Review of the remaining useful life prediction methods for lithium-ion batteries
As the energy and power density of lithium-ion batteries have gradually increased in recent years, the safety performance and prediction of remaining service life have become increasingly crucial. This review offers a comprehensive analysis of the current research status of predicting the remaining useful life of lithium batteries. It systematically introduces the existing forecast algorithms, focusing on the application of machine learning methods in this field. Model-based methods encompass electrochemical, equivalent circuits, and empirical models. In contrast, data-driven methods involve machine learning techniques such as support vector machines, Gaussian process regression, extreme learning machines, convolutional neural networks, recurrent neural networks, and transformers. We meticulously examine the advantages and disadvantages of each method, emphasizing on the application and evolution of machine learning methods in feature extraction and fusion techniques. This study summarizes and analyzes current-voltage-temperature, IC, and EIS curves regarding feature extraction. It subdivides and analyzes fusion methods into model-model, data-model, and data-data fusion methods. Finally, addressing the existing research challenges, this review proposes research suggestions for predicting remaining service life from three perspectives: early, online, and multioperating condition predictions. These suggestions provide insights into enhancing the accuracy and practicability of remaining service life prediction algorithms for Li-ion batteries.
万方数据
维普
