State of health (SOH) estimation of lithium-ion (Li-ion) batteries is of great importance for battery use, maintenance, management, and economic evaluation. However, the current SOH estimation methods for Li-ion batteries are mainly targeted at specific charging strategies by using deterministic estimation models, which cannot reflect uncertain information such as randomness and fuzziness in the battery degradation process. To this end, a method for estimating the SOH interval of Li-ion batteries applicable to different charging strategies is proposed, which extracts multiple feature parameters from the cyclic charging and discharging data of batteries with different charging strategies, and automatically selects the optimal combination of feature parameters for a specific charging strategy by using the cross-validation method. In addition, considering the limited number of cycles in the whole life cycle of Li-ion batteries as a small sample, support vector quantile regression (SVQR), which integrates the advantages of support vector regression and quantile regression, is proposed for the estimation of SOH interval of lithium-ion batteries. Li-ion battery charge/discharge cycle data with deep discharge degree is selected as the training set for offline training of the SVQR model, and the trained model is used for online estimation of the SOH of Li-ion batteries of different charging strategies. The proposed method is validated using three datasets with different charging strategies. The experimental results show that the proposed method is applicable to different charging strategies and the estimation results are better than those of quantile regression, quantile regression neural network and Gaussian process regression.
