Electricity consumption trend prediction based on a dual-branch long-short term Transformer and cross-attention
Advances in Engineering Technology Research · 12.03.2026 · 16 (1) · 36
DOI: 10.56028/aetr.16.1.36.2026 ↗Kurz gefasst
Fachartikel in der Zeitschrift Advances in Engineering Technology Research, online erschienen am 12.03.2026. Verfasst von Haiqiang Dong und 5 weiteren Autor:innen.
Abstract (Originalsprache)
Forecasting user electricity consumption trends is a critical task for smart grid management and power resource optimization. Transformer models are widely applied in this field due to their powerful sequence modeling capabilities. However, inherent stochasticity in existing Transformer models often leads to fluctuating outputs when processing time series data, compromising the accuracy and stability of predictions. To address this limitation, this paper proposes a novel user electricity consumption trend forecasting model called DCformer (Double-stream Cross-attention Transformer), based on a dual-branch Transformer architecture and designed to enhance model robustness through ensemble learning. Specifically, the DCformer employs a dual-branch structure to simultaneously process the target user's historical consumption data and the consumption data of other relevant users. This design effectively mitigates stochastic errors inherent in single-branch approaches. Furthermore, to bolster the model's ability to capture global dependencies, a cross-attention mechanism is introduced, enabling effective integration of electricity consumption patterns between the target user and other users within the same region. Experimental results demonstrate that the proposed DCformer significantly outperforms traditional methods across multiple datasets, achieving substantial improvements in both prediction accuracy and stability.
Zitieren
APA
Dong, H., Wang, N., Xiao, A., Xuan, P., Ye, P., & Liu, M. (2026). Electricity consumption trend prediction based on a dual-branch long-short term Transformer and cross-attention. Advances in Engineering Technology Research, 16(1), 36. https://doi.org/10.56028/aetr.16.1.36.2026
Vancouver
Dong H, Wang N, Xiao A, Xuan P, Ye P, Liu M. Electricity consumption trend prediction based on a dual-branch long-short term Transformer and cross-attention. Advances in Engineering Technology Research. 2026;16(1):36. doi:10.56028/aetr.16.1.36.2026
BibTeX
@article{Dong2026,
author = {Dong, Haiqiang and Wang, Niuwen and Xiao, Aiyi and Xuan, Pengfei and Ye, Peng and Liu, Mengdi},
title = {Electricity consumption trend prediction based on a dual-branch long-short term Transformer and cross-attention},
journal = {Advances in Engineering Technology Research},
year = {2026},
volume = {16},
number = {1},
pages = {36},
doi = {10.56028/aetr.16.1.36.2026}
}
EUVIMED