This master thesis focuses on developing a Physics-Informed Neural Network (PINN) to predict the degradation of lithium-ion batteries, aiming for a balance between computational efficiency, data requirements, and prediction accuracy. A thorough comparison between physical models and data-driven models was conducted, assessing their theoretical foundations, computational demands, and performance in predicting battery behavior. This evaluation led to the selection of the most suitable model for integration into the PINN framework, demonstrating the significant potential of PINNs as performance models for lithium-ion batteries.
The implementation revealed that while integrating physical laws into the neural network reduces data requirements and enhances interpretability, achieving high accuracy in degradation predictions remains challenging. The study highlights the strengths and limitations of PINNs for battery degradation prediction, showing they can effectively bridge the gap between purely data-driven approaches and traditional physical models. However, further advancements are necessary to refine these models and enhance their predictive capabilities. This thesis emphasizes the need for ongoing research and development to fully utilize PINNs in battery lifecycle management, pointing to their potential for more efficient and accurate degradation predictions.