LAUSR

Accurately estimating the capacity of lithium-ion batteries in electric vehicles (EVs) is critical for making correct management decisions. However, the randomness of the charging voltage range of EVs can lead to missing observations or reduced accuracy of capacity estimation methods. This article proposes an adaptive battery capacity estimation method suitable for arbitrary charging voltage range based on incremental capacity (IC) analysis and data-driven techniques. All charging conditions of EVs are divided into three categories according to the charging voltage range. Three data-driven estimation submethods with sequential application priority are designed for the three charging conditions separately, including back-propagation neural network with IC peak coordinates as input, ensemble learning with local high IC curve as input, and linear regression with ampere-hour coordinate transformation. The method is based on a priori knowledge to select a suitable estimation submethod under different charging conditions, so as to improve the adaptability. Experimental data is collected from eight commercial lithium-ion battery modules for model establishment and verification. Over 250 000 experimental samples at different states of health and random charging ranges show that the method can accurately estimate battery capacity under arbitrary charging conditions, with a maximum error of 2%.