An Improved Graph-CNN Based State of Health Estimation for Lithium-Ion Batteries

State of health (SOH) plays a vital role in the safety, reliability and lifetime of lithium-ion batteries (LIBs). In this paper, a novel SOH estimation method is proposed based on graph convolutional neural network (CNN). First, this method considers the correlation between health features (HFs) and SOH. Second, the selected HFs are connected to form a graph. Third, graph-CNN extracts key features from the graph and performs SOH estimation. The results demonstrate that the proposed SOH estimation method can enhance the estimation accuracy and reduce the computational load. Compared with other deep learning methods, the proposed graph CNN method has high accuracy with robustness against different conditions with the average RMSE below 0.8% on the CALCE dataset.