Monitoring the state of charge of all-vanadium redox flow batteries to identify crossover of electrolyte

Journal Pre-proof Monitoring the state of charge of all-vanadium redox flow batteries to identify crossover of electrolyte T. Haisch, H. Ji, C. Weidlich PII:

S0013-4686(19)32445-4

DOI:

https://doi.org/10.1016/j.electacta.2019.135573

Reference:

EA 135573

To appear in:

Electrochimica Acta

Received Date: 19 September 2019 Revised Date:

17 December 2019

Accepted Date: 23 December 2019

Please cite this article as: T. Haisch, H. Ji, C. Weidlich, Monitoring the state of charge of all-vanadium redox flow batteries to identify crossover of electrolyte, Electrochimica Acta (2020), doi: https:// doi.org/10.1016/j.electacta.2019.135573. This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. © 2019 Published by Elsevier Ltd.

CRediT author statement Theresa Haisch: Writing - Original Draft, Visualization, Data Curation Hyunjoon Ji: Investigation Claudia Weidlich: Conceptualization, Writing - Review & Editing, Supervision, Project administration

Monitoring the state of charge of all-vanadium redox flow batteries to identify crossover of electrolyte T. Haischa , H. Jia , C. Weidlicha,∗ a

DECHEMA Research Institute, Electrochemistry, Frankfurt am Main, Germany

Abstract During charging and discharging of an all-vanadium redox flow battery electrolyte components cross the membrane in the battery cell. This so called crossover leads to partial discharging and capacity loss. For the identification of electrolyte crossover and efficient operation of the battery the accurate and reliable determination of the state of charge is essential. In this study, state of charge estimation from open cell voltage measured currentless at a reference cell as well as from open circuit potentials measured at flow cells in the positive and negative electrolyte loop is discussed. Comparison of the state of charge obtained from the different potential measurements suggests, that monitoring the half-cell potentials of the battery allows timely detection of crossover processes and furthermore the determination of the direction of crossover between the half-cells. Keywords: Vanadium redox flow batteries, state of charge monitoring, Crossover, Open circuit potential

1. Introduction The necessity of storage systems for regenerative energy is steadily increasing due to the worldwide energy consumption and the challenges to mitigate climate change. The redox flow battery is an appropriate energy storage system that fulfills the requirements of a broad range of applications, mainly due to the characteristic of independent scalability of energy and performance [1, 2]. Commercially prevailed is the all-vanadium redox flow battery (VRFB) ∗

Corresponding author: [email protected], T+49-(0)69-7564-633

Preprint submitted to Electrochimica Acta

December 26, 2019

which was developed in the 1980s [3]. This battery is characterized by the use of vanadium electrolyte in both the negative half-cell (NHC) and positive half-cell (PHC) [4]. Charge

2+ −−−− → V 3+ + e− − ← −−−− −V Discharge

E0 (NHC) = -0.26 V vs. SHE (1)

Charge

+ + − −−−− → VO2+ + H 2 O − ← −−−− − VO2 + 2 H + e Discharge

E0 (PHC) = 1.00 V vs. SHE (2) During charging V(III) ions are reduced to V(II) in the NHC while the oxidation of V(IV) ions to V(V) occurs in the PHC. During discharging, the reactions in both half-cells occur in the opposite direction. The use of ions from the same chemical element offers the major advantage that crossover from one battery half-cell to the other does not lead to contamination and can be reversed by remixing of the electrolyte [5]. However, in addition to the vanadium ions other species can transfer through the membrane. Besides protons and sulfate ions crossing the membrane, water transport across the membrane contributes to capacity decay, which is associated with more frequent maintenance and limits the lifetime. In addition to capacity loss by crossover, several side reactions might irreversibly change the composition of the electrolyte in the single half-cells. Hydrogen or oxygen can evolve from the aqueous solution at low or high potentials, respectively, affecting the proton concentration. 2 H + + 2 e− −−→ H 2

(3)

2 H 2 O −−→ O2 + 4 H + + 4 e−

(4)

2

The presence of air or oxygen, respectively, has to be avoided, essentially in the negative half-cell, because it oxidizes the V(II) ions and lowers the SOC of the NHC. V 2+ + O2 + 4 H + −−→ V 3+ + 2 H 2 O

(5)

In general, the state of charge (SOC) is defined as the actual battery capacity (Qact ) as a percentage of the maximum capacity (Qtheo ). SOC =

Qact Qtheo

(6)

The term capacity indicates the amount of electric charge a battery is able to store or supply, respectively. The electric charge can be calculated by applying Faraday’s law with the number of electrons per reaction (z), the moles of active species dissolved in the electrolyte (n) and the Faraday constant (F). Q = z ·n ·F

(7)

The capacity of the VRFB is determined by the concentration of the vanadium ions in both half-cells. Hence, the SOC of a VRFB can be expressed by the ratio of the active species to the overall vanadium concentration in each half-cell. S O CNHC =

S O CPHC =

[V 2+ ] [V 2+ ] + [V 3+ ] [V O2+ ]

[V O2+ ] + [V O 2+ ]

(8)

(9)

