AI-based state-of-charge estimation

Abstract

Neural networks have been widely used for image recognition and result prediction based on input data. Compared with other SOC estimation methods, the neural network method does not need to accurately consider the electrochemical state inside the battery to estimate the SOC by self-learning capability. With the continuous development of artificial intelligence, neural network methods have received increasing attention from researchers in the battery field. However, few articles summarize neural network methods individually. This review examines recent articles that use neural network methods alone to estimate the SOC of lithium-ion batteries, dividing the methods into FFNN method, deep learning method, and hybrid method. Then, the neural network method is summarized and discussed from the aspects of network structure, algorithm principle, appropriate environment, advantages, disadvantages, and estimation errors. The challenges to neural network use in the EV field are discussed briefly, and the prospects and opportunities for neural networks are examined in closing.