Battery state of the charge estimation using Kalman filtering

Battery management system (BMS) requires an accurate prediction the remaining energy level or state of charge (SOC) of the cell or battery pack. However, in electric vehicles, batteries experience a dynamic operational environment whereby the simple algorithms employed in the portable devices to predict SOC, such as coulomb counting, are insufficient for this purpose. To address this problem, a Kalman filtering method is used to estimate the state of the charge of two different commercial lithium-ion batteries, with new physical insight being provided through an analysis of the Kalman filter covariance noise parameters. For example, the effect of geometry of the battery on value of these parameters is discussed. Different models are developed, tested and incorporated in the filter design. Subsequently, two types of Kalman filters including the extended Kalman filter and dual extended Kalman filter are implemented in order to predict the state of charge of the batteries. It is shown that the Kalman filtering can predict state of the charge of the battery with maximum 4% error.