane, dm2 s−1 of 60 mA cm−2 was used which is equivalent to 180 A in this case.
k3 = diffusion coefficient for V3+ across the membrane, dm2 s−1 Three different membranes were employed in the simulation with
k4 = diffusion coefficient for V4+ across the membrane, dm2 s−1 the diffusion coefficients summarized in Table 1 [5,25]. The val-
k5 = diffusion coefficient for V5+ across the membrane, dm2 s−1 ues of the Nafion 115 diffusion coefficients reported in [25] were
VS = volume of half-cell solution, L converted from units of cm2 min−1 to dms−1 by multiplying for
i = current density, A cm−2 the membrane’s thickness, 127 microns and by the corresponding
A = electrode surface area, cm2 conversion factor to obtain units of dms−1 as summarised in Table 1.
z = number of electrons transferred in reaction (1)
F = Faraday’s constant = 96485 C mol−1 4.1. Case 1: Vanadium ion diffusion with no side reactions
d = thickness of membrane, dm
S = membrane area, dm2 4.1.1. Selemion CMV membrane
E0 = standard cell potential, V The vanadium ion diffusion coefficients for Selemion CMV
T = temperature, K cation-exchange membrane were firstly employed in the simu-
R = gas constant = 8.314 Jmol−1 K−1 lation. In Fig. 3 it shows the capacity change for the VFB as a
r = resistance,  cm2 function of cycle number for 200 charge–discharge cycles. As seen,
‘+’ and ‘−’ apply to the process of charging and discharging respec- the capacity drops 15% after 200 cycles. The corresponding dis-
tively charging time variation can be observed in Fig. 4 in which the
initial charge–discharge curves are compared with those for cycles
Due to several ionic equilibria that affect the equilibrium con- 97–100 and 197–200. This gradual decrease in discharge time with
centrations of all of the ionic species in the in the VFB electrolyte, the continued cycling is due to the differential transfer of vanadium
concentration of hydrogen ions in the positive half-cell electrolyte ions across the membrane that produces an imbalance in the vana-
is not accurately known. For this reason, the proton concentration dium ion concentration in the two half-cells. This can be seen in
is removed from the logarithmic term in the Nernst Equations and Fig. 5 where a build-up in total vanadium concentration in the pos-
combined with the standard potential to produce the formal poten- itive half-cell electrolyte along with the corresponding loss in the
tial term E0 that can be determined experimentally. The modified negative electrolyte is shown.
 A. Tang et al. / Journal of Power Sources 196 (2011) 10737–10747 10741

Capacity vs. No. of cycles for CMV Concentrations of vanadium ions at 1.7V
50 2

Concentration / mol L−1

1.5 V2
49
V3
1 V4
48 V5
After 200 cycles, the capacity is 85.5% 0.5
47
Capacity / Ah

0
0 20 40 60 80 100 120 140 160 180 200
No. of cycles
46
Concentrations of vanadium ions at 1.1V
2.5

Concentration / mol L−1

45
2

44 1.5 V2
V3
1
V4
43 V5
0.5

42 0
0 20 40 60 80 100 120 140 160 180 200 0 20 40 60 80 100 120 140 160 180 200
No. of cycles
No. of cycles
Fig. 6. Vanadium ion concentrations vs. cycle number at 1.7 V and 1.1 V for CMV
Fig. 3. Capacity vs. No. of cycles for 200 cycles for CMV membrane. membrane.

Cell voltage curves for different cycles

1.7 Further concentration profiles as a function of cycle number at
the upper and lower voltage limits (corresponding to top of charge
and end of discharge respectively) can be observed in Fig. 6. This
1.6 shows that the concentration of V4+ increases while the other three
vanadium ions experience a slight decrease over time. This can
be attributed to the order of diffusion coefficient values which is
1.5
V2+ > V4+ > V5+ > V3+ such that given a certain amount of V2+ going
Cell voltage / V

