State of health estimation for Lithium-ion battery based on fusion features

An accurate battery state-of-health can effectively avoid the safety problems caused by lithium-ion batteries. In this paper, a novel feature extraction method based on autoencoder is proposed. First, measured features as well as calculated features are extracted from the battery current and voltage parameters as the initial feature set. Then the parameters which has low correlation to battery SOH in the initial feature set are removed. Finally, the remaining features are fused with an autoencoder to obtain the target feature set. The proposed feature extraction method is validated in a public dataset, and the results shows that the proposed method improves the SOH estimation accuracy by at least 26% compared to the current feature extraction method.