State of charge estimation of lithium-ion batteries using improved BP neural network and filtering techniques

Abstract The state of charge (SOC) is a critical parameter in the battery management system (BMS), and its accurate estimation essential for ensuring safety reliability batteries. This paper presents lithium-ion SOC method that combines an improved neural network with filtering algorithm. Firstly, backpropagation (BP) algorithm chosen as architecture hybrid due to strong nonlinear approximation ability, particle swarm optimization (PSO) used optimize it avoid falling into local optimal solutions. By combining search ability PSO learning BP network, accuracy model improved. proposed integrates PSO-BP extended Kalman filter based on minimum error entropy (MEE-EKF). utilized measurement equation MEE-EKF, while ampere-hour integration employed achieve closed-loop estimation. Finally, experimental validation conducted under four typical operating conditions one random condition across wide temperature range. results demonstrate achieves high all compared other algorithms, maximum absolute not exceeding 3.13%, mean less than 0.54%, root square no more 0.66%.