ABSTRACT State of charge (SOC) is a critical performance indicator for battery operation, and its precise estimation is essential for ensuring the safe operation of the battery system. This paper introduces a novel cascaded hybrid SOC estimation model that integrates an adaptive extended Kalman filter (AEKF) and a long short‐term memory (LSTM) network. The model employs RC circuit configuration and utilises a forgetting factor recursive least squares algorithm for parameter identification. Initially, the AEKF is used to derive SOC estimates from the circuit model. Subsequently, an LSTM network corrects the errors in these initial SOC estimates, resulting in improved accuracy. The paper provides a detailed model description and validates it across various operating conditions. Experimental results demonstrate that this model offers outstanding estimation accuracy and generalisation performance, with a root mean square error maintained within 0.34%, a maximum error within 2.10%, and a mean absolute error within 0.23%.
Yang et al. (Wed,) studied this question.