JAUE2022-111: Building Facilities Classification Using Unsupervised Learning at Kansai University

Abstract

Many buildings are starting to install building and energy management system (BEMS) which is key to managing energy effectively. But the data from BEMS are not utilized enough to minimize energy consumption. Moreover, in architecture, artificial intelligence (AI) is not widely studied and applied. So, in this study, we aimed to find the consumption patterns of energy equipment at Kansai University by using BEMS data and clustering, a kind of AI, and suggest evaluation of energy equipment as a new way to use the data. As a result, by using clustering and silhouette-analysis, a way to find the most optimal number of clusters, we discovered the seasonal changes in the number of clusters and the consumption patterns of the energy equipment. We classified the equipment based on the patterns of energy consumption, and this will make it easier to adjust these facilities and enable their efficient utilization.