During operation of the RFB knowledge of the SOC is essential to ensure efficient charging and discharging of the battery. SOC determination is also needed to avoid deep discharge and overcharge which might lead to accelerated aging and degradation of the battery [6–8]. To increase the lifetime of the battery, the SOC is significant for the development of effective operation strategies and battery management. Therefore, accurate and reliable online SOC estimation is essential to improve battery control and to achieve efficient long term operation 3

of the system [9, 10]. A simple method to monitor SOC is the measurement of the electrical charge during charging and discharging processes, the so called coulomb counting (CC) [11, 12]. This method relies on the time integral of the transferred charge during charging/discharging processes in relation to the maximum capacity of the battery. R t1 S O Ct =

t0

I (t )d t

Qtheo

+ S O C0

(10)

SOC estimation via CC is rather inaccurate because side reactions and capacity losses during operation of the battery due to crossover, self-discharge and degradation of the electrolyte alter the capacity of the system [13]. The resulting error accumulates over each additional cycle giving relatively high deviations. Furthermore, the precise predetermination of the initial SOC (SOC0 ) is required in order to obtain reliable results [14]. But the initial value of the SOC can not be extracted directly from coulomb counting. Therefore the predetermination of SOC0 has to be fulfilled with an additional technique like titration to receive exact values of the ion concentrations and thereby the initial SOC. Another simple and widely used method of SOC estimation is based on the measurement of the open circuit voltage (OCV) of the battery [15–18]. Since it is not possible to measure OCV of the battery currentless during operation a further cell equipped with the identical materials contained in the battery cell, a so called reference cell, is integrated into the electrolyte loop. The voltage at this cell is recorded continuously during charging and discharging but no current is flowing at the electrodes of this cell. The measured OCV is linked via the Nernst equation to the SOC of the battery using the definition of the SOC (Eq. 8 and Eq. 9). O C Vc e l l = ϕc0 e l l

 ‹ RT (1 − S O C )2 ln − zF S O C 2 · [H + ]2

(11)

ϕc0 e l l is the standard reduction potential of the cell (here 1.26 V [19]), R is the universal gas constant (8.314 J K−1 mol−1 ) and T is the temperature (298 K). The OCV method reflects the entirety of the system and thus is unable to differentiate 4

between the individual half-cells. This limits the method for practical implementation since it is not applicable to reflect concentration changes based on crossover or side reactions. To enable the determination of electrolyte crossover in earlier stages as well as to improve the accuracy of SOC estimation, the open circuit potentials (OCP) of the negative and positive half-cell are monitored individually. Knehr et al. proposed calculation models using the Nernst equation, by which the SOC dependent half-cell potential OCPNHC and OCPPHC are used for various experimental conditions [19]. O C PNHC = ϕ

O C PPHC = ϕ

 ‹ RT SOC − ln zF (1 − S O C )

(12)

 ‹ (1 − S O C ) RT ln − zF S O C · [H + ]2

(13)

0−

0+

In the following sections we examine and compare results obtained from OCV measurements at a reference cell and OCP measurements with auxiliary flow cells at the negative and positive half-cell as SOC monitoring methods and their suitability to determine electrolyte crossover.

2. Experimental section A VRFB single cell (DECHEMA Machine Shop) with an active area of 40 cm2 was assembled to conduct charging and discharging experiments. The cell consists of a membrane (fumasep® FAP-450 or fumapem® FS-930 FUMATECH BWT GmbH) sandwiched between two carbon felts (SIGRACELL® GFD 4.6, SGL Carbon, compressed (11 %)). Graphite bipolar plates (FU4369HT, Schunk Carbon Technology) and copper plates were used as current collector at both sides of the cell. The cell was connected to a Flow Battery Test Station (Greenlight Innovation). The tanks were each filled with 0.5 L of fresh vanadium electrolyte solution with a total vanadium concentration of 1.6 M (GfE Metalle und Materialien GmbH). The electrolyte was pumped with a flow rate of 3 L h−1 through the system. To minimize influence of oxygen, the system was purged with nitrogen (0.25 l min−1 ). Before the first charge cycle of the battery, the fresh electrolyte (mixture of V(III) and V(IV)) has been conditioned. Thereby, in the PHC V(III) ions are oxidized to V(IV) ions whereas V(IV) ions are reduced to V(III) in the NHC. The conditioning 5

Figure 1: Simplified diagram of Flow Battery Test Station.

