Capacity Prediction for Lithium-Ion Batteries Using Different Neural Networks Methods

For the purpose of assuring reliability and safety, the lifespan of a lithium-ion battery must be accurately predicted. Additionally, it serves as a mechanism of early warning to prevent the battery’s failure. New data-driven estimation techniques are made possible by recent advances in machine learning (ML). In this study, we propose two models that make use of the Long Short-Term Memory LSTM (LSTM) and Gated recurrent units (GRU) algorithms, respectively, to estimate the battery’s capacity and increase prediction accuracy. To demonstrate the superiority of the suggested GRU-based approach, a comparison against several ML estimation algorithms is made. Two statistical indicators, the MAE and RMSE, are used in order to carefully evaluate the accuracy of the prediction. Utilizing datasets of various lithium-ion batteries from NASA, experimental validation is carried out. The GRU technique performs better than the others, according to the results, and the suggested method is also effective at decreasing prediction error and enhancing forecasting performance.