Remaining Capacity Estimation of Lithium-ion Batteries based on Health Features Extraction and Gray Relation Analysis

Abstract

Lithium-ion battery remaining capacity estimation mainly adopts the model or data-driven method combined with feature extraction. In contemplation of deal with the issues of incomplete feature extraction procedure and poor estimation accuracy of extracted features, a data-driven lithium-ion battery remaining capacity estimation structure is suggested. To begin with, the charge and discharge data are fitted, time series analysis and frequency domain analysis are carried out to extract a set of health features. Then screen out features with high relation by gray relation analysis. Finally, the screened features are adopted as input to train a support vector regression model for estimating the lithium-ion batteries remaining capacity. Test and verify the proposed method on of NASA and CACLE lithium-ion battery cycle fading datasets, and the experimental results show the capability and superiority of the method.