Attention is being paid to attempts at predicting the degradation and life of lithium-ion batteries (LIBs). This paper focuses on the examination conducted on the features, advantages, disadvantages, etc., of a datadriven prediction model that combines feature extraction and regression by deep learning. Also described is a physics-based model that predicts the degradation progress by electrochemical reaction formula and the like. As a result, it was found that in the physicsbased model, the prediction accuracy is high when the degradation phenomena are relatively straightforward, but its application is difficult when the phenomena are complicated or unknown. On the other hand, the data-driven modeling can be done even when the phenomena are not sufficiently clear and is considered to have a great advantage in predicting degradation accurately. Further consideration of the constructed model has also turned out to be useful for elucidating hidden phenomena.
展开▼
