An Adaptive OCV-SOC Curve Selection Classifier for Battery State-of-Charge Estimation

Abstract

The State-of-Charge (SOC) estimation of lithium-ion batteries is crucial in battery management systems (BMS) for energy storage power stations. The open-circuit voltage (OCV)-SOC curve and the SOC estimation algorithm are important in SOC estimation. There exist two common OCV tests, including the low current OCV test and the incremental OCV test, for OCV-SOC curve acquirement, whose performances vary with different working conditions under $25^{\circ}{C}$. To make SOC estimation more effective, the least squares support vector machines (LS-SVM) method is presented as an adaptive classifier to decide which OCV test should be applied. Besides, an adaptive square-root unscented Kalman filter (ASRUKF) algorithm is proposed to improve square-root unscented Kalman filter (SRUKF) algorithm by updating the noise covariance matrixes in real-time. Based on the Center for Advanced Life Cycle Engineering (CALCE) public data set of University of Maryland of the 18650 LFP battery under $0^{\circ}{C}, 25^{\circ}{C}$ and $45^{\circ}{C}$, the SOC estimation algorithm based on adaptive OCV-SOC curve selection classifier is demonstrated to be precise, quick, robust and adaptive.