Adaptive modeling in the frequency and time domain of high-power lithium titanate oxide cells in battery management systems

Abstract

Lithium-ion batteries play a major role for the reliability and safety of the energy supply in vehicle power networks. Battery management systems (BMS) are needed to monitor the battery's inner states, such as state of charge or state of available power. State-of-the-art algorithms running on a BMS cannot fulfill all today's requirements completely such as qualification for an automotive safety integrity level, implementation on low-cost hardware, adaption to different kinds of aging effects and minimized excitation in a vehicle's power network. In this paper a new approach for online parameter estimation of a battery model for a lithium-ion battery with titanate oxide anode is presented. The approach combines simple mathematical equations and low memory consumption while still adapting dynamically to different aging effects. In order to handle different types of excitation in the power network, three methods are suggested for analyzing the battery behavior in the time and frequency domain. The algorithms are validated utilizing real driving profiles applied to aged batteries. Therefore, calendric and cyclic battery aging tests were carried out for about two years and up to 100,000 equivalent full cycles. The frequency-based methods are advantageous in terms of impedance parameter tracking while the methods using time-domain parameterization data benefit in terms of time predictions. All proposed methods show promising results regarding voltage predictions independent of the battery's state of health.