Battery technologies are central to applications such as electric vehicles, large-scale energy storage systems, and portable electronics, where understanding long-term performance is important. In a blog by Somayeh Toghyani, Gamma Technologies presents a workflow that combines physics-based simulation with machine learning to estimate battery aging.
The approach uses GT-AutoLion, an electrochemical battery model within GT-SUITE, to run simulations and generate virtual datasets under different operating conditions. These datasets are then used to train machine-learning metamodels through the ML Assistant tool in GT-SUITE. Training data can be imported from sources such as GT-Post, CSV, Excel, TXT, and MAT files.
The workflow is demonstrated through two scenarios: calendar aging, where batteries degrade during storage depending on temperature and state of charge, and cycle aging, where repeated charge–discharge cycles affect capacity loss. Using datasets created through a Design of Experiments (DoE), the trained models estimate battery state of health (SOH) across longer timeframes and varied operating conditions.
Image generated by: Gemini