A Data-Driven Method for Lithium-Ion Batteries Remaining Useful Life Prediction Based on Optimal Hyperparameters

For enhancing the prediction accuracy of remaining useful life (RUL) of lithium-ion batteries (LIBs), an innovative LIBs' RUL prediction framework based on whale optimization algorithm (WOA) and long-short term memory (LSTM) is proposed. The validity and exactitude are proved by NASA batteries data set. The WOA has faster convergence speed and higher convergence accuracy than Particle Swarm Optimization (PSO) and Artificial Bee Colony (ABC) algorithms. Explanatory results manifest the proposed WOA-LSTM model can markedly improve the prediction accuracy of RUL, which solves the problem that hyperparameter settings have widely different effects on LSTM prediction. The RMSE and MAE are basically less than 0.03 and 0.02, respectively. It can be selected as a recommended model for accurately predicting the RUL of LIBs.