through the membrane, more V5+ would be consumed according to

1.4 the reaction shown in Eq. (3), producing even more V4+ in the pos-
itive electrolyte. Similarly, as V5+ diffuses across the membrane, it
reacts with V2+ ions, producing V3+ according to the self-discharge
1.3
reaction given in Eq. (6). In addition, it is seen from these equations
that no V2+ and V5+ can be produced by the self-discharge side reac-
1.2
tions. Thus, the concentrations of V2+ and V5+ show a decrease with
the number of cycles at the upper limit of voltage in Fig. 6. Finally,
a decreasing trend in concentration change with respect to V2+ is
1.1
2 3 4 5 ~~ 97 98 99 100 ~~ 197 198 199 200
observed by comparing the two plots in Fig. 6, which is in accor-
dance with the capacity loss shown in Fig. 3. It is interesting to note
No. of cycles
that a rapid decrease in both the ion concentrations and cell capac-
Fig. 4. Comparison of cell voltage curves for different cycles for CMV membrane.
ity occurs in the first 100 cycles, but these changes slow down and
approach steady state at around 200 cycles.
Concentraions vs. time in positive and negative electrolytes
2.2
4.1.2. Selemion AMV membrane
The anion exchange membrane Selemion AMV was also used in
2.15
V2+V3
the simulation for comparison. Compared with the cation exchange
V4+V5 membrane Selemion CMV, the diffusion coefficients for all except
2.1 the V3+ ions are around an order of magnitude smaller, while the
order of the diffusion rates is V2+ > V5+ > V3+ > V4+ . This difference in
Concentraion / mol L−1

2.05 the diffusion rates compared with CMV can be explained in terms of
the fact that vanadium cations are repelled by the anion exchange
2.0 membrane, giving rise to much smaller diffusion coefficients.
Fig. 7 shows that the capacity drops about 19% after 200 cycles
1.95 resulting in a decrease trend of charge–discharge cycling time
observed in the cell voltage versus time curve. This is also due
1.9 to the diffusion effect within the cells that affects the long term
cycling performance. Interestingly, the capacity with this mem-
1.85 brane has dropped more than that of CMV shown above after 200
charge–discharge cycles in spite of the smaller diffusion rates for
1.8 AMV. Comparing Fig. 7 with Fig. 3, it is seen that the initial rate
0 20 40 60 80 100 of cell capacity decline is much smaller in the case of the AMV
Time / h
membrane, consistent with the lower diffusion coefficients com-
Fig. 5. Concentrations of positive and negative electrolytes vs. time for 200 cycles pared with CMV. On the other hand, comparing Fig. 8 with Fig. 6 it
for CMV membrane. is seen that despite the relatively high initial rate of capacity loss
10742 A. Tang et al. / Journal of Power Sources 196 (2011) 10737–10747

Capacity vs. No of cycles for AMV Concentration vs. time in positive and negative electrolytes
52 2.25
50
After 200 cycles, the capacity is 81.2% 2.20
Capacity / Ah

48

46 2.15

44 V2+V3
2.10 V4+V5

Concentration / mol L−1

42

40 2.05
0 20 40 60 80 100 120 140 160 180 200
No. of cycles
2.00
Cell voltage vs. No. of cycles for AMV
1.95
1.6
Cell voltage / V

1.5 1.90

1.4 1.85
1.3
1.80
1.2

1.1 1.75
2 3 4 5 ~~ 97 98 99 100 ~~ 197 198 199 200 0 20 40 60 80 100
No. of cycles Time / h

Fig. 7. Capacity and cell voltage vs. No. of cycles for 200 cycles for AMV membrane. Fig. 9. Concentration of positive and negative electrolytes vs. time for 200 cycles
for AMV membrane.

with the CMV membrane, steady state is achieved after within 200 concentration is observed in the positive half-cell with a corre-
cycles, resulting in a relatively small capacity loss compared with sponding dilution in the negative half of the vanadium redox flow
AMV where the concentrations of the vanadium ions continue to cell employing the AMV anion exchange membrane.
change without reaching steady state. Therefore, the differential
transfer of vanadium ions across the AMV membrane continues to 4.1.3. Nafion 115 membrane
produce an imbalance in the total vanadium ion concentrations in The diffusion coefficients of vanadium ions across Nafion
the cell as shown in Fig. 9, leading to a continuing decline in the 115 membrane have been measured and reported in [25] with
cell capacity. As a matter of fact, in the absence of other factors a decreasing order of V2+ > V4+ > V5+ > V3+ . As shown in Fig. 10
such as side reactions, the vanadium ion concentrations along with the capacity is 89.4% after 200 charge–discharge cycles. It is
the capacity should reach a steady state after a certain number of worth noting that the main capacity loss has occurred in the first
charge–discharge cycles, although this is not observed within the 100 charge–discharge cycles. This is due to the larger diffusion
200 cycles simulated here. An important observation from these coefficients of Nafion 115 membrane that contribute to a faster
simulations, however, has been that not only are the absolute val- convergence of vanadium ion concentrations at upper and lower
ues of the different diffusion coefficients important in determining voltage limits as observed in Fig. 11. Additionally, it is also worth
the rate of capacity decline, but in the absence of other factors, the noting from Fig. 11 that the concentration of V2+ is less than
relative values for the four different vanadium ions will also play that of V5+ at upper limit of 1.7 V while V4+ is less than V3+ at
an important role in the attainment of steady state in cell capac- lower limit of 1.1 V which differs from the trend observed for
ity. It is also interesting to note that as in the case of the CMV CMV membrane although the order of diffusion rates are same.
cation exchange membrane, with the values of the four vanadium Therefore, it is reasonably concluded that different relative values
ion diffusion rates used here, a steady build-up of total vanadium of diffusion coefficients will result in different half-cell vanadium

