This paper proposes a state of charge (SOC) estimation model that combines data-driven method with model-based filtering method. Firstly, an improved arithmetic optimization algorithm (AOA) is employed to optimize the initial values of the long short-term memory (LSTM) network, and the optimized LSTM network is utilized for the preliminary estimation of SOC. Then, an adaptive unscented Kalman filter (AUKF) is employed to correct the SOC estimation results. Experimental results demonstrate that the proposed model achieves accurate and smooth SOC estimation while being able to quickly respond to initial SOC errors.
Keywords State of charge, Long short-term memory network, Arithmetic optimization algorithm, Adaptive unscented Kalman filter.