A general framework for lithium-ion battery state of health estimation: From laboratory tests to machine learning with transferability across domains

Accurate State of Health (SOH) estimation is crucial for safe, efficient, and optimal operation of lithium-ion batteries (LIBs), yet it remains challenging in real-world applications. In this regard, this paper presents a novel approach to estimating the SOH of lithium-ion batteries using a convolutional neural network (CNN) model enhanced with 3D histogram feature extraction and transfer learning. Unlike traditional models, our method is uniquely capable of handling varying lengths of input time series data with the varying sliding window, making it highly adaptable to real-world scenarios where data may be irregular or incomplete. The integration of transfer learning further enhances the model’s adaptability, allowing it to efficiently generalize across different battery types and operational conditions with minimal retraining. Experimental results demonstrate the model’s accuracy and robustness, with significant improvements over existing methods in terms of estimation accuracy, computational efficiency, and adaptability to new data. This research offers a practical and scalable solution for battery health monitoring, supporting the advancement of reliable and efficient battery management systems. • Comprehensive datasets are generated from accelerated aging tests under various conditions. • Innovative use of 3D histogram feature extraction for SOH estimation. • Capability to handle varying lengths of input time series data. • Enhanced model adaptability through the integration of transfer learning. • Validated through comprehensive experimental testing across different battery types and conditions.