A reliable mixed-integer linear programming formulation for data-driven model predictive control in buildings

Area of Science:

• Building energy systems
• Renewable energy integration
• Demand-side flexibility

Background:

• Buildings with Thermally Activated Building Structures (TABS) and glass facades offer flexibility potential but face thermal balance challenges due to high inertia and solar gains.
• Data-driven Model Predictive Control (DMPC) is promising for enhancing demand-side flexibility in such buildings.
• Existing DMPC implementations for combined TABS and shading control require complex white-box models, hindering replicability.

Purpose of the Study:

• To present a detailed optimization algorithm for DMPC in buildings with TABS and shading systems.
• To facilitate the implementation of advanced control strategies for renewable energy integration.
• To address the limitations of existing white-box modeling approaches.

Main Methods:

• Development of a grey-box model for thermal zones, utilizing a reduced-order state-space representation.
• Formulation of a Mixed-Integer Linear Programming (MILP) optimization problem for joint control of TABS and shading.
• Integration of thermal comfort requirements as constraints within the optimization framework.

Main Results:

• The proposed algorithm enables joint control of TABS and shading systems within a DMPC framework.
• The grey-box model and MILP formulation provide a replicable and efficient approach.
• Thermal comfort is effectively managed as a constraint, simplifying the objective function.

Conclusions:

• The presented optimization algorithm facilitates the implementation of DMPC for combined TABS and shading control in buildings.
• This approach enhances demand-side flexibility for renewable energy integration while maintaining thermal comfort.
• The use of a grey-box model and MILP formulation overcomes limitations of previous white-box model-dependent methods.