Sparrow Search Algorithm-Optimized Long Short-Term Memory Neural Network for Predicting the Remaining Service Life of Lithium Batteries

The safety and reliability of lithium batteries in fields such as new energy vehicles and unmanned aerial vehicles (UAVs) depend on accurate remaining service life prediction. However, the nonlinear degradation process of batteries is complex, and the long short-term memory (LSTM) neural network model limits its prediction accuracy due to the limitation of hyperparameter selection. To this end, a hybrid model (SSA-LSTM) based on the sparrow search algorithm (SSA) to optimize the hyperparameters of LSTM is proposed. The model adjusts the number of neurons in the hidden layer and the learning rate through automatic optimization search, avoids the blindness of manually selecting the hyperparameters, and effectively improves the model performance. Based on the lithium battery dataset made public by CALCE, the experimental results show that the SSA-LSTM model outperforms the traditional LSTM and PSO-LSTM models in terms of RMSE, MAE, and MAPE. The prediction accuracy is improved by more than 9 % and 5 %, respectively, demonstrating a stronger accuracy and suitable for complex battery life prediction tasks.