State-of-charge (SOC) estimation based on a reduced order electrochemical thermal model and extended Kalman filter

Accurate estimation of SOC of a battery is one of the most important issues to prevent the battery from being overcharged and undercharged. Currently, SOC is estimated using Coulomb counting or a stored set of data that includes a relationship between open circuit voltage (OCV) and SOC, where the OCV for a cell in operation is estimated based on the first or second order of electrical equivalent circuit model (EECM). The typical error percentage of those methods range 5-10% because of the integration error of the current over time and ignored effects in the EECM such as lithium ion concentrations, temperature, and other electrochemical phenomena. A new method for estimation of SOC is developed that uses a reduced order model (ROM) derived from a quasi-three dimensional full order physical model (FOM), and Extended Kalman filter (EKF) to minimize errors caused by inaccuracy of the ROM. Static and dynamic responses of the new approach at different charging and discharging current in a single cycle and multiple cycles are compared with those of Coulomb counting, ROM and ROM with EKF. The results show that the accomplished average error is approximately less than 4%.