A Multitimescale Kalman Filter-Based Estimator of Li-Ion Battery Parameters Including Adaptive Coupling of State-of-Charge and Capacity Estimation

Abstract

This article deals with coupled, state, and parameter estimation for lithium-ion batteries described by an equivalent circuit model, including polarization dynamics. Since the model parameters depend on the battery state-of-charge (SoC) and temperature operating point, as well as on the battery state-of-health, all states and parameters need to be estimated simultaneously for an accurate overall estimation during the battery lifetime. The proposed estimation algorithm is structured in two timescales: 1) slow-scale, sigma-point Kalman filter (KF)-based estimation of battery capacity and 2) fast-scale, dual-extended KF-based estimation of SoC and model parameters. A particular emphasis is on the adaptive parameterization of SoC and capacity estimators, which provides robust coupling between two timescales and ensures favorable convergence and robust capacity tracking in conditions of SoC and model parameters’ estimation errors. In support of estimation accuracy analysis, an algebraic observability analysis of impedance parameters is conducted. Also, by introducing an observability index calculated in each simulation timestep, a comparison of degrees of observability of different impedance parameter subsets is allowed for. The proposed estimation algorithm is verified both by simulation and experimentally for an electric scooter Li-NMC battery pack.