Remaining useful life prediction of lithium-ion batteries based on peak interval features and deep learning

Abstract

Anticipating the lifespan of a lithium-ion battery is a challenging task due to the unstable external working conditions and complex internal response. To enhance the accuracy of the remaining usable life prediction of lithium-ion batteries, this study proposes a feature extraction method that relies on the peak value of the incremental capacity (IC) curve for a data-driven prediction approach. This method can significantly reduce the amount of data required for training the model. Furthermore, the grey correlation method (GCM) is employed to select features that have a high correlation with the target value, thereby further reducing the amount of data required for training the model. It is explored how well characteristics may be extracted from various peak intervals. The time series prediction-capable long short-term memory (LSTM) network is used to create the data-driven model. Finally, the experimental findings demonstrate that the proposed data-driven model's prediction error is smaller than 1.18 %. The proposed data-driven in this work is found to have a superior prediction impact on the same data set in different deep learning methods when compared to other data-driven approaches.