نهمین کنفرانس بین المللی فناوری و مدیریت انرژی

رفتن به صفحه اصلی

ارائه مجازی مقاله

کد مقاله icemt9-1234

عنوان Building-Level Multi-Energy Load Forecasting: A Graph Representation Learning Convolution-Based Approach in Smart Meter-Enabled Environments

نویسندگان Mir Karim Aboutalebi Zonouz - Mehrdad Tarafdar Hagh - Sajad Najafi Ravadanegh

چکیده The field of load forecasting has recently witnessed a surge in interest, with graph neural network techniques emerging as a powerful tool for achieving high performance and efficiency. While existing research has primarily concentrated on power systems at the transmission and distribution scale, this study pioneers the application of a graph representation learning convolution-based method for electricity and gas demand forecasting at the building level. The proposed approach demonstrates exceptional accuracy in multi-energy load forecasting by modeling apartment floor plans as graphs and integrating internal space data with outdoor weather conditions. In this innovative approach, a graph neural network captures intricate relationships within the building structure, offering a holistic representation of the energy consumption dynamics. To assess the efficacy of the proposed graph network, comprehensive evaluations are conducted across 15 smart buildings, comparing its prediction results with those of a widely used long short-term memory benchmark. The results showcase the superior performance of the graph representation learning convolution-based method, affirming its potential to revolutionize multi-energy load forecasting in smart meter-enabled buildings.

فایل ها

تصویر نویسنده سوم

محتوای ارائه مجازی مقاله

نظرات