Prediction of State of Charge in Electric Vehicle Lithium-ion Batteries using Deep Learning

The lithium battery state of charge (SOC) is a crucial state parameter in the battery management system. It serves as a vital reference for the safety management of electric vehicle batteries. To investigate the application of deep learning in assessing the state of charge (SOC) of electric vehicle power batteries. The precision of state-of-charge (SOC) estimation will impact the efficiency of the battery management system. Greater accuracy leads to improved performance. The initial step was a comprehensive examination of the precise technical procedures employed in the deep learning approach for SOC estimation. The processes included in this research were gathering data, preprocessing of data, training of models, and model evaluation. Furthermore, a summary was provided for the existing lithium battery dataset that is widely utilised by the public. A strategy is proposed to enhance the precision of SOC estimate by utilising a particle swarm optimisation (PSO) approach to optimise the long short term memory. The PSO algorithm is first employed to develop a PSO-Long Short Term Memory (LSTM) estimation model, which allows for the determination of the most effective iteration parameters for the model. Subsequently, the PSO-LSTM estimation model is simulated across several operational situations, encompassing Dynamic Stress Test and US06. The voltage, current, and discharge data of the lithium-ion battery are inputted into the PSO-LSTM neural network model to be compared with the LSTM algorithm. Performance metrics such as RMSE (Root Mean Square Error), MAE (Mean Absolute Error), and MAPE (Mean Absolute Percentage Error) are used for evaluation. The results suggest that the optimised PSO-LSTM algorithm model achieves an estimation accuracy of 1.5%. The accuracy has been improved.