Capacity vs. No. of cycles for Nafion 115 membrane

Concentration vs. No. of cycles at 1.7V 49
2
Concentration / mol L 1

48
−

After 200 cycles, the capacity is 89.4%

Capacity / Ah

1.5 V2 47
V3
46
1 V4
V5 45
0.5 44

43
0 0 20 40 60 80 100 120 140 160 180 200
0 20 40 60 80 100 120 140 160 180 200
No. of cycles
No. of cycles
Cell voltage vs. No. of cycles for Nafion 115 membrane
Concentration vs. No. of cycles at 1.1V
2.5
Concentration / mol L 1

1.6
−

2
Cell voltage / V

1.5
1.5 V2
1.4
V3
1 V4 1.3
V5 1.2
0.5

0 1.1
2 3 4 5 ~~ 97 98 99 100 ~~ 197 198 199 200
0 20 40 60 80 100 120 140 160 180 200
No. of cycles No. of cycles

Fig. 8. Vanadium ion concentrations vs. No. of cycles at 1.7 V and 1.1 V for 200 cycles Fig. 10. Capacity and cell voltage vs. No. of cycles for 200 cycles for Nafion 115
for AMV membrane. membrane.
 A. Tang et al. / Journal of Power Sources 196 (2011) 10737–10747 10743

Concentration vs. No. of cycles at 1.7V Capacity vs. No. of cycles for 100 cycles
2 50
Concentration / mol L−1

40 After 100 cycles, the capacity is 11.1%

1.5

Capacity / Ah
V2
V3 30
1 V4
20
V5
0.5
10

0 0
0 20 40 60 80 100 120 140 160 180 200 0 10 20 30 40 50 60 70 80 90 100
No. of cycles No. of cycles
Concentration vs. No. of cycles at 1.1V Cell voltage vs. No. of cycles for 100 cycles
2.5
Concentration / mol L−1

2 1.6

Cell voltage / V
1.5
1.5 V2
V3 1.4
1
V4 1.3
0.5 V5
1.2
0 1.1
0 20 40 60 80 100 120 140 160 180 200 2 3 4 5 ~~ 47 48 49 50 ~~ 97−100
No. of cycles
No. of cycles

Fig. 11. Vanadium ion concentrations vs. No. of cycles at 1.7 V and 1.1 V for 200
Fig. 13. Capacity and cell voltage vs. cycle number for CMV membrane with 0.98
cycles for Nafion 115 membrane.
current efficiency factor for V2+ charging reaction.

