Cimen, Halil2025-02-102025-02-1020251551-32031941-0050https://doi.org/10.1109/TII.2024.3520180https://hdl.handle.net/20.500.13091/9850CIMEN, HALIL/0000-0003-0104-3005Li-ion batteries play a crucial role in green energy goals, but estimating their parameters is challenging due to their nonlinear structure, aging effects, and varying chemistries. In this article, a distribution, scale and context sensitive, convolutional neural network-based state of charge estimation model is proposed. First, the proposed model improves generalization by addressing data distribution shifts in batteries across different temperatures through individual sample handling. Second, by stacking convolutional layers with varied receptive fields, the model captures both local and global dependencies, providing the model with multiscale features and hierarchical representation. Finally, we add a self-attention module to enhance learning of input sequences by focusing on relevant parts and understanding the global context of features. Experiments were performed on single-domain and cross-domain settings to prove the effectiveness of the model. The results obtained demonstrate that the proposed model significantly outperforms state-of-the-art approaches in terms of both accuracy and generalization capability.eninfo:eu-repo/semantics/closedAccessState of chargeData modelsEstimationContext modelingPredictive modelsAnalytical modelsFeature extractionComputational modelingAdaptation modelsAccuracyConvolutional neural network (CNN)deep learningenergy storagegeneralization capabilityli-ion batteryself-attentionstate of charge (SOC) estimationArticle10.1109/TII.2024.35201802-s2.0-85214094292