Short-Term Load Forecasting for Commercial Buildings Using 1D Convolutional Neural Networks

Abraham Kaligambe, Goro Fujita

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Many Commercial Buildings have employed smart meters to measure load consumption data at real-time intervals and then utilized by the Energy Management System (EMS). Load Forecasting based on historical load data is of key importance for effective operation, planning, and optimization of energy for Commercial Buildings. However, designing an accurate Load Forecasting Model is still an on-going challenge. Our methodology involved the usage of Deep Neural Networks (DNN) for Short-Term Load Forecasting. A special architecture of 1-Dimensional Convolutional Neural Networks (1D CNN) known as WaveNet was employed in our method because of its ability to extract rich features from historical load data sequences. A benchmark load consumption dataset of a Commercial Building for the fiscal year 2017 in Kyushu-Japan was used as a case study. Our model was evaluated and compared to other Machine Learning techniques for Forecasting. When tested on the same dataset, it outperformed them all.

Original languageEnglish
Title of host publication2020 IEEE PES/IAS PowerAfrica, PowerAfrica 2020
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9781728167466
DOIs
Publication statusPublished - 2020 Aug
Event7th Annual IEEE PES/IAS PowerAfrica Conference, PowerAfrica 2020 - Virtual, Nairobi, Kenya
Duration: 2020 Aug 252020 Aug 28

Publication series

Name2020 IEEE PES/IAS PowerAfrica, PowerAfrica 2020

Conference

Conference7th Annual IEEE PES/IAS PowerAfrica Conference, PowerAfrica 2020
CountryKenya
CityVirtual, Nairobi
Period20/8/2520/8/28

Keywords

  • 1D Convolutional Neural Networks
  • Commercial Buildings
  • Deep Neural Networks
  • Energy Management System
  • Short-Term Load Forecasting

ASJC Scopus subject areas

  • Safety, Risk, Reliability and Quality
  • Geography, Planning and Development
  • Energy Engineering and Power Technology
  • Renewable Energy, Sustainability and the Environment
  • Electrical and Electronic Engineering

Fingerprint Dive into the research topics of 'Short-Term Load Forecasting for Commercial Buildings Using 1D Convolutional Neural Networks'. Together they form a unique fingerprint.

Cite this