accumulations in the simulation despite the same order in relative

capacity was simulated here by inserting a current efficiency factor
diffusion rates. Finally, referring to Figs. 5 and 12, there is an
into the charging current for the V2+ and V3+ mass balance equa-
accumulation of vanadium in the negative half-cell for the Nafion
tions in the dynamic model. The Selemion CMV and Nafion 115
115 membrane, opposite to the observed trend for CMV. In the
membranes were employed in the simulation to investigate the
case of Nafion, the total vanadium concentration in the negative
effects on performance.
half-cell solution shows an increase of approximately 7% after 200
charge–discharge cycles and the concentrations of each ion in the
two electrolytes tend to reach steady-state within this time-frame. 4.2.1. Selemion CMV membrane
Once again, it is apparent that not only are the magnitudes and Initially, a factor of 0.98 was used to simulate the side reac-
order of the four diffusion coefficients important in determining tion effect on capacity for the CMV membrane. Fig. 13 shows the
the rate of capacity loss and attainment of steady state, but in the capacity has dropped to 11% of the original capacity after only
absence of other factors, the relative values will also determine 100 charge–discharge cycles because of the side reactions in the
the direction of vanadium accumulation in the cell. negative half-cell. This is further highlighted by the corresponding
variation in the charge–discharge voltage curves that also reflect
4.2. Case 2: effect of side reactions on capacity loss the dramatic capacity loss during the first 100 cycles. Correspond-
ingly, a continued decrease in V2+ concentration during cycling can
Side reactions inevitably occur in all electrochemical energy be observed in Fig. 14, showing that a 2% current efficiency loss
storage systems, especially during the process of charging, thereby caused by side reaction will lead to a capacity loss of approximately
leading to further capacity loss during operation of the VFB. The 89% after 100 cycles. For comparison, a current efficiency factor of
effect of air oxidation of V2+ and hydrogen evolution on cell 0.9 was also employed in the simulation. As shown in Fig. 15, the

Concentration vs. time for 200 cycles Concentration vs. No. of cycles at 1.7V
2.2 2.5
Concentration / mol L−1

2
2.15
1.5 V2
V3
2.1 1 V4
V5
0.5
Concentration / mol L−1

2.05 V2+V3
V4+V5 0
0 20 40 60 80 100
No. of cycles
2.0
Concentration vs. No. of cycles at 1.1V
2.5
Concentration / mol L−1

1.95
2

1.9 1.5 V2
V3
1 V4
1.85 V5
0.5

1.8 0
0 20 40 60 80 100 0 20 40 60 80 100
Time / h No. of cycles

Fig. 12. Concentrations of positive and negative electrolytes vs. time for 200 cycles Fig. 14. Vanadium ion concentrations vs. cycle number at 1.7 V and 1.1 V for CMV
for Nafion 115 membrane. membrane with 0.98 current efficiency factor for V2+ charging reaction.
10744 A. Tang et al. / Journal of Power Sources 196 (2011) 10737–10747

Capacity vs. No. of cycles for 20 cycles Capacity vs. No. of cycles for 40 cycles
50 50

40
After 20 cycles, the capacity is 13% 45
Capacity / Ah

30
40
20 After 40 cycles, the capacity is 12.5%
10 35

Capacity / Ah
0 2 4 6 8 10 12 14 16 18 20 30
No. of cycles
Concentration vs. time for 20 cycles
2.1 25
Concentration / mol L 1
−

20
1.4 V2
V3
15
V4
0.7 V5
10

0
0 1 2 3 4 5 5
0 5 10 15 20 25 30 35 40
Time / h
No. of cycles

Fig. 15. Capacity and concentration curves for 20 cycles for CMV membrane with
Fig. 17. Capacity vs. cycle number for 40 cycles for CMV membrane with 0.95 current
0.9 current efficiency factor for V2+ charging reaction.
efficiency factor for V2+ charging reaction.

capacity loss is approximately 87% after only 20 charge–discharge the effect of gas side reaction in Nafion 115 membrane.
cycles. Fig. 16 further shows the variations of voltage and current Fig. 18 shows that the capacity has dropped dramatically for
curves for 20 cycles which has a trend similar to the experimental 40 charging–discharging cycles which can be observed from the
results shown in Fig. 2a. Fig. 2b shows a capacity loss of approxi- decreasing charging–discharging time in Fig. 19 as well. Further-
mately 86% after 20 cycles for the small laboratory flow cell. As both more, referring to Fig. 20 V2+ has experienced a continuous decrease
experiment and simulation employed cation exchange membrane as a function of cycle number at the upper voltage limit due to
and results were extremely similar, it is reasonable to conclude that the loss of current efficiency caused by the gas side reaction that
side reactions caused by the open electrolyte reservoirs and large would lower the conversion of V2+ during charging. Correspond-
current density led to 10% loss of current efficiency in the nega- ingly, V5+ will increase as a function of cycle number in discharging
tive half-cell for the laboratory flow cell. It is also seen in Fig. 15 process which can be observed at the lower voltage limit in Fig. 20.
that the concentration of V2+ at the end of 20 cycles is five times Finally it is seen from Fig. 21 that rather than reaching a relative
smaller than that observed in Fig. 14 at the 20th cycle. Therefore, steady state as shown with 100% current efficiency in the previ-
it is reasonable to predict that 40 charge–discharge cycles might ous case study, the vanadium ions have kept accumulating in the
be required for the same amount of capacity loss with a current negative half-cell electrolyte while decreasing in the positive elec-
efficiency factor of 0.95. This is in agreement with the simulation trolyte accordingly as a consequence of the gas side reactions. It is
result given by Fig. 17. worth noting that this predicted trend does not match the results
from a recent experimental study of a kilo-watt scale VFB system
4.2.2. Nafion 115 membrane with Nafion 115 membrane in which the accumulation of vanadium
To start with, a charging current efficiency of 98% for the ions was seen to occur in the positive half-cell electrolyte [25]. It
negative electrode reaction was used in effort to simulate should be noted, however, that a low SOC range was employed

