Research Projects

The exponential interest and rapid adoption of electric vehicles (EVs) nationwide have accentuated the need for an efficient, reliable, and durable battery pack, and a robust BMS is pivotal to ensure reliable, efficient, and durable battery operation. One of the primary tasks of the BMS is to estimate the battery's current SOH, which quantifies its age or health degradation since the beginning of its operation. Knowing the correct value of battery SOH is extremely important because it can predict the battery's remaining useful life, provide early warnings, and ensure its safety. This project aims to accurately estimate the SOH of an electric vehicle's battery cell and pack during real-world driving and charging, more specifically, whenever a drive cycle finishes or the vehicle is in charging conditions. Reduced-order electrochemical model (ROEM) is the overall framework's central element, which will be implemented online in the battery management system (BMS). The other elements of the framework are a physics-informed deep learning (DL) method, an unscented Kalman filter (UKF), and transfer learning (TL). The estimation framework has two main phases, i.e., the offline phase and the online phase. ROEM has several parameters, such as electrochemical kinetic parameters (exchange current density, charge transfer coefficients, reaction rate constant), transport parameters (Solid Diffusion Coefficient, electrolyte diffusion coefficient, electrolyte conductivity), thermodynamic parameters (open circuit voltage and activity coefficients), geometric parameters (particle radius, porosity, electrode thickness), degradation parameters (SEI layer growth constant, lithium plating rate constant, active material loss rate), and ohmic resistance parameters. These parameters will be offline identified through DL trailed on electrochemical impedance spectroscopy (EIS) data and other battery operational data (e.g., voltage, current, and temperature) collected from different battery-aged conditions. Once the DL is trained on a mapping between the variation of ROEM parameters and corresponding aging-based measurable features, such as EIS data and dis/charging data, the DL and the ROEM can be deployed onboard to the BMS. For the online phase, the trained DL will detect real-time variation of ROEM parameters through real-time dis/charging data and periodically measured impedance spectra. For this phase, one additional hardware is required for the fast EIS method for onboard impedance measurement. Then, the updated ROEM with updated parameters will be used for accurate SOH estimation. The proposed SOH estimation framework will improve battery reliability and safety by closely monitoring the degradation process. Moreover, the framework will promote sustainability by providing the foundation for a battery SOH-aware energy management system, enabling second-life applications, and reducing e-waste.