Testing and demonstration of model predictive control applied to a radiant slab cooling system in a building test facility

Abstract

Radiant slab systems have the potential to significantly reduce energy consumption in buildings. However, control of radiant slab systems is challenging. Classical feedback control is inadequate due to the large thermal inertia of the systems and heuristic feed-forward control often leads to unacceptable indoor comfort and may not achieve the full energy savings potential. Model predictive control (MPC) is now attracting increasing interest in the building industry and holds promise for radiant systems. However, an often-cited barrier to its implementation in the building industry is the high computational cost and complexity relative to the feedback controls used in conventional systems. The objectives of this study were to (i) verify the correct operation of an open source MPC toolchain developed for radiant slab systems, and (ii) demonstrate its efficacy in a test facility. A matched pair of cells in the FLEXLAB building test facility at the Lawrence Berkeley National Laboratory was used in the study. The proposed MPC toolchain was implemented in one cell and the performance compared to that of the other cell, which used a conventional heuristic control strategy. The results showed that the simplified MPC approach applied in the toolchain worked as expected and realized energy savings over the conventional control strategy. The MPC yielded 42% chilled water pump power reduction and 16% cooling thermal energy savings, while maintaining equal or better indoor comfort.

Indoor Temperature Control of Radiant Ceiling Cooling System Based on Deep Reinforcement Learning Method

“…MPC could consider multiple environmental parameters and suppress disturbances, which had received widespread attention and had been proven to be an effective control method [10][11][12]. Pang et al [13] established MPC arithmetic based on radiant cooling experiments with large thermal inertia radiant slab and compared it with traditional heuristic control, and the MPC achieved better energy-saving and thermal comfort effects. Joe et al [14] established a MPC arithmetic to achieve dynamic estimation and prediction based on regional load and temperature, outdoor weather and HVAC system models, and the results showed that MPC saved 34% of the cost and 16% of the energy compared to baseline feedback control.…”

R&D and Implementation Outcomes from the U.S.-India Bilateral Center for Building Energy Research and Development Program

“…The Cool Roof Calculator tool calculates energy savings and life cycle costs for solar reflecting envelope materials and optimizes roof insulation [21] as both a new construction and retrofit opportunity, making it relevant for both countries. The Model Predictive Control (MPC) software toolchain uses artificial intelligence and radiant cooling strategy [22] to help solve this complex controls challenge and bridge the design-operations gap. ECBC Code Compliance Ruleset for the Indian Energy Conservation Building Code (ECBC) is incorporated in performance-based compliance software, and implemented in DOE's OpenStudio modeling environment.…”

Neural network, ARX, and extreme learning machine models for the short-term prediction of temperature in buildings

“…For residential buildings, it has been shown that heating demand can be decreased by up to 25% in the case of high indoor comfort criteria, and by up to 49% in the case of low indoor comfort criteria (Smarra et al 2018). A 26% reduction in heating demand in residential buildings, by using the adaptive predictive control of thermo-active building systems, was presented in (Schmelas et al 2017), whereas in (Pang et al 2018) a 16% reduction in the cooling demand in case of an applied MPC to a radiant slab cooling system was reported. In commercial buildings, the implementation of MPC in building services has resulted in yearly energy savings of between 31% and 36% (Killian and Kozek 2018).…”