Capacity vs. No. of cycles for Nafion 115 membrane

Cell voltage vs. time for 20 cycles 50

1.6
Cell voltage / V

45
1.5
After 40 cycles, the capacity is 33.5%
1.4

1.3 40

1.2
Capacity / Ah

1.1 35
0 1 2 3 4 5
Time / h
Current vs. time for 20 cycles 30

180
120 25
Current / A

60
0
−60 20
−120
−180
15
0 1 2 3 4 5 0 5 10 15 20 25 30 35 40
Time / h No. of cycles

Fig. 16. Cell voltage and current vs. time for 20 cycles for CMV membrane with 0.9 Fig. 18. Capacity vs. cycle number for 40 cycles for Nafion 115 membrane with 0.98
current efficiency factor for V2+ charging reaction. current efficiency factor for V2+ charging reaction.
 A. Tang et al. / Journal of Power Sources 196 (2011) 10737–10747 10745

Cell voltage vs. time for Nafion 115 membrane Concentrations vs. time for Nafion 115
1.7 2.2
Cell voltage / V

1.5 2.15

2.1
1.3

Concentration / mol L−1

V2+V3
2.05 V4+V5
1.1
0 2 4 6 8 10 12 14
Time / h 2.0
Current vs. time for Nafion 115 membrane

180 1.95
120
Current / A

60 1.9
0
−60
−120 1.85
−180
1.8
0 2 4 6 8 10 12 14 0 2 4 6 8 10 12 14
Time / h Time (hours)

Fig. 19. Cell voltage and current vs. time for 40 cycles for Nafion 115 membrane Fig. 21. Concentrations of positive and negative electrolytes vs. time for 40 cycles for
with 0.98 current efficiency factor for V2+ charging reaction. Nafion 115 membrane with 0.98 current efficiency factor for V2+ charging reaction.

in [25] such that side reactions would be minimised during charg- it is still worth to investigate the potential factors that restrict the
ing, while considerable volumetric transfer was also observed from accuracy of the model and further improve performance prediction.
the negative into the positive half-cell electrolyte reservoirs. This
bulk electrolyte transfer has been attributed to water transfer asso- 5.1. The effect of diffusion coefficients
ciated with osmotic pressure effects across the membrane that
also drag bulk electrolyte from the negative to the positive half- First of all, considerable errors can arise from the experimental
cell [13,14,25]. Furthermore, the difference in conditions between procedures used by different groups to measure diffusion coef-
the present simulation results and the experimental study in [25] ficients of the different vanadium ions, and these will have a
makes direct comparison difficult. What is clear, however, is that considerable influence on the accuracy of any prediction. In par-
volumetric transfer is very significant in the VFB and this phe- ticular, V2+ ions are always difficult to study because of their
nomenon will greatly affect the net accumulation of vanadium extreme sensitivity to air which converts V2+ to V3+ making accu-
ions in the VFB system. A more accurate dynamic model should rate measurements very difficult. In reality, membrane aging will
therefore include the effect of variable electrolyte volume during also change membrane properties, leading to higher or lower per-
charge–discharge cycling of the VFB. The dynamic model which meability of the vanadium ions depending on degree of swelling or
accounts the volumetric transfer is currently under investigation. fouling during operation. As a result, the initial diffusion rate mea-
surements from static diffusion tests will not accurately describe
5. Discussion the real ion diffusion effects on self-discharge and ion accumulation
during long-term cycling and operation. In addition, the diffusion
Despite the ability to use the dynamic model for prediction of rates also vary with temperature. As the cells are being cycled, the
loss of capacity under specified operating conditions in the VFB, temperature of the electrolytes will increase as a result of the reac-
tions and ohmic losses in the cells. This will gives rise to the swelling
Concentration vs. No. of cycles at 1.7V for membrane and further increase the diffusion rate of all the vana-
2
dium ions or even change the order of diffusion rates recorded in
Concentration / mol L 1
−

