A SOC Estimation Method for Lithium-ion Batteries Based on Optuna optimized Long Short-Term Memory Neural Network

Estimating a battery's state of charge (SOC) accurately is essential for determining an electric vehicle's safety and remaining range. In this paper, a battery SOC (state of charge) estimation method based on Optuna optimised LSTM is proposed. In response to the problem of low efficiency and difficulty in achieving global optimization through manual parameter tuning in traditional methods, this paper applies the Optuna automated hyperparameter tuning framework and combines it with a random search algorithm to optimize key hyperparameters in the LSTM model. The research results indicate that the optimal setting (Look back: 43, Epoch: 175, Batch size: 65, Learning rate: 0.0029) achieved the best estimation accuracy. Finally, the performance of the model was validated on two different lithiumion battery datasets. The experimental results showed that the LSTM model based on Optuna optimization exhibited good generalization ability and high accuracy under different temperature and dynamic load conditions, significantly improving the reliability and accuracy of SOC estimation.