Mobile QR Code QR CODE : Korean Journal of Air-Conditioning and Refrigeration Engineering
Korean Journal of Air-Conditioning and Refrigeration Engineering

Korean Journal of Air-Conditioning and Refrigeration Engineering

ISO Journal TitleKorean J. Air-Cond. Refrig. Eng.
  • Open Access, Monthly
Open Access Monthly
  • ISSN : 1229-6422 (Print)
  • ISSN : 2465-7611 (Online)
Title Building Energy Demand Prediction Model using Fourier Transform based Schedule Analysis Algorithm
Authors Yusun Ahn ; Yong-jun Lee ; Eun Joo Oh ; Byungseon sean Kim
Page pp.386-397
ISSN 1229-6422
Keywords 기계 학습; 건물 에너지 소비량 예측; 스케줄 분석 알고리즘; 푸리에 변환 Machine learning; Building energy demand prediction; Building schedule analysis algorithm; Fourier transform
Abstract Building energy demand currently accounts for 40% of the total energy demand, which has a great influence on the planning and operation of the energy market by energy suppliers, and its importance has increased significantly with the advent of the smart grid. Variables affecting the use of building energy include the identification of environmental conditions, historical conditions, and schedule conditions, and these factors have a sophisticated effect on buildings. One of the most influential variables is the building schedule. Because each building has its own schedule, standardized schedules cannot be applied to various buildings, and it is difficult for non-experts to analyze or predict schedules in these cases. The aim of this paper is to propose a high-precision building energy demand prediction model by using a Fourier transform-based time series prediction model, that automatically analyzes and predicts the schedule to be applied when predicting building energy demand. In order to compare with the existing method, the six buildings are divided into schedules when they are not scheduled, weekdays/weekends, days of week and when schedule analysis algorithm are applied. Machine learning is performed using the LSTM model, and prediction accuracy is verified through the CvRMSE and MBE. There was an average difference of 15.37% based on the CvRMSE, and all predictions were predictable when the automated prediction model was applied. This study can be used as a building energy operation plan for the creation and implementation of a future energy-efficient smart grid system.