the experiment. As a matter of fact, extra simulations have shown

1.5
V2 that incorporating different relative values of diffusion coefficients
V3
1 into the simulation could possibly result in different trends even
V4
V5
when the same order of the diffusion rates is maintained. Thus, a
0.5 dynamic diffusion rate based on battery cycling experiments would
be required in order to simulate a more realistic situation in terms
0
0 10 20 30 40 of the ion diffusion within the cells. Furthermore, additional fac-
No. of cycles tors such as electric field and co-existence of different ions during
Concentrations vs. No. of cycles at 1.1V charge–discharge cycling also affect the accuracy of the measured
2.5
diffusion coefficients and highlight the need for further experimen-
Concentration / mol L 1
−

2 tal measurement under conditions that simulate real cell operation.

V2
1.5
V3
5.2. The effect of flow rate and volumetric transfer
1 V4
V5
0.5 Flow rate plays an important role in real-life systems since insuf-
ficient flow rate will limit the reactions within the half-cells and
0
0 10 20 30 40 reduce the cycling time, thereby leading to loss of capacity that
No. of cycles
is not predicted by the present model. Differential flow rates or a
Fig. 20. Vanadium ion concentrations vs. cycle number at 1.7 V and 1.1 V for Nafion high different pressure drop across the membrane could also cause
115 membrane with 0.98 current efficiency factor for V2+ charging reaction. significant bulk electrolyte transfer from one half-cell to the other
10746 A. Tang et al. / Journal of Power Sources 196 (2011) 10737–10747

that swamps the effects of ion transfer, leading to further incon- 1.8
sistencies between experimental and simulated trends. Moreover, constant hydrogen ion concentration
1.6
variable hydrogen ion concentration
the water carried by the vanadium ions and dragged by the proton

Cell voltage / V
across the membrane can also have a slight impact on the imbal- 1.4
ance of the two half-cell electrolytes and this has been neglected in 1.2
the present model. Last but not least, water is also associated with
the charge–discharge reactions in the positive half-cell electrolyte 1
where water is consumed and generated during the charging and 0.8
discharging processes as shown in Eq. (1). In real-life VFB sys- 0 0.2 0.4 0.6 0.8 1
State of charge
tems therefore, the volumes of the two half-cell electrolytes keep
changing during charge–discharge cycling, but laboratory obser- 40
vations show that a stead state is reached after a certain number
of cycles depending on the type of membrane and cycling condi- 30

Voltage / mV
tions used. Nevertheless, the exact volumes of the electrolytes in
both half-cells as a function of time are difficult and even impossi- 20
ble to determine mathematically due to the complicated equilibria
in the electrolytes that make it difficult to accurately calculate the 10
total ionic strength of the solutions during cycling and the resultant
changes in osmotic pressure. In the present case, the accuracy of the 0
0 0.2 0.4 0.6 0.8 1
model will be affected by the assumption of constant electrolyte State of charge
volume. But given the complexity of the solution chemistry, theo-
Fig. 22. Comparison of theoretical cell voltages calculated by assuming constant
retical predictions would be very inaccurate [27]. Further studies and variable hydrogen ion concentration in the Nernst equation.
on the effect of volumetric transfer on the model should there-
fore incorporate experimental measurements of electrolyte volume
changes during charge–discharge cycling of a VRB battery system shape of the dynamic cell voltage curve, the variation in hydrogen
in order to improve the accuracy of prediction. Unfortunately very ion concentration during charge–discharge cycling has no effect on
little data on dynamic volume changes is available in the literature, cell capacity that is the main purpose of the present dynamic model.
so for the purpose of the present simulation, this effect has been
neglected. 5.4. Other effects and preventive actions

