Vanadium redox flow batteries (VRFBs) are increasingly deployed for grid-scale energy storage due to their long duration, scalability, and ability to decouple power and capacity ratings. However, multi-unit VRFB stations face a persistent challenge: state-of-charge (SOC) inconsistency across individual units triggers voltage violations and forces premature charge or discharge cutoffs, reducing overall station efficiency and dispatch performance.
Researchers have now proposed a unified control framework that addresses this problem through coordinated model predictive control (MPC) integrated with neural network surrogates and advanced bound-tightening algorithms. The strategy operates at the station level and combines three key innovations: optimized open-circuit-voltage (OCV) reference selection to operate in high-sensitivity regions, real-time SOC equalization across units, and computational efficiency through hybrid feasibility-based and optimization-based bound tightening (FBBT + OBBT).
Rather than using expensive nonlinear electrochemical models in the online MPC computation, the team replaced them with trained ReLU neural surrogates embedded via Big-M constraints. This reduces computational burden while maintaining safety. The hybrid bound-tightening scheme improved neuron bounds by up to 59% and reduced preprocessing time by 14% compared to optimization-based methods alone.
Simulation results on a 100 MWh dispatch scenario demonstrate tangible benefits: charging time decreased by roughly 10%, voltage efficiency improved from 75.6% to 80%, and energy efficiency rose from 71.2% to 74.6%. Station-level switching actions also reduced from 31 to 29 cycles, indicating smoother operations and lower wear.
These gains translate to extended battery lifetime, lower operational costs, and improved grid reliability. As energy storage becomes essential to integrating renewable generation, control strategies that maximize efficiency and safety in multi-unit systems will be increasingly valuable for operators and utilities seeking to optimize their storage investments.
