Temperature Dependent State of Charge Estimation of Lithium-ion Batteries Using Long Short-Term Memory Network and Kalman Filter

In this paper, a new state of charge (SOC) estimation method is proposed combining Long Short-Term Memory Network (LSTMN) battery model and Kalman filter (KF) considering temperature dependency. The technique has been compared to the equivalent circuit model (ECM) and the combined model (CM) using dynamic stress test data. A KF is applied to each model to realize the dynamic estimation of battery states. Based on the collected data from the federal urban driving schedule, terminal voltage approximation and SOC estimation are carried out, and the results are compared among the models. This paper includes the following contributions: (1). A LSTMN battery model that shows stronger robustness against temperature is implemented. (2). A LSTMN-KF method is proposed for SOC estimation at different temperatures and is compared with ECM-KF method and CM-KF method. (3). The proposed method eliminates the need for SOC-OCV lookup table and does not rely on the chemical characteristics of batteries.