State Estimation, Performance Prediction, and Health Evaluation of Commercial Lithium-ion Batteries

商用鋰離子電池的電量估計、性能預測和健康評估

Student thesis: Doctoral Thesis

View graph of relations

Author(s)

Detail(s)

Awarding Institution
Supervisors/Advisors
Award date15 Sep 2017

Abstract

In the automotive industry, electric vehicles (EVs) have attracted increasing attention and achieved ever better development as an important technology to reduce greenhouse gas emissions and consumption of natural resources. An energy storage system, as one of the key parts of EVs, plays a significant role in determining the efficiency, reliability, and safety of EV systems. Lithium-ion batteries are electrical energy storage devices characterized by their high energy density, good power capability, long lifespan, and environmental friendliness, and thus meet the wide range of specific requirements in automotive applications. Due to demanding driving operations, a battery management system (BMS) is required to ensure safe and reliable operation, and it guarantees the good performance of batteries. Accurate state of charge (SOC) estimation, remaining useful life (RUL) prediction, and battery health evaluation are three main concerns related to BMSs.
The aim of this thesis is to develop methods for improving the functional modules of BMSs, particularly in terms of state estimation, performance prediction, and health evaluation of the battery system, based on experimental data. The main contributions of this thesis are as follows.
• Comparing and testing the robustness of three model-based filtering algorithms in estimating the SOC of a battery under a new dynamic loading profile
Accurate SOC estimation is critical for the safety and reliability of BMSs in EVs. As SOC cannot be directly measured and its estimation is affected by many factors, such as ambient temperature, battery aging, and current rate, a robust SOC estimation must be developed to deal with time-varying and nonlinear battery systems. Three popular model-based filtering algorithms, including the extended Kalman filter, unscented Kalman filter, and particle filter (PF), are respectively used to estimate SOC, and their performance in terms of tracking accuracy, computation time, robustness against the uncertainty of initial SOC values, and battery degradation is compared. To evaluate the performance of these algorithms, a new combined dynamic loading profile composed of the dynamic stress test, the federal urban driving schedule, and the US06 is proposed. The comparison results show that the unscented Kalman filter is the most robust to different initial SOC values, while the PF has the fastest convergence ability when an initial SOC estimate is far from the true initial SOC.
• Developing a two-term logarithmic model to predict the RUL of a battery with two-phase degradation
Some lithium-ion battery materials show two-phase degradation behavior with evident inflection points, such as lithium nickel manganese cobalt oxide (NMC) cells. A model-based Bayesian approach is proposed to predict the RUL for these types of batteries. First, a two-term logarithmic model is developed to capture the degradation trends of NMC batteries. By fitting the battery degradation data, it is experimentally demonstrated that the developed model is superior to existing empirical battery degradation models. A PF-based prognostic method is then incorporated into the model to estimate the batteries’ possible degradation trajectories. Correspondingly, the RUL values of NMC batteries are expressed in terms of the probability density function. The collected experimental data verify the effectiveness of the developed method. The results indicate that the proposed prognostic method can achieve higher predictive accuracy than the existing two-term exponential model.
• Revealing the relationship between coulombic efficiency (CE) and capacity degradation of commercial lithium-ion batteries under cycle life tests
High CE usually indicates a long battery cycle life. However, the relationship between CE evolution and battery degradation is not yet fully understood. To explore the behavior of long-term CE and clarify its relationship with capacity degradation, cycle life tests of two types of mainstream commercial lithium-ion batteries representing different degradation shapes are conducted. Experimental observations are made, and the results are discussed in depth. Incremental capacity analysis is conducted to identify battery-aging mechanisms. It is observed that CE can indicate capacity fading rates inside a cell. Additionally, the experimental results show that, in addition to a loss of lithium inventory, a loss of active material accelerates battery degradation and brings down the CE value.
The experimental data were collected based on the established battery test bench. For each aforementioned topic, a group of battery samples was tested for battery parameter identification, while another group of battery samples was used to validate the developed methodology. The proposed models and algorithms for SOC estimation, RUL prediction, and health evaluation should provide insights into the development of an advanced BMS. They are expected to be applicable to the current BMS for EV application.