Machine Learning‐Based Lifetime Prediction of Lithium‐Ion Cells

• DOI: 10.1002/advs.202200630
• Publication Date: 2022-08-26T18:50:44Z
• Authors: Kai Schofer; Florian Laufer; Jochen Stadler; Severin Hahn; Gerd Gaiselmann; Arnulf Latz; Kai P. Birke
• Keywords: Interpretability; Symbolic Regression; Overfitting

Abstract Precise lifetime predictions for lithium‐ion cells are crucial efficient battery development and thus enable profitable electric vehicles a sustainable transformation towards zero‐emission mobility. However, limitations remain due to the complex degradation of cells, strongly influenced by cell design as well operating storage conditions. To overcome them, machine learning framework is developed based on symbolic regression via genetic programming. This evolutionary algorithm capable inferring physically interpretable models from aging data without requiring domain knowledge. novel approach compared against established approaches in case studies, which represent common tasks prediction cycle calendar 104 automotive pouch‐cells. On average, predictive accuracy extrapolations over time energy throughput increased 38% 13%, respectively. For other stress factors, error reductions up 77% achieved. Furthermore, generated meet requirements regarding applicability, generalizability, interpretability. highlights potential algorithms enhance insights.