A K-nearest Neighbours Inspired Direct MPC for SOC Balancing in Smart Batteries

Lithium-ion batteries dominate energy storage systems. Working circumstances and parameter variations in single cells can result in state of charge imbalances that reduce the battery’s lifetime. Modular batteries have been proven to be capable of actively reducing the imbalance among the cells by connecting or bypassing the individual batteries, sharing the work load in real-time. This paper presents a machine learning-inspired direct model predictive control for reconfigurable batteries to balance the cells that solves the underlying optimization problem in polynomial time.