State estimation, performance prediction and health assessment of Lithium-ion batteries for advanced battery management systems

Abstract

The development of battery technologies in the last two decades has reinvigorated the application of energy systems, including consumer electronics, energy storage systems, and electric vehicles (EVs). To enlarge the market share of the applications mentioned above, the top concerns are focused on safety, durability and cost issues of batteries. A battery management system (BMS) plays a significant and irreplaceable role in a battery-powered system. An advanced BMS not only has basic functions such as monitoring and charge/discharge control, but also provides accurate battery state information for timely recharging, replacement, and predictable maintenance. The information mainly includes an estimation of remaining charge, a prediction of remaining useful performance (RUP), and a health assessment. The uncertainty of a battery's performance poses a challenge to the implementation of these functions in current BMSs because a battery acts differently under different operational and environmental conditions. Without a mature benchmark in the battery industry, the study on advanced BMSs is undoubtedly a demanding task. This thesis aims to improve the functionalities of BMSs, particularly state evaluation of the battery system, based on the experimental data. The main contributions of this thesis are as follows: • Developing temperature-based models to estimate remaining charge of a battery over a discharge cycle Ambient temperature is a significant factor that influences the accuracy of battery state of charge (SOC) estimation, which is critical for predicting the remaining driving range of electric vehicles (EVs) and the optimal charge/discharge control of batteries. Two temperature-based equivalent circuit models (ECMs) are developed to estimate battery SOC. Their corresponding unscented Kalman filtering (UKF) estimators are designed to update model parameters and hidden states to address various uncertainties. The results indicate that the proposed temperature-based models provide a more accurate SOC estimation in comparison with the original models without taking into account the operating temperatures. • Proposing an ensemble model to predict the remaining performance of a battery over a life cycle An ensemble model is developed to characterize degradation and predict the RUP of lithium-ion batteries. The model fuses an empirical exponential and a polynomial regression model to track the trend of battery degradation over its life cycle based on the experimental data. Model parameters are adjusted online by using a particle filtering (PF) approach. Experiments are conducted to compare the prediction performance of the developed ensemble model with the individual results of the empirical models mentioned above. Another set of experimental capacity data is used to validate the estimation performance of our developed model. • Proposing a health indicator to assess battery health under different discharge rates A weak cell in a battery accelerates the deterioration of the battery, which usually consists of several cells in series or in parallel. Real-time monitoring and timely replacement of the individual cell are essential to guarantee a cost-saving and effective operation of a battery system. Although the deliverable capacity of the cell is used to characterize its degradation and evaluate its SOH, capacity is a confusing indicator because it is dependent on the discharge rate, which is determined by the usage profile. According to our investigation, the Peukert coefficient (PC) can be extracted to characterize the battery health. Behavior of the PC can be more effectively used to identify the weak cell than other existing indicators, such as available capacity, internal resistance, and battery temperature. An exponential weighted moving average (EWMA) method is employed to monitor the PC profile, and thus identify the weak cell. The PCbased EWMA approach can be used for early detection of the weak cell or battery fast screening before delivery. This method is expected to provide insight into the timely replacement of the weak cell in a battery system. The experimental data are collected based on the established battery monitoring system. For each topic mentioned above, several battery samples are tested for system identification while several other samples are used to validate the developed methodology. The proposed models and methods for state estimation, performance prediction, and health assessment will provide insights into the development of an advanced BMS. It is expected that they could be applied to the current BMS for EV applications and energy storage systems.

Date of Award	15 Jul 2014
Original language	English
Awarding Institution	City University of Hong Kong
Supervisor	Kwok Leung TSUI (Supervisor) & Michael Gerard PECHT (Co-supervisor)