The aim of this thesis is to develop a framework for training reinforcement learning (RL) agents for the management of energy systems and to evaluate the feasibility of applying reinforcement learning in this domain. The project will be carried out in collaboration with the Faculty of Mechanical Engineering at BUT. The student will analyze an existing simulator developed by a research group at BUT and propose and implement the modifications necessary to make it suitable for efficient training of reinforcement learning agents. The work will include an experimental evaluation demonstrating the simulator’s readiness for RL-based training. The thesis will also include an experimental comparison and feasibility study of several RL deployment regimes, such as direct control versus control through an energy management system (EMS) intermediary, and direct control versus RL-based correction of a rule-based controller.