6

of the electrolyte was directly followed by the first charge cycle. The cell was charged to 1.65 V. The discharge limit was set to 0.8 V. Five cycles were performed with a current density of ± 50 mA cm−2 . Between each cycle step (charge/discharge), the OCV of the battery cell was monitored for 5 min. Figure 1 shows the process flow diagram of the test station and equipment. OCV measurements were done in a reference cell (Micro Flow Cell, ElectroCell A/S) with an active area of 10 cm2 . The same materials as for the battery cell were used. For OCP measurement at the NHC and PHC, two single flow cells (DECHEMA Machine Shop) with a volume of 5 mL were placed as auxiliary cells directly at the outlets of the PHC and NHC of the battery cell to avoid different retention times. The auxiliary flow cells were equipped each with a glassy carbon rod (SIGRADUR® G, Hochtemperatur-Werkstoffe GmbH, 2 mm diameter) as working electrode and a Hg/HgSO4 reference electrode (HgE 11-S, Sensortechnik Meinsberg Xylem Analytics GmbH & Co. KG). The reference electrodes were placed in luggin capillaries filled with 2 M H2 SO4 to prevent the reference electrodes from vanadium contamination. The glassy carbon rods have been polished with carbide paper (grid 2000), pretreated with an electrochemical polarisation routine to ensure reproducible and homogeneous composition of the carbon surface. Furthermore the GC rods have been investigated with cyclic voltammetry in fresh vanadium electrolyte as well as in charged and discharged electrolytes of the NHC and PHC. More information about these results are given in the supplementary information (SI). 3. Results and Discussion To investigate the capability of the described setup for online estimation of SOC and identification of crossover, OCV at the reference cell as well as OCP at the auxiliary flow cells at the PHC and NHC were measured simultaneously during charging and discharging of the battery cell with 50 mA cm−2 . The battery and reference cell were either equipped with an anion-exchange membrane (AEM) or a cation-exchange membrane (CEM). Figure 2 represents the charge and discharge cycles of the battery cell. Charging of the battery is proceeded until the cut-off voltage of 1.65 V at the battery cell is reached and discharging is proceeded until the lower voltage limit, which is set to 0.8 V, is reached. 7

Figure 2: Cell voltage and current density of the VRFB with an AEM (FAP-450) during 5 charge and discharge cycles at 50 mA cm−2 (A = 40 cm2 ).

8

Figure 3 compares the OCV measured at the reference cell (OCVmeas ) with the OCV calculated from the difference of the OCP of the individual half-cells (OCVcalc ) which are depicted in figure 4. In addition the OCV measured at the battery cell at the end of each charging and discharging process (OCVbat ) during a short (300 s) currentless break is shown. Comparison of measured and calculated OCV (Fig. 3) suggest an accurate prediction of the cell OCV by measuring OCP of the half-cells. A small deviation between the potentials measured at the discharged state occurs where OCVmeas reaches lower potentials than OCVcalc . This might be due to the different carbon materials used for OCV determined at the reference cell (carbon felt) and OCP measured at the auxiliary flow cell (GC rod). Furthermore, a certain ohmic drop is caused by the membrane in the reference cell whereas the OCP at the auxiliary flow cell is not influenced by a membrane. The SOC can be calculated from the measured OCV of the reference cell by the Nernstbased equation (Eq. 11). The standard redox potential of the cell used in the Nernst equation is 1.26 V. The equation also includes the concentration of protons which is mostly unknown due to the involvement of protons in charging and discharging of the PHC electrolyte (Eq. 2). Also crossover of protons through the membrane affects the proton concentration in the half-cells. For simplification, these two parameters, proton concentration and standard redox potential, are often used with default values or completely disregarded. Figure 5 shows the effect of these two parameters on the resulting OCV-SOC relationship. The theoretical calculation of the relation between SOC and OCV without considering the proton concentration and assuming a standard redox potential behaviour is shown in Figure 5a (black curve). Another often used approach is the use of a constant proton concentration. The concentration of the sulfuric acid solution is 2 M. Sulfuric acid is a polyprotic acid and dissociates in two steps. Since both H2 SO4 and HSO−4 can be considered as strong acids, at low pH values the acid dissociates completely giving a proton concentration of 4 M in the electrolyte solution. The theoretical calculated curve (red) is shifted to higher values on the yaxis compared to the curve without considering the proton concentration. Due to the reaction in the PHC, the proton concentration is not constant during charging and discharging. Based on the reaction in the PHC (Eq. 2) regarding the proton concentration the assumption is made, 9

Figure 3: OCV monitoring during charging and discharging of a battery cell with AEM. Comparison of OCVcalc (from Figure 3), OCVmeas and OCVbat .

10

Figure 4: OCP monitoring during charging and discharging of a battery cell with AEM. Half-cell potentials OCPPHC and OCPNHC measured at auxiliary flow cells.

11

(a)

(b)

Figure 5: Theoretical calculated OCV curves with different assumptions regarding (a) the proton concentration and (b) the formal potential E0∗ .

