Online state-of-charge estimation for lithium-ion batteries considering model inaccuracies under time-varying current conditions

Abstract

Load currents of batteries undergo real-time change depending on the operations of battery-powered devices, such as acceleration, deceleration, and sudden braking of electric vehicles. With the variation in current magnitude, parameters in an equivalent circuit model of the battery change depending on its chemical characteristics. Therefore, accurately identifying model parameters offline for various sizes and patterns of load currents with few experiments is difficult. Besides, there is a model uncertainty due to the difference between the linearity of the model structure and the nonlinearity of the real battery; hence, an experimentally obtained open-circuit voltage (OCV)-state-of-charge (SOC) curve has difficulty in providing an accurate OCV value for the model. Consequently, these difficulties lead to inaccuracies in the model and SOC estimation. In this article, novel methods are proposed to reduce these inaccuracies under time-varying current conditions, where the magnitude of the load current changes in real time. These are (1) an OCV-SOC curve estimation for reducing the model uncertainty and (2) a current-adaptive extended Kalman filter algorithm for decreasing the effect of inaccurately identified model parameters. The experimental results show that the maximum-absolute-errors of the proposed SOC estimation were reduced to <1.5%. Consequently, battery-powered applications can obtain highly accurate SOC values even the magnitude of the load current changes in real time.