锂离子电池状态估计机器学习方法综述

This paper aims to give a comprehensive review on the research progress in the field of the estimation of the states of lithium-ion battery, including the state of charge (SOC), state of health (SOH) and residual useful life (RUL). Firstly, the application status of machine learning method to the estimation of battery states are expounded. Then, five specific implemental links of machine learning methods for battery state estimation are summarized, including data preparation, model selection and evaluation, hyperparameter determination, data preprocessing and model training, and an evaluation method of learning algorithms is proposed in terms of fusion accuracy, implementation cost and robustness. Finally, the problem of scene adaptability in determining hyperparameters is pointed out, with a suggestion put forward: establishing multi-regional, cross-seasonal, multi-mode and long-term driving cycle database of traction battery, so as to promote the research on the practicability and universality of machine learning algorithms for battery state estimation.