A thermal control methodology based on a machine learning forecasting model for indoor heating

To take advantage of the data generated in buildings, this document proposes a methodology based on machine learning model to improve thermal comfort and energy efficiency. This uses measured (e.g., indoor/outdoor temperature, relative humidity, etc.) forecast meteorological data) train multiple linear regression indoor temperature space under study. Using genetic algorithm optimization method, is then used evaluate different heating strategies generated. For each strategy, score assigned according user-defined criteria order prioritize them select best one. By studying an office building simulated TRNSYS software, was implemented with errors less than 1% adjusted R2 coefficient close 0.9. Compared conventional can by up 43%.