12

that in the PHC during charging two moles of protons are released per mole vanadium. During discharging the process occurs vice versa, so two moles protons are required. The proton concentration in the NHC is not affected, diffusion to balance the new build charge difference is neglected. The resulting curve concerning a variable proton concentration (green curve) shows also slightly higher OCV values, especially at rather high SOC levels. Another assumption includes permeation of ions to balance the charge difference (orange curve). Depending on the membrane type the crossing ions are either protons or anions as SO2− or HSO−4 , both 4 influencing the pH in the PHC and NHC as well. The permeation of protons results in a pH change of one mole protons per mole vanadium in both half-cells. Since the Nernst equation for the NHC does not depend on the proton concentration at low pH values, the pH change does not influence the OCP-SOC relation in the NHC and thereby has a lower effect on the OCV-SOC relation. Comparison of the received curves indicates that the proton concentration exerts a strong effect on the associated OCV-SOC pair of values and therefore neglecting the proton concentration results in a strong deviation. Similar considerations have also been discussed by Knehr et al. and Corcurea et al. [19, 20]. Another parameter that can be modified in the calculation is the standard redox potential. It accounts 1.26 V for an aqueous VRFB but is often replaced by the formal potential, E0∗ , to consider activity coefficients of the ions [20, 21]. The arithmetic mean of the charged and discharged state of the first cycle (Fig. 6) of the OCV measurement is considered as the formal potential assuming that it correlates to a SOC of 50 %. Based on the investigation of a battery cell with an AEM (Fig. 3) 1.35 V is obtained as formal potential. The effect of using a formal potential instead of the standard redox potential on the SOC-OCV relation can be seen in Figure 5b (blue curve) as well. The curve is shifted on the y-axis to a higher OCV for a certain SOC by the difference between the potential values (here 0.9 V) but the slope of the curve remains the same. As the previous analysis shows, there are terms like the formal potential and the proton concentration which influence the estimation of SOC based on Nernst equation. Their influence has to be considered carefully since the conversion of the potential into the SOC uses the ion concentrations instead of the activities as the equation originally requires. As a result, 13

Figure 6: OCV measured at the reference cell during charging and discharging and SOC calculated from the OCV data using Equation 11 assuming constant proton concentration (4 M) and a formal potential of 1.35 V vs. SHE. Battery cell and reference cell are equipped with an AEM.

deviations may occur. For further calculation of the SOC from measured potential (OCV at reference cell and OCP at flow cells) the formal potential (OCV: 1.35 V, OCP: -0.28 V and 1.07 V vs. SHE) as well as a constant proton concentration (4 M) are inserted into the Nernst equation.

SOC estimation based on measured OCV data is illustrated in Figure 6. In general, a vanadium redox flow battery is cycled between 20-80 % SOC to keep the system under mild conditions and avoid imbalances [22–25]. To accelerate electrolyte crossover through the membrane, the upper voltage limit during charging was set to 1.65 V, corresponding to a SOC of about 95 %. During discharging, 0.8 V, which equals a SOC of 0-5 %, was chosen as lower limit of voltage [23, 26, 27]. The occurrence of crossover is proved by the potential shift of the OCV 14

Figure 7: SOC estimation from OCP of each half-cell (Fig. 4) separated by an AEM measured in auxiliary flow cells during charge and discharge cycles.

curve especially for the discharged state. Since the measurement of the OCV or the estimation of the associated SOC only consider the net balance of the system, detailed information about the crossover, e.g. amount, direction or type of ions can not be received. Crossover can be investigated more detailed by measuring the redox potential, respectively the open circuit potential (OCP) of the half-cell electrolytes [13, 19, 20]. The OCP measured in the PHC and NHC using auxiliary flow cells is depicted in Figure 4. SOC estimation from OCP data applies the Nernst-based equation as well, but for the individual half-cells (Eq. 12 and Eq. 13). The standard redox potential is replaced by the formal potential likewise to the SOC calculation from OCV giving -0.28 V for the NHC and 1.07 V for the PHC. As well as for the SOC calculation from OCV data, constant proton concentration can be 15

assumed for the half-cells. The calculated SOC resulting from the OCP measured at the PHC and NHC based on OCP data are shown in Figure 7. The curves reflect a shift of the maximum and minimum SOC over time. The SOC of the charged NHC is shifting to lower values whereas the SOC of the discharged PHC is shifting to higher values. That means with ongoing cycling, the NHC can not be completely charged whereas the PHC can not be completely discharged any more. In comparison, the SOC calculated from the OCV measured at the reference cell (Fig. 6) does not depict such an shift of SOC for the charged and discharged state during cycling; only a slight increase of SOC can be observed for the discharged state as a difference of the individual cells. In a battery cell containing an AEM, vanadium ions cross the membrane from the PHC to the NHC during charging [28]. The transfer of ions into the NHC results in two effects. On the one hand, the theoretical capacity in the PHC is lowered by the reduced amount of vanadium ions and simultaneously increased in the NHC. On the other hand, the transferred ions react with the charged species V(II) to uncharged V(III) ions [29, 30].Therewith during charging the crossover induces also discharging of the NHC electrolyte by the conversion of V(II) ions to V(III) under the consumption of the crossed species. The capacity of the entire cell is limited by the half-cell with the lowest capacity. Due to the crossover from PHC to NHC, the capacity of the PHC defines the cell capacity. During charging, the PHC can be completely charged but the NHC remains partially uncharged since its capacity can not be fully utilized. Therefore, the maximum SOC of the NHC decreases over time. During discharging, the NHC can be completely discharged because it has not been fully charged and a part of the charged ions were already discharged by the ions from the crossover. Though, the PHC can no longer be completely discharged so that the SOC shifts in the discharge state of the PHC. A shift in the SOC of the NHC may additionally occur due to the oxidation of V(II) ions by air giving a possible explanation for the different slopes of the SOC shift over the number of cycles (Fig. 7). These results suggest that OCP half-cell monitoring allows the determination of the crossover direction after a few cycles. The observed SOC shift is dedicated to crossover of electrolyte components through the membrane of the battery cell. In general, the species which are crossing the membrane and 16

(a) AEM

(b) CEM

Figure 8: Direction of crossover in dependence on the membrane type. The crossover of the electrolyte ions (H+ and SO2+ 4 ) depends on charged or discharged state. The tank volume changes due to water transport across the membrane. AEM = anion-exchange membrane, CEM = cation-exchange membrane.

