Feature-Based and Modeling Dataset for Lithium-Ion Battery Remaining Useful Life Prediction Across Multiple Public Datasets

This dataset contains feature representations and machine learning-ready modeling data for lithium-ion battery Remaining Useful Life (RUL) prediction. The dataset is derived from multiple publicly available battery datasets (CALCE, HUST, MATR, RWTH, XJTU, among others) and includes both charge and discharge cycle information. It is organized into two main components:(1) Feature dataset: curve-based and hierarchical feature representations extracted from voltage, current, and capacity trajectories.(2) Modeling dataset: machine learning-ready tabular data including features and target RUL values prepared for supervised learning experiments. The dataset supports reproducible benchmarking under multiple evaluation protocols including global training, per-dataset evaluation, and leave-one-dataset-out (LODO) validation. This release does not include raw intermediate preprocessing artifacts to ensure clarity and compactness of the reproducible pipeline.