BLformer: a short-term electrical bus load forecasting method based on enhanced Patch-TSTransformer

This study addresses the challenges in short-term electrical bus load forecasting. We propose a novel BLformer framework based on an enhanced Patch-TSTransformer. The framework quantifies the importance of temporal features across three load types and filters key input dimensions to reduce redundant...

Full description

Saved in:
Bibliographic Details
Main Authors: Hui Liu, Qian Chen, Di Zhang, Haiyun Wang, Xinchen Zhao, Zaichi Zhang, Lei Fu, Wei Wang, Sizhuang Chen
Format: Article
Language:English
Published: Frontiers Media S.A. 2025-06-01
Series:Frontiers in Energy Research
Subjects:
Online Access:https://www.frontiersin.org/articles/10.3389/fenrg.2025.1622991/full
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:This study addresses the challenges in short-term electrical bus load forecasting. We propose a novel BLformer framework based on an enhanced Patch-TSTransformer. The framework quantifies the importance of temporal features across three load types and filters key input dimensions to reduce redundant information interference. A sparse attention mechanism is designed to dynamically allocate computational resources, balancing efficiency and robustness. Innovatively, we integrate DCNN into the Patch-TST module, combining the advantages of local feature extraction and global temporal modeling to enhance the learning capability of time-frequency coupling characteristics. Furthermore, a coupled prediction strategy is developed to explore high-accuracy bus load forecasting models that incorporate multiple heterogeneous loads. Experiments demonstrate that BLformer significantly outperforms baseline models in terms of RMSE and MAPE metrics. Notably, the indirect prediction strategy substantially reduces errors compared to direct prediction, validating its effective learning ability for multi-load characteristics.
ISSN:2296-598X