Multi-Dimensional Feature Fusion Framework for Robust Lithium-Ion Battery State of Health Estimation Using Partial Voltage Segments

Accurate estimation of lithium-ion battery state of health (SOH) is critical for efficient and safe battery operation. However, existing methods often lack generalization capability, depend on complete charge-discharge cycles, and use overly simple feature-fusion strategies, limiting their ability to extract aging information from partial voltage data. To overcome these challenges, we propose a hybrid deep learning model based on an Inception-BiGRU architecture for accurate and robust SOH estimation.The method first applies a multi-dimensional feature fusion strategy to extract three effective voltage-segment feature sequences as health indicators. Then, an improved Inception-BiGRU model is designed: multi-scale parallel Inception convolution modules efficiently capture voltage features while avoiding the fixed receptive field limitations of traditional CNNs, and bidirectional GRUs enhance modeling of long-term temporal dependencies. Comprehensive experiments systematically evaluate the contribution of different voltage segments, showing that high-voltage regions provide more informative aging features. We also investigated the effects of sequence length and feature combinations. Experimental results demonstrate that the proposed approach significantly outperforms conventional temporal and hybrid models, while exhibiting strong robustness to noise and excellent generalization.