LiPB Dynamic Cell Models for Kalman-Filter SOC Estimation

SOC Estimation Gregory L. Plett, consultant to Compact Power Inc., and Assistant Professor, University of Colorado at Colorado Springs Abstract hHEV environment harsh: Rates up to ±25C, very dynamic rate profiles. hVery different from low-rate / constant-rate portable electronics. hSOC estimation must be done differently — if precise SOC estimation is required by the HEV, then a very accurate cell model is necessary. hSeveral different cell models are proposed. All may be used to predict SOC using an extended Kalman filter. Extended Kalman Filter hState estimator for nonlinear dynamic system. System model must have form: xk+1 = f (xk , uk ) + wk yk = g(xk ,uk ) + vk where xk is the system “state”, uk is the input, wk is process noise, vk is sensor noise, f ( ) and g( ) are known (possibly nonlinear) functions. hKalman filter recursively estimates system state with plus / minus error bounds. Models hInputs to models includes cell current; output is (loaded) terminal voltage.