클러스터링 기반의 인공신경망을 이용한 건축물 가스에너지소비량 예측 및 평가

Abstract

Purpose: Prediction of building energy consumption is an important indicator of effective energy use and savings. Accordingly, studies are actively carried out to find energy use patterns in buildings or to predict energy consumption, especially machine learning technique have been successfully applied in analyzing building energy consumption data. In this paper, we propose a clustering method based on the dynamic time warping(DTW) algorithm, and applied to Artificial neural network(ANN) prediction model to improve the prediction accuracy of gas energy consumption in buildings. Method: To establish cluster-based ANN prediction model, monthly gas energy consumption data of buildings provided by the Korean government were collected. Using the collected data, this study analyzed representative patterns of gas energy consumption through DTW clustering method, and built an ANN prediction model based on these patterns. In addition, the accuracy of the DTW clustering-based ANN predictive model was compared with the general model to confirm whether the accuracy of the predictive model would be improved. Result: It was confirmed that cluster-based ANN prediction model showing representative pattern was analyzed about 12.7% more accurate than general model.