Energy Schedule Setting Based on Clustering Algorithm and Pattern Recognition for Non-Residential Buildings Electricity Energy Consumption

Building energy modelling (BEM) is crucial for achieving conservation in buildings, but occupant energy-related behaviour often oversimplified traditional engineering simulation methods and thus causes a significant deviation between prediction actual consumption. Moreover, the conventional fixed schedule-setting method not applicable to recently developed data-driven BEM which requires more flexible data-related multi-timescales boost its performance. In this paper, data-based schedule setting by applying K-medoid clustering with Principal Component Analysis (PCA) dimensional reduction Dynamic Time Warping (DTW) distance measurement comprehensive building historical dataset, partitioning data into three different time scales explore usage profile patterns. The Year–Month were partitioned two clusters; Week–Day Day–Hour clusters, matrix was based on result. We have compared performance of proposed default settings calendar using single-layer neural network (NN) model. findings show that predictive BEM, results-based performs significantly better than (with 25.7% improvement) advantageous 9.2% improvement). conclusion, study demonstrates cluster results profiles can be suitable establishment improve BEMs