Comparative Analysis of Algorithms for Predicting the Remaining Useful Life of Lithium-Ion Batteries

Lithium-ion batteries are widely used in portable devices, electric vehicles, and large-scale energy storage systems. Predicting remaining useful life (RUL) of these batteries is crucial for ensuring the reliability and safe operation of battery-powered systems. This study aims to identify the most effective method for prediction RUL by evaluating predictions derived form the CALCE dataset. We compare the performance of Bi-LSTM, a deep learning algorithm, with model-based approaches (Kalman filter) and data-driven methods (Support Vector Machine: SVM), Recurrent Neural Networks (RNN), and Long Short-Term Memory (LSTM). Data processing will vary based on the type of non-linearity being analyzed. Model performance will be assessed using validation metrics such as RMSE, MAE, and relative error. The evaluation indicates that while other models enhance the temporal dependence of battery data, Bi-LSTM consistently outperforms them in accuracy and in capturing the temporal and nonlinear relationships within the data. This research is significant for identifying effective machine learning strategies to enhance the performance and reliability of lithium-ion battery management systems (IBMS).