Dissertation Defense: Model Order Reduction of Multi-Dimensional Partial Differential Equations for Electrochemical-Thermal Modeling of Large-Format Lithium-ion Batteries

Committee

• Professor Marcello Canova, Chair (ME)
• Professor Giorgio Rizzoni (ME)
• Professor Vishnu Baba Sundaresan (ME)
• Professor Soheil Soghrati (ME)

Abstract

Lithium ion batteries are considered the state of the art for energy storage in electric and hybrid vehicles. However, there are still several major challenges, such as battery safety, durability and cost, limiting the widespread application of Li-ion batteries in electrified vehicles. Understanding and predicting the chemical and physical processes in Li-ion cells is possible through multi-scale characterization methods. However, "in-situ" quantification of such processes on a vehicle is not yet achievable due to the absence of direct measurements. Hence, high-fidelity, first-principles models are an essential investigation tool for the prediction of the battery performance and life. While such multi-scale, multi-dimensional first-principles models allow one to characterize the distribution of electrochemical and thermal properties within the cell, they require significant calibration effort and computation time, due to the presence of large scale coupled Partial Differential Equations (PDEs) and nonlinear algebraic equations, ultimately preventing their application to estimation and control algorithm design and verification. This dissertation presents the reduced order electrochemical-thermal models derived from first principles and suitable for real-time simulation, estimation and control design, through the systematic use of projection methods to achieve direct Model Order Reduction (MOR) from linear and nonlinear parabolic PDEs to low-order Ordinary Differential Equations (ODEs). The proposed methodology is applied to an electrochemical-thermal model for the simulation of large-scale Lithium ion battery cells. The resulting reduced-order multi-scale, multi-dimensional model is validated against numerical solutions and experimental data at various input current conditions. The physics-based, ultra-fast modeling tools developed within this research will enable accurate prediction of the electrochemical and thermal distributions within the battery cells, supporting simulation and analysis of performance and remaining usable life of the Li-ion batteries in electrified vehicles.