Learning-Based Models for Intelligent Control Over Air Conditioning Units in a Smart Building

In power and energy systems, artificial intelligence is commonly used for energy forecasting and decision-making. However, the real deployment of artificial intelligence models for operation in smart buildings is not common, and there is a significant gap between the literature-proposed models and deployed models. In this research, the authors apply multiple machine learning models to learn user-profiles and comfort levels regarding air conditioning usage: decision tree, random forest, support vector machine, and artificial neural network. To validate the models, a real use case with a one-year dataset was used. The dataset preparation is also part of this research. The models were analyzed using simulation and the best solution, in this case support vector machine, was deployed in a real uncontrollable environment using a containerized-based solution. The promising results demonstrate that the proposed methodology can be deployed in real buildings and learn the air conditioning usage profiles within one year of use. The authors also proceed with the deployment of a learning model in a smart building.