On-board state of health estimation for lithium-ion batteries based on random forest

Abstract

Precise estimation of state of health (SOH) for lithium-ion batteries is critical to improve the state of charge (SOC) estimation precision, guarantee safety and improve the usage lifetime. In this paper, considering actual usage of lithiumion battery in electrical vehicles (EVs), a novel SOH estimation method is proposed based on the dataset of partial charging voltage under the constant-current (CC) mode and current value under the constant voltage (CV) mode. A random forest (RF) model is built to perform the battery SOH estimation, whose number of regression trees and number of input variables per node are optimized by the out of bag (OOB) estimation. In addition, the support vector machine (SVM) and least squares support vector machine (LSSVM) are employed for comparison of the SOH estimation accuracy. The results show that RF can achieve more accurate estimation of SOH, compared with those based on the SVM and LSSVM.