Historical data-driven approach for energy storage system state of charge estimation

State of charge (SOC) is a key parameter in lithium battery management technology, and accurate estimation of the SOC of an energy storage system facilitates its grid-connected operation and control. In this paper, a historical data-driven SOC estimation method is proposed to analyse the relationship between the storage system voltage, current, average temperature and other factors on the SOC, and to obtain the predicted value of the SOC. In this paper, a bidirectional long and short-term memory multi-head neural network is used for prediction, and particle swarm optimization algorithm (PSO) is used to find the optimal of its parameters, which makes the predicted SOC of energy storage system closer to the real value compared with other models, with a better training effect, and the performance is better than that of the traditional single neural network model, and improves the accuracy of prediction.