Data-Driven State-of-Health Estimation of Li-ion Batteries Based on Cross Recurrence Plot

State-of-health (SOH) is an important indicator for evaluating the degradation and security of Lithium-ion batteries. This paper proposed a new online battery SOH estimation method based on cross recurrence plots (CRP) and the convolutional neural network (CNN) regression model that is adaptive to different battery chemistries and robust to data noise. The Region of interest (ROI) is selected as the peak of the incremental capacity curve and requires a narrow voltage range of 0.1V that is suitable for online application. The two-dimensional health features based on CRPs are generated by combining the initial reference trajectories from training data and online trajectories to comprehensively evaluate the degradation of batteries. The proposed method was tested on the MIT-Stanford and Oxford datasets and showed a high accuracy of RMSE of less than 0.8% and MaxAE of less than 1.9% and presents a significant improvement compared to the other two benchmark methods. © 2024 IEEE.