Abstract
Reinforcement learning (RL) has proven effective for AI-based building energymanagement. However, there is a lack of flexible framework to implement RLacross various control problems in building energy management. To address thisgap, we propose BuildingGym, an open-source tool designed as aresearch-friendly and flexible framework for training RL control strategies forcommon challenges in building energy management. BuildingGym integratesEnergyPlus as its core simulator, making it suitable for both system-level androom-level control. Additionally, BuildingGym is able to accept externalsignals as control inputs instead of taking the building as a stand-aloneentity. This feature makes BuildingGym applicable for more flexibleenvironments, e.g. smart grid and EVs community. The tool provides severalbuilt-in RL algorithms for control strategy training, simplifying the processfor building managers to obtain optimal control strategies. Users can achievethis by following a few straightforward steps to configure BuildingGym foroptimization control for common problems in the building energy managementfield. Moreover, AI specialists can easily implement and test state-of-the-artcontrol algorithms within the platform. BuildingGym bridges the gap betweenbuilding managers and AI specialists by allowing for the easy configuration andreplacement of RL algorithms, simulators, and control environments or problems.With BuildingGym, we efficiently set up training tasks for cooling loadmanagement, targeting both constant and dynamic cooling load management. Thebuilt-in algorithms demonstrated strong performance across both tasks,highlighting the effectiveness of BuildingGym in optimizing cooling strategies.