Cost, performance prediction and optimization of a vanadium flow battery by machine-learning

Abstract

Performance optimization and cost reduction of a vanadium flow battery (VFB) system is essential for its commercialization and application in large-scale energy storage. However, developing a VFB stack from lab to industrial scale can take years of experiments due to the influence of complex factors, from key materials to the battery architecture. Herein, we have developed an innovative machine learning (ML) methodology to optimize and predict the efficiencies and costs of VFBs with extreme accuracy, based on our database of over 100 stacks with varying power rates. The results indicated that the cost of a VFB system (S-cost) at energy/power (E/P) = 4 h can reach around 223 $ (kW h)⁻¹, when the operating current density reaches 200 mA cm⁻², while the voltage efficiency (VE) and utilization ratio of the electrolyte (UE) are maintained above 90% and 80%, respectively. This work highlights the potential of the ML methodology to guide stack design and optimization of flow batteries to further accelerate their commercialization.