Remaining Useful Life Prediction for Lithium-ion Batteries Based on Health Indicators and Hybrid Bi-LSTM-NAR Model

In order to accurately predict the remaining useful life of lithium-ion batteries and reduce the risk of battery operations, a novel model was proposed for online remaining useful life prediction of lithium-ion batteries. On the basis of historical operation data of lithium-ion batteries, six types of health indicators were extracted to characterize the degradation of batteries. The random for-cst(RF) algorithm was adopted to evaluate and screen the health indicators. The generalized regression neural network(GA-GRNN), which was optimized by genetic algorithm, was used to estimate the residual capacity of the battery. Then, a hybrid model combining bidirectional long short-term memory(Bi-LSTM) network model and nonlinear autoregressive (NAR) neural network (hybrid Bi-LSTM-NAR model) was used to predict the remaining useful life for lithium-ion batteries. A case study was conducted with the NASA open data. The results show that by way of screening the indicators, the accuracy of capacity estimation and remaining useful life prediction of lithium-ion batteries arc ensured. Compared with the prediction results of existing methods, the prediction accuracy of the proposed hybrid prediction model is improved effectively. © 2024 Chinese Mechanical Engineering Society. All rights reserved.