Cross-Domain Transfer Learning For Lithium-Ion Battery SOH Estimation With Partial Charging and Limited Data

State of Health (SOH) estimation of Lithium-Ion batteries is crucial for battery management systems (BMS) to ensure optimal performance and safety. This paper presents a cross-domain transfer learning approach for SOH estimation of Lithium-Ion batteries employing Deep Neural Networks (DNNs) with rolling window analysis and polynomial feature engineering. The proposed framework leverages partial charging data, a common scenario in real-world applications, and trains entirely on the source domain. Further, domain adaptation was used through fine-tuning utilizing only 5% of target data with K-fold cross-validation, highlighting the critical role of limited data scenarios. Using three distinct battery datasets in 56 source-target domain pair configurations, the developed framework achieved a high prediction accuracy with 94.64% of MAE (Mean Absolute Error) values below 2.5 and 92.86% of R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> values above 0.9. These results revealed the efficiency of the proposed transfer learning model using practical partial charging data and limited target data, saving extensive costs for degradation experiments.