Remaining Useful Life and State of Health Assessment for Lithium Ion Batteries Using CNN-BILSTM-DNN Hybrid Method

Abstract

Accurate prediction of Remaining Useful Life (RUL) and State of Health (SOH) of lithium-ion batteries play an increasingly crucial role in intelligent battery health management systems. It also serves as a battery failure early warning system. For electrical vehicles, lithium-ion batteries serve as the primary energy source. Li-ion battery safety requires the use of a battery management system (BMS), which typically rests on RUL and SOH. This work suggests a hybrid method, consisting of a Convolutional Neural Network (CNN), Bidirectional Long Short-Term Memory (BiLSTM), and Deep Neural Network (DNN), to estimate the remaining useful life (RUL) and state of health (SOH) of the battery. A comparative analysis has been done with another existing hybrid method consisting of a Convolutional Neural Network, Long Short-Term Memory, and Deep Neural Network. Three analytical indices are chosen to evaluate the prediction results numerically. They are MAE, R2, and RMSE. The suggested method is experimented with and validated on the NASA lithium-ion battery health dataset. When compared with the existing method, it is observed that the suggested technique has greater accuracy.