Abstract

As wholesale electricity prices on the Nord Pool day-ahead market become increasingly volatile, shifting a commercial building’s energy consumption to cheaper hours offers significant economic benefits. However, conventional reactive controllers, such as static weather-compensation curves, lack the predictive foresight required to utilize a building’s heavy structural mass as a thermal battery for load shifting. This thesis investigates the potential of implementing a Model Predictive Control (MPC) strategy for a 40 000 m2 commercial office building in Stockholm. The facility features a heavy concrete frame and is equipped with a reversible heat pump plant connected to a borehole thermal energy storage field. To accurately represent the building’s thermal dynamics, four grey-box Resistance Capacitance (RC) model structures were evaluated. Both a single-state lumped model and a three-state model demonstrated high predictive stability, maintaining a 24-hour prediction error below 0.4°C.

The predictive controller minimizes a quadratic cost function based on hourly electricity spot prices, utilizing soft constraints to keep indoor temperatures comfortably between 21°C and 23°C. The proposed MPC and the legacy controller were evaluated over the 2025 calendar year using identical inputs within a thermodynamic sandbox model, ensuring that differences in energy performance reflect true algorithmic advantages.

By proactively pre-heating the structural mass during cheap nighttime hours and curbing compressor usage during expensive morning price peaks, the MPC effectively decoupled comfort delivery from peak energy costs. Under conservative operational adjustments, the MPC achieved a 39% reduction in compressor energy and a 62% decrease in overall operating costs compared to the legacy system. These substantial savings were realized without deteriorating the indoor thermal climate, demonstrating the potential in MPC as a highly viable framework for demand- side management in commercial buildings