AUTHOR=Wang Zhiyu , Zhu Zhen , Xiao Geyang , Bai Bing , Zhang Yinjie 

TITLE=A Transformer-Based Multi-Entity Load Forecasting Method for Integrated Energy Systems

JOURNAL=Frontiers in Energy Research

VOLUME=Volume 10 - 2022

YEAR=2022

URL=https://www.frontiersin.org/journals/energy-research/articles/10.3389/fenrg.2022.952420

DOI=10.3389/fenrg.2022.952420

ISSN=2296-598X

ABSTRACT=Energy demand forecasting is a critical component of energy system scheduling and optimization. Classified as a time-series forecasting problem, this technique uses previous features as inputs to estimate future energy demands. An integrated energy system, unlike a classical single-target scenario, has a hierarchy of multiple interrelated energy consumption entities as prediction targets. Existing data-driven approaches typically interpret entity indexes as suggestive features, which fail to adequately represent entity correlations. This paper, therefore, proposes a neural network model named CETFT (Cross-entity Temporal Fusion Transformer) that leverages a cross-entity attention mechanism to model inter-entity correlations. This enhanced attention module can describe the relationships between multiple entities within a time window and inform the decoder about which entity in the encoder to concentrate on. In order to reduce the computational complexity, shared variable selection networks are adapted to extract features from different entities. A data set obtained from 13 buildings on campus is used as a case study to verify the performance of the proposed approach. Compared to the comparative models, the proposed model achieves the smallest error on most horizons and buildings. Furthermore, variable importance, temporal correlations, buildings correlations and time-series patterns in data are analyzed with the attention mechanism and variable selection networks, and the interpretability of the model is verified.