Hybrid observer design for online battery state-of-charge estimation

Abstract

This paper presents a method for estimating battery state-of-charge based on linear hybrid estimation theory. Contrasting with existing battery estimation techniques, hybrid linear models permit systematic introduction of additional dynamics online, which can capture cell aging or temperature effects. Piecewise regression with free-knots was employed to segment the nonlinear model into linear regions, enabling a decoupled hybrid observer scheme to independently identify the continuous and discrete hybrid states. Simulation and experimental studies demonstrate the observer efficacy.