17

the direction of crossover depend on the membrane type (Fig. 8). Using an AEM leads to a net crossover of ions from the PHC to the NHC. In contrast, net crossover takes place from the NHC to the PHC with the use of a CEM [29, 31–33]. However, this relates only to the net crossover respectively the change of electrolyte volume and does not reflect all occurring processes. The ion transport across the membrane is mainly driven by two forces: the concentration-gradient causes ion diffusion and the electric-field induces ion migration [31, 32]. The direction of diffusion through the membrane is independent on charging or discharging whereas migration changes its direction in dependence on charging and discharging [29, 34]. Hence, crossover processes alter the SOC in dependence on the type of membrane and the charging and discharging of the battery. To study crossover in dependence on the membrane type, experiments were repeated with the same parameters and setup (Fig. 1) with a battery containing a cation-exchange membrane (CEM). As described (Fig. 8) crossover from NHC to PHC is expected which leads to partial discharge of the PHC and decrease of capacity in the NHC. During these experiments measuring the OCP at the PHC was connected with some difficulties leading to unstable potential values. Therefore the OCP was calculated using the OCV measured at the battery cell (OCVb a t , Fig. 9) and OCP measured at the NHC (Fig. 10). OCVb a t was used for the calculation because of the higher agreement between OCVb a t and OCVc a l c in the experiment with an AEM (Fig. 3). Figure 11 shows the SOC of the NHC of battery cells equipped with either AEM or CEM during the first five cycles. As already described the SOC of the charged NHC decreases from cycle to cycle using a cell with an AEM, whereas with a CEM, the SOC of the discharged NHC increases. With the CEM crossover takes place mainly during the discharging process from the NHC to the PHC. This lowers the capacity in the NHC and an additional discharging caused by the crossover occurs in the PHC. Therefore, at the end of the discharge cycle the PHC is already completely discharged whereas some (state of) charge remains in the NHC. With the CEM crossover exhibits a stronger influence on the SOC during discharging of the battery because diffusion and migration proceed in the same direction. In comparison, crossover using an AEM has a stronger influence on the SOC during charging as the directions of migration and crossover coincide. 18

Figure 9: OCV monitoring during charging and discharging of a battery cell with CEM. Comparison of OCVmeas and OCVbat .

19

Figure 10: OCP monitoring during charging and discharging of a battery cell with CEM. Half-cell potentials OCPPHC,calc and OCPNHC measured at auxiliary flow cell.

20

Figure 11: SOC estimation from OCP data of NHC with two different membranes, an AEM and a CEM during charging and discharging.

21

The changes in SOC describe a temporary capacity loss. At which state of the cycle, either charge or discharge, the capacity decreases, provides information about the direction of the net crossover. The slope of capacity fade might offer predication about the amount of crossover. The thinner membrane (FS-930, 25-35 µm) which usually suffers from higher crossover flux displays a stronger SOC shift considering the sum of the NHC and PHC slopes than the thicker membrane (FAP-450, 45-55 µm) at discharged state (NHC). However, regardless of the membrane type, the SOC shift in the NHC is stronger. This might indicate a strong influence of the different diffusion coefficients. The smaller size of V2+ and V3+ compared to VO2+ and VO+2 facilitate their transport [32, 35, 36]. Another aspect which might explain the stronger changes of the SOC in the NHC is the sensitivity to oxygen leading to a reduction of the amount of charged V(II) species. Comparison of the SOC estimated for both battery cells with different types of ion exchange membranes shows that electrolyte components cross the AEM from PHC to NHC mainly during charging of the battery whereas crossover through the CEM takes place mainly during discharging from the NHC to the PHC of the battery. Another aspect, in which the results for the AEM and CEM differ, is the duration of the charging and discharging cycles. The cycles are slightly shortened following the use of a CEM. A lower resistance or higher membrane conductivity, respectively, might be the source for this. However, the relative change in the duration of a charge and discharge cycle might also give an indication of the amount of crossover. The shortening of cycles is significantly stronger with the thinner CEM than with the thicker AEM. These results will be further studied with different membrane thicknesses and verified by other SOC estimation methods in order to obtain and develop not only qualitative but also quantitative statements for crossover. 4. Conclusions The use of Nernst equation enables the estimation of the state of charge (SOC) by measuring the open circuit voltage (OCV) of a reference cell or open circuit potentials (OCP) at negative (NHC) and positive half-cell (PHC), respectively. However, the importance to consider the proton concentration and the correct formal potential of the present system to determine 22

accurate results is shown. The calculation of the SOC based on OCP data from NHC has the advantage that it is theoretically independent on the proton concentration and minimizes a potential error. Our results show that both OCV and OCP measurements are suitable for SOC estimation. Rather than the widely used OCV monitoring at reference cells, monitoring OCP of the individual half-cells provides additional information about crossover processes, especially the direction of crossover, timely and reliable. Despite neither OCV nor OCP measurements may distinguish between the crossing electrolyte species, the net crossover and therefore changes in the electrolyte volume of the negative and positive half-cells can be determined and furthermore be forecasted by OCP monitoring. Therefore it is a suitable tool for electrolyte management, respectively remixing and rebalancing VRFB electrolytes. Moreover SOC estimation and crossover determination from OCP monitoring can be conveyed to other redox flow electrolytes and is therefore also suitable for an application in research and development of alternative flow battery electrolytes.

