Prediction of Standardized Energy Consumption of Existing Buildings Based on Hybrid Systems Modeling and Control

With the emergence of smart grids and the development of legislation related to the energy consumption of buildings, the need for accurate and reliable energy prediction models has increased in order to support decision making processes. In this work, we present a novel automated methodology for predicting the energy consumption of existing buildings based on real measurements. The methodology relies mainly on the modeling of the indoor air temperature using a hybrid switching system approach with a Piece Wise Auto-Regressive eXogeneous inputs (PWARX) technique. This technique is afterwards coupled with a classification technique, namely, Support Vector Machine (SVM), and integrated in a closed loop with PID controllers designed for each one of the continuous sub-models. The estimation of the energy consumption using the proposed approach based on measurements collected from a test building located in Angers, France is close to the one computed by physics-based methods.