Advances in EKF SOC Estimation for LiPB HEV Battery Packs

This paper describes advances in models and methods used to estimate hybrid-electric-vehicle (HEV) battery-pack state-of-charge (SOC) using extended Kalman filtering (EKF). The electrochemical cells in the battery pack are Lithium Ion Polymer based, have a nominal capacity of about 7.5 Ah, and are optimized for power-needy applications. The discharge curve of these cells is very flat, which is desirable for some reasons, but also makes SOC estimation quite challenging and motivates the use of advanced methods. In earlier papers [1–2], we presented several cell models that may be used with the EKF method, and the EKF method of SOC estimation itself. Our best cell model was a “Radial Basis Function” (RBF) type, whose parameters were optimized by a “black-box” system identification procedure. We have subsequently found that although this model worked well on a cell level, it was too slow and exhibited unreliable performance when ported to the battery management system of the pack. In this paper, we describe a new cell model that alleviates these problems. The model parameters are optimized by “gray-box” system identification and most have direct physical interpretation. The model structure includes effects of: internal resistance, hysteresis, and relaxation time constants. It is greatly simplified with respect to the RBF model, allowing the SOC estimation algorithm to execute in about 1/50 of the time it did before. Results indicate that the new cell model with EKF provides SOC estimates that are about as precise as those made with RBF, but are more accurate and reliable. Copyright 2003 EVS20