Acknowledgments The authors gratefully acknowledge financial support from Federal Ministry for Economic Affairs and Energy BMWi (Project "DegraBat" No. 03ET6129A).

References References [1] A. Z. Weber, M. M. Mench, J. P. Meyers, P. N. Ross, J. T. Gostick, Q. Liu, Redox flow batteries: a review, J. Appl. Electrochem. 41 (2011) 1137. [2] L. Arenas, C. P. De León, F. Walsh, Engineering aspects of the design, construction and performance of modular redox flow batteries for energy storage, J. Energy Storage 11 (2017) 119–153. [3] M. Skyllas-Kazacos, M. Rychcik, R. G. Robins, A. Fane, M. Green, New all-vanadium redox flow cell, J. Electrochem. Soc. 133 (1986) 1057. [4] M. Skyllas-Kazacos, L. Cao, M. Kazacos, N. Kausar, A. Mousa, Vanadium electrolyte studies for the vanadium redox battery—a review, ChemSusChem 9 (2016) 1521–1543.

23

[5] Y.-W. D. Chen, K. Santhanam, A. J. Bard, Solution redox couples for electrochemical energy storage i. iron (iii)-iron (ii) complexes with o-phenanthroline and related ligands, J. Electrochem. Soc. 128 (1981) 1460–1467. [6] M. Rychcik, M. Skyllas-Kazacos, Evaluation of electrode materials for vanadium redox cell, J. Power Sources 19 (1987) 45–54. [7] M. Skyllas-Kazacos, F. Grossmith, Efficient vanadium redox flow cell, J. Electrochem. Soc. 134 (1987) 2950–2953. [8] A. Parasuraman, T. M. Lim, C. Menictas, M. Skyllas-Kazacos, Review of material research and development for vanadium redox flow battery applications, Electrochim. Acta 101 (2013) 27–40. [9] S. Piller, M. Perrin, A. Jossen, Methods for state-of-charge determination and their applications, J. Power Sources 96 (2001) 113–120. [10] H. He, R. Xiong, J. Fan, Evaluation of lithium-ion battery equivalent circuit models for state of charge estimation by an experimental approach, Energies 4 (2011) 582–598. [11] Z. Wei, R. Xiong, T. M. Lim, S. Meng, M. Skyllas-Kazacos, Online monitoring of state of charge and capacity loss for vanadium redox flow battery based on autoregressive exogenous modeling, J. Power Sources 402 (2018) 252–262. [12] M. Murnane, A. Ghazel, A closer look at state of charge (soc) and state of health (soh) estimation techniques for batteries, Internet: http://www. analog.com/media/en/technical-documentation/technical-articles/ACloser-Look-at-State-Of-Charge-and-State-Health-Estimation-Techniques-.... pdf (2017). [13] S. Ressel, F. Bill, L. Holtz, N. Janshen, A. Chica, T. Flower, C. Weidlich, T. Struckmann, State of charge monitoring of vanadium redox flow batteries using half cell potentials and electrolyte density, J. Power Sources 378 (2018) 776–783. [14] Z. Wei, K. J. Tseng, N. Wai, T. M. Lim, M. Skyllas-Kazacos, Adaptive estimation of state of charge and capacity with online identified battery model for vanadium redox flow battery, J. Power Sources 332 (2016) 389–398. [15] M. J. Watt-Smith, P. Ridley, R. Wills, A. Shah, F. Walsh, The importance of key operational variables and electrolyte monitoring to the performance of an all vanadium redox flow battery, J. Chem. Technol. Biotechnol. 88 (2013) 126–138. [16] V. Pop, H. J. Bergveld, D. Danilov, P. P. Regtien, P. H. Notten, State-of-the-art of battery state-of-charge determination, in: Battery Management Systems: Accurate State-of-Charge Indication for Battery-Powered Applications, Springer, 2008, pp. 11–45. [17] K. Ngamsai, A. Arpornwichanop, Measuring the state of charge of the electrolyte solution in a vanadium redox flow battery using a four-pole cell device, J. Power Sources 298 (2015) 150–157. [18] M. Mohamed, H. Ahmad, M. A. Seman, Estimating the state-of-charge of all-vanadium redox flow battery using a divided, open-circuit potentiometric cell, Elektron. Elektrotech. 19 (2013) 37–42.

