Model-Based Condition Monitoring and Power Management for Rechargeable Electrochemical Batteries

Abstract

Rechargeable multicell batteries have been used in various electrical and electronic systems, e.g., renewable energy systems, electric-drive vehicles, commercial electronics, etc. However, there are still concerns about the reliability and performance degradation of rechargeable batteries caused by low thermal stability and the aging process. A properly designed battery management system (BMS) is required for condition monitoring and control of multicell batteries to ensure their safety, reliability, and optimal performance. The goal of this dissertation research was to develop a novel BMS for rechargeable multicell batteries.

First, this research developed high-fidelity battery models for online condition monitoring and power management of battery cells. The battery models were capable of capturing the dynamic circuit characteristics, nonlinear capacity and nonlinear open-circuit voltage effects, hysteresis effect, and temperature effect of the battery cells.

Second, this research developed a novel self-X, multicell battery design. The proposed multicell battery can automatically configure itself according to the dynamic load/storage demand and the condition of each cell. The proposed battery can self-heal from failure or abnormal operation of single or multiple cells, self-balance from cell state imbalances, and self-optimize to improve energy conversion efficiency. These features were achieved by a highly efficient cell switching circuit and a high-performance condition monitoring and control system.

Moreover, this research developed several model-based condition monitoring algorithms based on the proposed battery models. First, a particle swarm optimization-based parameter identification algorithm was developed to estimate the impedance and state of charge (SOC) of batteries using the proposed hybrid battery model. Second, an algorithm combining a regression method for parameter identification, a sliding-mode observer for SOC estimation, and a two-point capacity estimation method were proposed. In addition, an electrical circuit with hysteresis model-based condition monitoring algorithm was proposed. It systematically integrates: a fast upper-triangular and diagonal recursive least square for online parameter identification, a smooth variable structure filter for SOC estimation, and a recursive total least square for maximum capacity and state of health estimation. These algorithms provided accurate, robust condition monitoring for lithium-ion batteries. Due to the low complexity, the proposed second and third algorithms are suitable for the embedded BMS applications.