Battery State of Health (SOH)

Battery State of Health (SOH) is a critical metric used to evaluate the remaining capacity and performance of a battery relative to its original condition. As batteries age, their capacity to store and deliver energy decreases due to chemical and structural degradation, directly impacting the performance and lifespan of the systems they power. Reliable SOH estimation is crucial, especially in applications like electric vehicles, grid energy storage, and consumer electronics, where battery reliability, safety, and efficiency are paramount. Without accurate SOH estimation, batteries may degrade unnoticed, leading to unexpected failures, reduced efficiency, or even safety hazards due to overheating or short-circuiting.

This project, conducted in collaboration with Cummins Inc., aims to enhance SOH estimation using a Scientific Machine Learning (SciML) Method such as Physics Informed Neural Network (PINN). Traditional data-driven models often require large datasets and may struggle with accuracy across different battery types and usage scenarios. To address these challenges, the project integrates machine learning with battery electrochemical principles, creating a hybrid model that can reliably predict SOH, even with limited data, and capture complex degradation patterns over time. The Python Battery Mathematical Model (PyBaMM), a new library, is utilized along with public datasets to simulate battery behavior and enhance model training. The main goal of the project is to improve SOH estimation accuracy using lab data, followed by the consideration of more realistic cycling conditions, such as drive cycles. Additionally, the project will incorporate pack-level data to better reflect real-world battery performance, ultimately delivering a scalable and reliable solution for SOH estimation across diverse applications.