A Spatio-Temporal Prediction Model for Exhaust Gas Temperature in Regenerative Aluminum Smelting Furnaces Towards Energy Efficiency and Carbon Reduction
Sustainability · 24.07.2026 · 18 (15) · 7539
DOI: 10.3390/su18157539 ↗Kurz gefasst
Fachartikel in der Zeitschrift Sustainability, online erschienen am 24.07.2026. Verfasst von Jiayang Dai und 3 weiteren Autor:innen. Die Arbeit nennt 34 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Guangxi Natural Science Foundation, Guangxi Science and Technology Major Program, National Natural Science Foundation of China.
Abstract (Originalsprache)
Exhaust gas temperature is a critical indicator of combustion efficiency and waste heat recovery in regenerative aluminum smelting furnaces, directly governing the energy intensity and carbon footprint of secondary aluminum production—a cornerstone of the circular economy. However, accurate prediction is challenged by pronounced long-range thermal lag and dynamically evolving spatial dependencies among process variables during operational transitions, which hinder real-time process optimization for energy savings. To address these challenges and advance sustainable manufacturing, a novel prediction model termed ChronoClassGAT (Chronological Class-aware Graph Attention Network) is proposed in this paper. The model integrates three key components: (1) a Temporal Convolutional Autoencoder (TCN-AE) with a Gaussian Mixture Model (GMM) for unsupervised identification of distinct operating regimes, providing categorical priors for dynamic graph construction; (2) a multi-graph fusion mechanism that builds operating-condition-specific spatial structures, enabling a Graph Attention Network (GAT) to adaptively model evolving inter-variable dependencies; and (3) a ChronoSwish activation function that modulates LSTM-based temporal encoding with time-aware periodic and switching signals, enhancing responsiveness to transient dynamics. Validated on real-world industrial datasets, ChronoClassGAT achieves superior prediction accuracy (RMSE of 2.2712, MAE of 1.8023, and R2 of 0.9985) over state-of-the-art baselines. By enabling precise and robust exhaust gas temperature forecasting, our framework provides the decision-support intelligence needed for optimizing regenerator switching, minimizing thermal losses, and reducing fuel consumption, thereby contributing significantly to the operational energy efficiency and environmental sustainability of the energy-intensive non-ferrous metal industry.
Förderung
- Guangxi Natural Science Foundation
- Guangxi Science and Technology Major Program
- National Natural Science Foundation of China
Zitieren
APA
Dai, J., Wang, L., Li, S., & Wu, T. (2026). A Spatio-Temporal Prediction Model for Exhaust Gas Temperature in Regenerative Aluminum Smelting Furnaces Towards Energy Efficiency and Carbon Reduction. Sustainability, 18(15), 7539. https://doi.org/10.3390/su18157539
Vancouver
Dai J, Wang L, Li S, Wu T. A Spatio-Temporal Prediction Model for Exhaust Gas Temperature in Regenerative Aluminum Smelting Furnaces Towards Energy Efficiency and Carbon Reduction. Sustainability. 2026;18(15):7539. doi:10.3390/su18157539
BibTeX
@article{Dai2026,
author = {Dai, Jiayang and Wang, Lei and Li, Shenwang and Wu, Thomas},
title = {A Spatio-Temporal Prediction Model for Exhaust Gas Temperature in Regenerative Aluminum Smelting Furnaces Towards Energy Efficiency and Carbon Reduction},
journal = {Sustainability},
year = {2026},
volume = {18},
number = {15},
pages = {7539},
doi = {10.3390/su18157539}
}
EUVIMED