24

[19] K. Knehr, E. Kumbur, Open circuit voltage of vanadium redox flow batteries: Discrepancy between models and experiments, Electrochem. Commun. 13 (2011) 342–345. [20] S. Corcuera, M. Skyllas-Kazacos, State-of-charge monitoring and electrolyte rebalancing methods for the vanadium redox flow battery, Eur. Chem. Bull. 1 (2012) 511–519. [21] C. P. De Leon, A. Frías-Ferrer, J. González-García, D. Szánto, F. C. Walsh, Redox flow cells for energy conversion, J. Power Sources 160 (2006) 716–732. [22] S. Eckroad, Vanadium redox flow batteries: an in-depth analysis, Electr. Power Res. Inst. (Palo Alto, Calif.) 1014836 (2007). [23] S. Kim, J. Yan, B. Schwenzer, J. Zhang, L. Li, J. Liu, Z. G. Yang, M. A. Hickner, Cycling performance and efficiency of sulfonated poly (sulfone) membranes in vanadium redox flow batteries, Electrochem. Commun. 12 (2010) 1650–1653. [24] H. Liu, Q. Xu, C. Yan, Y. Qiao, Corrosion behavior of a positive graphite electrode in vanadium redox flow battery, Electrochim. Acta 56 (2011) 8783–8790. [25] H. Cho, H. M. Krieg, J. A. Kerres, Performances of anion-exchange blend membranes on vanadium redox flow batteries, Membranes 9 (2019) 31. [26] J. N. Noack, L. Vorhauser, K. Pinkwart, J. Tuebke, Aging studies of vanadium redox flow batteries, ECS Trans. 33 (2011) 3–9. [27] F. J. Oldenburg, E. Nilsson, T. J. Schmidt, L. Gubler, Tackling capacity fading in vanadium redox flow batteries with amphoteric pbi/nafion bilayer membranes, ChemSusChem 12 (2019) 2620–2627. [28] T. Mohammadi, M. S. Kazacos, Modification of anion-exchange membranes for vanadium redox flow battery applications, J. Power Sources 63 (1996) 179–186. [29] R. M. Darling, A. Z. Weber, M. C. Tucker, M. L. Perry, The influence of electric field on crossover in redox-flow batteries, J. Electrochem. Soc. 163 (2016) A5014–A5022. [30] R. P. Brooker, C. J. Bell, L. J. Bonville, H. R. Kunz, J. M. Fenton, Determining vanadium concentrations using the uv-vis response method, J. Electrochem. Soc. 162 (2015) A608–A613. [31] X.-G. Yang, Q. Ye, P. Cheng, T. S. Zhao, Effects of the electric field on ion crossover in vanadium redox flow batteries, Appl. Energy 145 (2015) 306–319. [32] Q. Luo, L. Li, Z. Nie, W. Wang, X. Wei, B. Li, B. Chen, Z. Yang, In-situ investigation of vanadium ion transport in redox flow battery, J. Power Sources 218 (2012) 15–20. [33] M. Skyllas-Kazacos, L. Goh, Modeling of vanadium ion diffusion across the ion exchange membrane in the vanadium redox battery, J. Membr. Sci. 399 (2012) 43–48. [34] E. Agar, K. Knehr, D. Chen, M. A. Hickner, E. Kumbur, Species transport mechanisms governing capacity loss in vanadium flow batteries: Comparing nafion® and sulfonated radel membranes, Electrochim. Acta 98 (2013) 66–74.

25

[35] J. grosse Austing, C. N. Kirchner, L. Komsiyska, G. Wittstock, Investigation of crossover processes in a unitized bidirectional vanadium/air redox flow battery, J. Power Sources 306 (2016) 692–701. [36] C. Sun, J. Chen, H. Zhang, X. Han, Q. Luo, Investigations on transfer of water and vanadium ions across nafion membrane in an operating vanadium redox flow battery, J. Power Sources 195 (2010) 890–897.

26

Supporting Information For measuring OCP with the auxiliary flow cells, glassy carbon rods have been applied as working electrodes. The smooth and non-porous surface of the glassy carbon allows fast potential response in dependence on changes in the electrolyte during charging and discharging whereas the porous carbon felt retains electrolyte components with higher or lower state of charge than the electrolyte which passes the flow cell, these residues contribute to the potential measurement. Before employing the glassy carbon rods (SIGRADUR® G, Hochtemperatur-

Figure 12: Cyclic voltammograms (ν = 50 mV s−1 ) of glassy carbon rods in commercial vanadium electrolyte (GfE): fresh (V(III/IV)), discharged (V(III) and V(IV)) and charged (V(II) and V(V)) electrolyte.

Werkstoffe GmbH, 2 mm diameter) they have been polished with carbid paper (grid 2000), pretreated with an electrochemical polarisation routine (50 cycles with 500 mV s−1 , -0.9 – 1.7 27

V vs. NHE) and investigated by cyclic voltammetry (CV) (Potentiostat Gamry 1000) with a three electrode arrangement (reference electrode HgE11-S, Sensortechnik Meinsberg Xylem Analytics GmbH & Co. KG, counter electrode carbon felt SGL carbon) . Figure 12 shows cyclic voltammograms of a glassy carbon rod in fresh vanadium electrolyte (V(III/IV)) as well as in charged and discharged electrolytes of the NHC (V(II) and V(III)) and the PHC (V(IV) and V(V)) (GfE, total vanadium concentration of 1.6 M). The results reveal that the electrode material is suitable to indicate the oxidation state of vanadium and therefore for measuring OCP in the PHC and NHC electrolytes in dependence on the SOC of the battery.