5.3. The effect of concentration of hydrogen ions Apart from the factors mentioned above, other factors such as
charge–discharge current magnitude and the effect of shunt cur-
The concentration of hydrogen ions in the positive half-cell rents in a bipolar stack will also have a considerable influence on
electrolyte varies during charge–discharge cycling due to the trans- cell capacity and further add to discrepancies between experimen-
fer of hydrogen ions across the membrane for charge balance as tal and simulated trends. Hence, such additional factors need to be
well as changes associated with the redox reaction at the posi- included in the further development of the model so as to not only
tive electrode. As discussed above, the chemistry of the vanadium perform a more reasonable capacity prediction, but provide deeper
electrolytes is very complex with several ionic equilibria interact- insight into the real-life system operation as well. From the above
ing to change the equilibrium concentration of hydrogen ions in case studies, in the meanwhile, it is seen that side reactions in nega-
the positive electrolyte during cycling [27]. Given that equilibrium tive electrolyte during charging could have a significant impact on
constants for many of these reactions are not accurately known, capacity loss if appropriate operational limits and battery design
accurate calculation of equilibrium proton concentrations during considerations are not implemented. In practical VFB systems,
cycling is very difficult. Changes in the hydrogen ion concentration however, many actions can be taken to eliminate the effect of side
during change-discharge cycling are expected to affect the Nernst reactions on capacity loss. Examples include the employment of an
potential for the positive half-cell reaction and therefore the shape inert gas or other type of air blanket within the negative electrolyte
of the dynamic voltage curve produced by the simulation. In order reservoir to minimise air oxidation of V2+ , avoiding overcharging
to understand the magnitude of this effect, the cell voltage curve and elimination of electrolyte impurities that will catalyse hydro-
was modelled for both situations of constant and varying proton gen evolution. None of them can, however, completely eliminate
concentration. Assuming that the electrolyte contains 2 M vana- the influence of side reactions over many thousands of cycles. As
dium in 2.5 M total sulfate, for instance, the following proton mass such, the prediction of capacity loss will help with automated elec-
balance for the positive half-cell electrolyte is incorporated into the trolyte management decisions for periodic electrolyte rebalancing
Nernst Equation in order to calculate the effect of varying hydrogen to maintain and restore capacity.
ion concentration on the theoretical cell voltage curve:
6. Conclusions
[H+ ] = [H+ ]0 + 2 × SOC (13)
In Eq. (13), it is assumed that during charging, one mole proton A dynamic model based on molar mass balance equations for
transfers across the membrane from the positive half-cell to the the four vanadium species in the all-vanadium redox battery has
negative half-cell for every mole of vanadium ions reacting, while been developed and presented in this paper. Using the Nernst Equa-
2 moles of protons are also produced by reaction (1). Assuming that tion, a dynamic model for cell potential has been developed and
the hydrogen ions are not affected by other ionic equilibria in the used to simulate and predict the capacity loss caused by diffusion
solutions, Eq. (13) can be incorporated into the Nernst Equation to of vanadium ions across the membrane as well as side reactions
estimate the effect of varying hydrogen concentration in the pos- in the negative half-cell over long term charge–discharge cycling
itive half-cell electrolyte on the dynamic cell voltage. As seen in and operation of the VFB. Three membranes with different sets of
Fig. 22, the voltage difference between the variable and constant reported diffusion coefficients have been employed in the simula-
proton concentration cases varies from zero to 30 mV as a func- tion. The results have shown that the rate of capacity loss versus
tion of state of charge. Importantly however, while affecting the cycle number under preset limits of cell voltage is influenced by
 A. Tang et al. / Journal of Power Sources 196 (2011) 10737–10747 10747

