Smart Building Educational Platform

This thesis provides information on advances in the field of building energy management and efficiency by developing an educational platform that uses machine learning tech- niques such as surrogate modeling and reinforcement learning, for the students and users to better understand how energy management and efficiency works depending on different building parameters by introducing the use of IoT capable sensors and actuators making it possible the user awareness at all times. Also, this platform is intended to facilitate the EneryPlus input data files (IDF) edition by delivering an intuitive graphical user interface (GUI) to add the relevant components for making experiments on building behavior and for exporting models into FMUs in order for the user to work with external interfaces. This work uses existing building energy simulation tools like "EnergyPlus" that is the heart and simulation engine of this platform combined with python, Gym, openHAB, in- fluxdb, MQTT and other new mainstream technologies delivering the students/users the ability to improve their understanding and learning process in a faster and better way by removing precious time and effort on coding and configuring their learning environment to start working and experimenting right away on changing different building parameters and on creating different control strategies on a specific building. In this document, a series of results are reported . They were obtained from testing the correct function of the developed educational platform on how adaptive sampling outperforms static sampling in surrogate modeling for building design optimization and on how reinforcement learning is able to learn control strategies in order to improve building performance in the building operational phase. This tool presents students an opportunity to understand how ICT, IoT and AI are key technologies to transit places into smart buildings and how machine learning could accelerate the building design and operational phases optimization by replacing the use of white box complex simulation models such as EnergyPlus with trainable machine learning models and gives ICT in building design teachers a resource to accelerate student learning by using a tool that simplifies testing and setting up coding environments in an all in one ready to install dockerized platform.