Hybrid CC-LSTM Algorithm for SOC Estimation and Its Dual-Mode Hardware Design

Coulomb counting method (CC) and Long ShortTerm Memory (LSTM) network are widely used for battery State of Charge (SOC) estimation. This paper develops a hybrid CC-LSTM algorithm for SOC estimation and designs a dual-mode hardware architecture based on this algorithm. Additionally, a test platform based on a field-programmable gate array (FPGA) is built to validate the proposed hardware. In the proposed CC-LSTM model, CC is employed for feature extraction, while LSTM is used for SOC estimation. In the proposed hardware design, CC-LSTM is utilized in dynamic mode, whereas the open circuit voltage method is used in sleep mode. The proposed design achieves an execution time of 2.86ms on FPGA, with an average root mean square error (RMSE) of 0.912% under various operating conditions.