both the absolute and relative magnitudes of the different vana- systems. Further investigation with respect to all the above issues
dium ion diffusion coefficients. In the absence of side reactions, is in progress and results will be presented in a future publication.
larger diffusion rates tend to level off faster and attain a steady-
state capacity over a smaller number of charge–discharge cycles. References
Additionally, the concentration variation of vanadium ions and the
direction of net vanadium accumulation are associated with the [1] M. Skyllas-Kazacos, G. Kazacos, G. Poon, H. Verseema, Int. J. Energy. Res. 34
(2010) 182–189.
relative values of the diffusion coefficients of the four different [2] M. Skyllas-Kazacos, Encyclopedia of Electrochemical Power Sources, Elsevier,
vanadium ions in the VFB. With respect to side reactions, air oxida- Amsterdam, 2009, pp. 444–453.
tion of V2+ and evolution of hydrogen during charging will result in [3] M. Skyllas-kazacos, M. Rychick, R. Robins, All-vanadium redox battery, United
States Patent, No. 4,786,567 (1988).
significant capacity loss during ongoing cycling. Even a small loss [4] M. Skyllas-Kazacos, D. Kasherman, D.R. Hong, M. Kazacos, J. Power Sources 35
of current efficiency for the negative half-cell reactions will pro- (1991) 399–404.
duce a dramatic drop in capacity, highlighting the importance of [5] F. Grossmith, P. Llewellyn, A. Fane, M. Skyllas-Kazacos, Proceedings of Sym-
posium on Energy Storage: Load Levelling and Remote Applications, The
good voltage control to reduce gassing side reactions and negative
Electrochemical Society, 1988, p. 11.
electrolyte blanketing to prevent air oxidation of V2+ . [6] S.C. Chieng, M. Kazacos, M. Skyllas-Kazacos, J. Power Sources 39 (1992)
The dynamic model will play an essential role in predicting the 11–19.
scheduling of periodic electrolyte rebalancing in the VFB so as to [7] S.C. Chieng, M. Kazacos, M. Skyllas-Kazacos, J. Membr. Sci. 75 (1992) 81–91.
[8] T. Mohammadi, M. Skyllas-Kazacos, J. Membr. Sci. 98 (1995) 77–87.
restore the capacity loss caused by ion diffusion and side reac- [9] T. Mohammadi, M. Skyllas-Kazacos, J. Power Sources 56 (1995) 91–96.
tion effects, thereby enabling long term charge–discharge cycling [10] T. Mohammadi, M. Skyllas-Kazacos, J. Membr. Sci. 107 (1995) 35–45.
to be carried out without a large loss of capacity. Besides, this [11] T. Mohammadi, M.S. Kazacos, J. Appl. Electrochem. 27 (1997) 153–160.
[12] T. Mohammadi, M.S. Kazacos, J. Power Sources 63 (1996) 179–186.
study highlights the importance of rebalance cell technology as well [13] T. Sukkar, M. Skyllas-Kazacos, J. Membr. Sci. 222 (2003) 235–247.
as advanced battery control systems that can monitor electrolyte [14] T. Sukkar, M. Skyllas-Kazacos, J. Membr. Sci. 222 (2003) 249–264.
imbalance and maintain electrolyte chemistry in order to manage [15] T. Sukkar, M. Skyllas-Kazacos, J. Appl. Electrochem. 34 (2004) 137–145.
[16] J. Xi, Z. Wu, X. Qiu, L. Chen, J. Power Sources 166 (2007) 531–536.
the long-term operation and performance of the VFB. Laboratory [17] Q. Luo, H. Zhang, J. Chen, P. Qian, Y. Zhai, J. Membr. Sci. 311 (2008) 98–103.
observations have, however, shown that other factors such as vol- [18] C. Jia, J. Liu, C. Yan, J. Power Sources 195 (2010) 4380–4383.
umetric transfer of solutions from one half-cell to the other can also [19] A.A. Shah, M.J. Watt-Smith, F.C. Walsh, Electrochim. Acta 53 (2008) 8087–8100.
[20] H. Al-Fetlawi, A.A. Shah, F.C. Walsh, Electrochim. Acta 55 (2009) 78–89.
be caused by differential pumping pressure across the membrane [21] H. Al-Fetlawi, A.A. Shah, F.C. Walsh, Electrochim. Acta 55 (2010) 3192–3205.
and this will also give rise to an additional impact on capacity and [22] A.A. Shah, H. Al-Fetlawi, F.C. Walsh, Electrochim. Acta 55 (2010) 1125–1139.
performance of the VFB. These issues will be taken into account in [23] M. Li, T. Hikihara, IEICE Trans. Fund. Electron. Commun. Comput. Sci. E91-A
(2008) 1741–1747.
the further development of the VFB dynamic modelling that con-
[24] D. You, H. Zhang, J. Chen, Electrochim. Acta 54 (2009) 6827–6836.
tinues to be evolved. Furthermore, the effect of electrolyte flow [25] C. Sun, J. Chen, H. Zhang, X. Han, Q. Luo, J. Power Sources 195 (2010) 890–897.
rate on vanadium ion concentrations in the cells, which in turn [26] M. Skylas-Kazacos, M. Kazacos, R. Wegner, R. Burford, G. Lin, Final Report NSW,
influences the charge–discharge cycling time and capacity has also Dept. of Minerals & Energy, 1991, pp. 200.
[27] C. Blanc, Modelling of a vanadium redox flow battery electricity storage system,
been ignored in this paper, but will be incorporated into the refined Ph.D. Thesis, EPFL, Switzerland, 2009.
model to better simulate the capacity and performance of real VFB