Capacity estimation of lithium-ion battery based on charging voltage and GAF-LSTM

Abstract

ABSTRACT Accurate estimation of the capacity of lithium-ion battery is crutial for the health monitoring and safe operation of electronic equipment. However, it is difficult to ensure a complete charge-discharge cycle because of the randomness of the battery working state under actual working conditions. Insufficient capture of feature information will also lead to low prediction accuracy of the model. Aiming at the above problems, a method for estimating the capacity of lithium-ion battery based on charging voltage, Gramian Angular Fields (GAF) and Long Short-Term Memory Network (LSTM) is proposed. To make up for the randomness of battery data cycle, this method selects a fixed voltage segment from the charging process and extracts three characteristic parameters closely related to capacity decline. GAF is used to transform one-dimensional feature parameters into two-dimensional image structure, and the neglected hidden features are fully mined. It emphasizes the global correlation between features and enhances the ability of feature expression. Then, the feature map is input into the LSTM network for training and capacity estimation. This method is verified on two public datasets. The results show that this method has achieved good prediction effect and proved its superiority.