Parameter Identification Of The Multi-physics Model And Health Feature Extraction For Lithium-ion Battery

Lithium-ion batteries are widely used in electric vehicle and energy store, accurate evaluation of the internal health status is the prerequisite for ensuring the safety of the batteries. Thus, detailed information of the internal health status is needed. However, Lithium-ion battery is a typically dynamic and nonlinear electrochemical system, it is difficult to obtain its internal health status. Researchers on electrode material study the degradation mechanism by using “post-mortem” method, which does not apply here because it has to disassemble the batteries. Data-driven based remain useful life prediction can not describe the degradation mechanism inside batteries, therefore it is not able to describe the internal health status. This dissertation focus on parameter identification and health feature extraction of lithium-ion batteries based on the multi-physics model. It aims at the key issues on how to accurately obtain the internal health features of Lithium-ion batteries without destroying them, and mining the relationship between the internal health features and battery aging, then quantitative analyzing the factors which cause degradation of battery performance. Detailed health status information is then obtained via those studies.Firstly, a multi-physics model was developed using P2 D model, thermal model and temperature distribution model of lithium-ion battery. The P2 D model and heat generation equations were used to calculate the heat behavior, and the temperature distribution was calculated with the thermal impedance model, Arrenius’ law was used to establish the coupling of electrochemical and heat behavior inside the battery. This model can accurately simulate the external performance and internal physical or chemical process on any operating condition, especially the simulation of shell temperature which crucial to the research on heat behavior inside the battery. Simulation analysis was conducted by using the multi-physics model, at the same time a relationship between the model parameters and battery performance is established.For different influences on battery performance from model parameters, sensitivity analysis was carried out for both Li Fe PO4 and Li Co O2 batteries. Model simulations were used to obtain terminal voltage and shell temperature curves when the parameters change under different operating conditions. And the “Sensitivity values matrix” was proposed to quantitatively describe the relationship between parameter sensitivity and operating conditions. Then the parameter sensitivity was used as one of the objectives to optimize the identification condition using NSGA-II algorithm. The optimal identification conditions of two types of batteries were obtained. The parameter sensitivities under those conditions are higher than that under traditional conditions, it is beneficial for accurate parameter identification.For accurately identifying multi-physics model parameters, a multi-objective method was established. A cataclysm operator and a searching range expansion operator were used to improve search ability, and the parallelized genetic algorithm was used to accelerate the identification process. The terminal voltage and shell temperature, which measured under optimal conditions at two ambient temperatures were proposed as four optimization objectives. The effectiveness and result accuracy of multi-objective method was verified via “synthetic experimental data”. Comparison study shows that the accuracy of the proposed method is superior to traditional methods. Then the proposed method was used for two types of commercial batteries, the effectiveness was also verified by experimental results.In order to extract the battery internal health features, a multi-mode cycle life test was conducted for Li Co O2 batteries. The parameter sets of batteries in different aging stages and different aging modes were achieved using multi-objective parameter identification. Two-factor analysis of variance was carried out for those parameter sets, showing that there were 8 key parameters significant change with battery aging, and thus they were defined as internal health features. Degeneration trends of health features were obtained by fitting the values of key parameters, and the cycle aging mechanisms were qualitatively analyzed with the principle of side reaction in electrodes.At last, quantitative analysis of the factors which cause the performance degradation of Li Co O2 battery was conducted using the degradation of internal health features. The identified parameter sets were put into the multi-physics model, then ascertained the three factors of capacity loss, and then quantatively calculated their change pattern under different aging modes. By using the disassemble equation of total overpotential, the ratio changes of five parts of overpotential were quantitatively studied, and the main reason of total overpotential rises was also found in different aging modes and different aging stages. The relationship between the heat behavior and battery aging was also quantitatively studied with the heat generation and heat exchange equations.