Multicycle SOH prediction for lithium batteries based on CRQA-driven dynamic features

To address the insufficient modelling of cross-cycle dynamic correlations, this article introduces Cross-Recurrence Quantification Analysis (CRQA) as a feature extraction method for lithium-ion battery State of Health (SOH) assessment. By constructing cross-cycle cross-recurrence plots for multivariate charge-discharge signals (current, voltage, temperature), dynamic similarity statistical features are extracted to model the State of Health variation between adjacent cycles, overcoming the reliance of traditional methods on single-cycle static features. Furthermore, Spearman's rho is used to calculate the correlation between cross-cycle features and the change in SOH of lithium-ion battery, which allows us to choose the appropriate offset parametera. At last, predictive models are constructed using Random Forest, Support Vector Machine, and Back Propagation Neural Network based on the NASA battery aging datasets. Experimental results validate that this study proposes a new dynamic modeling framework for lithium-ion battery SOH assessment via revealing the nonlinear dynamic correlations inherent in charge-discharge signals.