28

DECHEMA Research Institute Electrochemistry Theodor-Heuss-Allee 25 D-60486 Frankfurt am Main Germany

December 30, 2019

Electrochimica Acta

Dear Professor Sergio Trasatti,

we want to extend our appreciation for taking the time and effort reviewing our manuscript. We are very pleased that Reviewer #3 witnesses that all previous questions are well addressed with reasonable justifications. The Reviewer also gave recommendations for references for the bibliography part: 1„ Evaluation of redox flow batteries goes beyond round-trip efficiency: A technical review, Journal of Energy Storage, Volume 16, April 2018, Pages 108-115. 2„ Thermally Stable Positive Electrolytes with a Superior Performance in All‐Vanadium Redox Flow Batteries, ChemPlusChem, Volume 80, February 2015, pages 354-358. We will comment on these recommendations in our reply to the reviewer, but we would also like to comment to you as the editor of the journal. The comments of Reviewers #1 and #2 have been valuable and helpful to the manuscript. The recommended references had been added to the manuscript for completion because they were concerned with the same subject as our work.

The literature recommendations of Reviewer #3 deal with interesting aspects of the research on vanadium redox flow batteries. However, we have not included any of them in our work since in our opinion their scopes present another research field compared to the topic we presented in our work. We will describe our concerns in detail in our response to the reviewer. Best regards Claudia Weidlich and Theresa Haisch

Reviewer Comments, Author Responses and Manuscript Changes Reviewer #3 #3 All questions are well addressed with reasonable justifications. I would like to give the authors some references for the bibliography part: Response: We are glad that the new additions have been well received and have been witnessed to satisfy questions and comments of Reviewer #1 and #2. Thank you for your recommendation of references.

Comment 1: Evaluation of redox flow batteries goes beyond round-trip efficiency: A technical review, Journal of Energy Storage, Volume 16, April 2018, Pages 108-115 Response 1: This is a very useful review about the basic necessity of proper evaluation criteria for redox flow batteries. We agree that the most meaningful assessment must include consideration of the entire system and the system energy efficiency is a very helpful indicator. But our work deals with another scope. The determination of the SOC has in our opinion a different scientific background compared to the calculation of efficiencies. In addition, we investigate the behavior of the half-cells individually to get information about crossover processes. So, we see no correlation or contradiction between these topics which is why we did not include this paper to our bibliography.

Comment 2: Thermally Stable Positive Electrolytes with a Superior Performance in All‐ Vanadium Redox Flow Batteries, ChemPlusChem, Volume 80, February 2015, pages 354-358

Response 2: This paper describes the use of an organic additive to the vanadium electrolyte which should prevent the precipitation of VO2+. It would be a useful innovation if the vanadium redox flow battery can also be operated at elevated temperatures. However, this aspect has no link to our topic. Of course it would be interesting to see whether the additive creates interactions that influence the SOC determination or the crossover, but we think this is the subject of more distant research. Therefore we also did not add the second paper to our bibliography. Kind regards Claudia Weidlich and Theresa Haisch

DECHEMA Research Institute Electrochemistry Theodor-Heuss-Allee 25 D-60486 Frankfurt am Main Germany

December 30, 2019

Reviewer #3 #3 All questions are well addressed with reasonable justifications. I would like to give the authors some references for the bibliography part: Response: We are glad that the new additions have been well received and have been witnessed to satisfy questions and comments of Reviewer #1 and #2. Thank you for your recommendation of references.

Comment 1: Evaluation of redox flow batteries goes beyond round-trip efficiency: A technical review, Journal of Energy Storage, Volume 16, April 2018, Pages 108-115 Response 1: This is a very useful review about the basic necessity of proper evaluation criteria for redox flow batteries. We agree that the most meaningful assessment must include consideration of the entire system and the system energy efficiency is a very helpful indicator. But our work deals with another scope. The determination of the SOC has in our opinion a different scientific background compared to the calculation of efficiencies. In addition, we investigate the behavior of the half-cells individually to get information about crossover processes. So, we see no correlation or contradiction between these topics which is why we did not include this paper to our bibliography.

Comment 2: Thermally Stable Positive Electrolytes with a Superior Performance in All‐ Vanadium Redox Flow Batteries, ChemPlusChem, Volume 80, February 2015, pages 354-358

Response 2: This paper describes the use of an organic additive to the vanadium electrolyte which should prevent the precipitation of VO2+. It would be a useful innovation if the vanadium redox flow battery can also be operated at elevated temperatures. However, this aspect has no link to our topic. Of course it would be interesting to see whether the additive creates interactions that influence the SOC determination or the crossover, but we think this is the subject of more distant research. Therefore we also did not add the second paper to our bibliography. Kind regards Claudia Weidlich and Theresa Haisch