Data-driven Optimization of Energy Efficiency and Comfort in an Apartment

• Published in: 2018 International Conference on Intelligent Systems (IS) pp. 174 - 182
• Main Authors: Avendano, Diego Nieves, Ruyssinck, Joeri, Vandekerckhove, Steven, Van Hoecke, Sofie, Deschrijver, Dirk
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01-09-2018
• Subjects: Actuators; Atmospheric modeling; Deep reinforcement learning; Energy efficiency; Heating systems; Home automation; Model predictive control; Multi-objective optimization; Optimization; Ventilation; Windows
• DOI: 10.1109/IS.2018.8710456

An important challenge in home automation is the energy efficient optimization of the indoor environment. This relies on the solution of a multi-objective optimization problem where energy efficiency and comfort parameters are maximized simultaneously. This paper presents three data-driven control algorithms based on machine learning techniques, which offer an alternative to traditional control methods. The results demonstrate that some data-driven methods can achieve similar results than rule-based systems. Moreover, they require no prior expert knowledge and have better scalability than standard approaches.