Membrane evaluation and thermal modelling of the vanadium redox flow battery

Abstract: In this study, different membranes were tested for their use in a vanadium redox flow battery (VRB). The membranes were first tested under charge-discharge operation of the VRB over four charge-discharge cycles at different current densities (20 mAcm-2, 40 mA cm-2 and 60 mAcm-2) and the corresponding cell performance was evaluated in terms of total energy efficiency (EE), coulombic efficiency (CE) and voltage efficiency (VE). The anion exchange Fumasep® series FAP-450 as well as the cation exchange series, Fumapem®, F-930 by Fumatech™ (located in Baden-Württemberg, Germany) and the GN-115, GN-212, GN-212C (all from General Energy and New Materials Co Ltd,. Nanjing, China) and the VB1 (supplied by V-Fuel Pty Ltd) were all evaluated during charge-discharge operation. It was found that the F930 membrane outperformed all other membranes at all current densities tested in this study with a peak average EE of 89,6% at 500 mA (20 mAcm-2) and CE of 98,8% at 1500 mA (60 mAcm-2). The second best at charge-discharge evaluation was the FAP450 with an overall EE of 81,0%. In order to further quantify any differences between the membranes, these membranes were tested for permeability rates at constant temperature for all vanadium ions present in a VRB without influence of any current. The diffusion coefficients through the FAP-450 and the F930, VB2 (same as VB1 but thicker) and the GN-115 membrane were determined for the following ions V(II), V(III), V(IV) and V(V). It was found that the VB2 had the lowest measured permeability rates for the V(II), V(III) and V(IV) ions. Furthermore, the FAP450 membrane was the membrane with the second best diffusion coefficients which was consistent with the high coulombic efficiencies observed in the laboratory scale VRB cell. To predict the overall capacity loss during long term operation, the battery was simulated in MATLAB™ using mathematical models developed at UNSW Australia for 60 cycles (without electrolyte remixing). It was found that the FAP450 performed better than the F930 and the GN-115 membrane based on experimentally determined diffusion coefficients. The self-discharge reactions led to almost 0% capacity loss after about 60 cycles for the FAP450 membrane. For the F930, which had about 10 times higher diffusion coefficients compared to the FAP450, about 90% capacity was lost over the same number of cycles. The GN-115 membrane, which had higher diffusion coefficients than the F930 gave better results at simulations and had lost about 50% of the capacity after 60 cycles. This is believed to be due to the F930 having a factor of 20 difference between the diffusion coefficients which leads to a buildup of vanadium ions on one side of the cell and a deficit of the other. As the F930 had overall smaller diffusion coefficients compared to the GN115, it was unexpected that the latter would prove better than the F930 at extended use. It can thus be concluded that not only do the diffusion coefficients have to be low, they also have to be the same order of magnitude for all the different ions to prevent severe capacity drop. As the vanadium electrolytes may irreversibly precipitate at low respectively high temperatures it is imperative to understand how non-electrochemical, exothermic side reactions as well as surrounding temperatures interact with the electrolytes. Therefore, thermal modelling was also undertaken using the MATLAB model developed at UNSW. It was found that higher permeability rates gave much higher temperatures in the cell stacks and the electrolyte tanks. After 5 days, the electrolyte temperature in the tanks of the VRB using the F930 was about 30°C and increasing while the stacks reached temperatures of as high as 38°C and increasing. The FAP450, which had lower permeability rates than the F930, reached a stack temperature of only a few more degrees than the tank temperature due to the low influence of exothermic self-discharge reactions inside the cell stack. By comparison, previous thermodynamic simulation studies using Nafion showed that the temperature of the stack increased to about 40°C which indicates that external cooling should be considered when the pumps are turned off and the battery is at standby. The GN-115 membrane from General Energy © gave good results in the VRB cycling tests as well as for permeability rates and was therefore further evaluated by immersion in 1 M V(5), 2,5 M H2SO4 for 7 weeks in order to test its chemical resistance to the oxidizing V(5) solution. It displayed a 26% increase in thickness and about 6% increase in length and width. The weight increased by about 10%. This indicates that the pore size of the membrane might have changed during immersion which could influence the performance of the membrane in VRBs.