Abstract
The increasing integration of renewable energy systems into the power grid necessitates enhanced demand-side flexibility in buildings. Data-driven model predictive control (DMPC) has emerged as a promising approach for such energy management task. To advance research in this field, this paper introduces a comprehensive simulation environment designed to facilitate the development and evaluation of DMPC solutions for buildings. This development framework: • encompasses a customizable zone model for multiple room configurations, • is designed with a particular emphasis on thermally activated building structures (TABS) and solar shading control, • integrates DMPC algorithms, with adaptable state space model structures, or using reinforcement learning algorithms, supporting various optimization architectures (centralized, decentralized…). To ensure real-world applicability, the simulation environment has been validated using over one year of data from a living-lab office building. Validation results of the thermal zone models demonstrate good accuracy, with mean absolute errors below 0.5 °C across all zones and simulation time steps (1 and 15 min). The simulation environment exhibits robust performance in simulating complex building systems, effectively capturing the dynamics of thermally activated components and solar shading mechanisms. This versatile framework enables researchers and practitioners to address the challenges posed by increasing building complexity and growing need for decentralized control approaches.