| The building sector is the primary consumer of energy in France. In order to reduce energy consumption and gas emissions, it is necessary to optimize the management and control of energy systems. This study presents a Multi-Agent System (MAS) approach to dynamically detect daily profiles in time series data issued from sensors deployed in buildings. Daily profiles help buildings managers to understand the building behavior, detect anomalies, and better control the energy systems. The proposed model applies a dynamic and adaptive clustering method based on a threshold derived from historical data. The methodology has been validated on three types of data issued from three datasets: the heating consumption from a private office building dataset in Toulouse, France; the outdoor temperature; and the electricity consumption from two publicly available datasets. The results demonstrate the effectiveness of the proposed approach in identifying meaningful daily profiles that characterize different operational building patterns. Beyond profiles detection, our objective is to develop a global adaptive system capable of predicting future energy demand. |
*** Title, author list and abstract as submitted during Camera-Ready version delivery. Small changes that may have occurred during processing by Springer may not appear in this window.