Fractional Order BPNN for Estimating State of Charge of Lithium-ion Battery under Temperature Influence

Abstract

Abstract State of charge (SOC) estimation for lithium-ion battery (LIB) is always a vital issue for battery management system (BMS) of LIBs. Due to the complex nonlinear characteristics of LIBs, data-driven model and estimation methods have been proposed. Among them, back propagation neural network (BPNN) is one of the widely used machine learning (ML) method. To enhance the performance of BPNN of LIBs, a fractional-order BPNN (FO BPNN) based on fractional-order gradient method is designed for SOC estimation of LIB in this paper. Moreover, temperature acting as key factor is also taken into consideration. Hence, the charging or discharging current, voltage, and temperature are applied as the inputs of the proposed FO BPNN, and SOC is obtained from the network. By Dynamic Stress Test (DST) experiments under five different temperatures of four 18650 LIBs, it proves that the proposed FO BPNN is able to estimate SOC of LIBs accurately in a data-driven way.