Control of heating in a low energy single-family house
Authors
Larsen, Kristian Helmer Kjær ; Nielsen, Christian Mølgaard
Term
4. term
Education
Publication year
2022
Pages
133
Abstract
Denne specialeafhandling præsenterer en to-niveaus styringsløsning til opvarmning af et lavenergi-enfamiliehus i forbindelse med OPSYS 2.0. Huset opvarmes med vandbåren gulvvarme forsynet af en varmepumpe, der delvist drives af solceller (fotovoltaiske paneler). Det øverste (planlægnings-)niveau fastsætter et varmebudget ud fra prognoser for energipriser, husets solcelleproduktion og vejret samt en gennemsnitsmodel af bygningens varmedynamik. Det nederste (rumniveau) fordeler dette varmebudget mellem de enkelte rum ved hjælp af en samlet parametermodel af huset opdelt i flere zoner (multi-zone) og vejrudsigter for at opretholde komfortable temperaturer. Til designet er der udviklet diskrete, ikke-lineære grå-boks-modeller for begge niveauer (grå-boks betyder, at modellerne kombinerer fysisk viden med data). Ud fra disse modeller er der konstrueret tilstandsobservatører i form af Kalman-filtre til at estimere ikke-målte tilstande. Begge niveauer anvender modelprædiktiv styring (MPC), som ser frem i tiden og optimerer de kommende styringshandlinger. Ved at omskrive modellerne til den såkaldte mixed logical dynamical (MLD)-ramme kunne styringsproblemerne formuleres som blandet-heltal kvadratisk programmering (MIQP), hvilket gør dem mere effektive at løse. Styringsløsningen er testet ved hjælp af en højdetaljeret husmodel leveret af OPSYS 2.0-deltagerne og viste både forstyrrelsesundertrykkelse og stabilisering af systemet.
This thesis presents a two-level heating control scheme for a low-energy single-family house within OPSYS 2.0. The house uses hydronic (water-based) underfloor heating supplied by a heat pump that is partially powered by solar photovoltaic panels. The upper (planning) level sets a heating budget using forecasts of energy prices, the home’s photovoltaic production, and the weather, together with an average model of the building’s heat dynamics. The lower (room-level) controller allocates this budget across rooms using a lumped-parameter multi-zone model of the house and weather forecasts to maintain comfortable temperatures. For each level, discrete-time, nonlinear grey-box models were developed (grey-box means combining physical knowledge with data). Based on these models, state observers in the form of Kalman filters were designed to estimate unmeasured states. Both levels use model predictive control (MPC), which looks ahead and optimizes future control actions. By reformulating the models in the mixed logical dynamical (MLD) framework, the control problems can be posed as mixed-integer quadratic programs (MIQP) that are solved more efficiently. The control scheme was tested with a high-fidelity model of the house provided by OPSYS 2.0 participants and demonstrated disturbance rejection and stabilization of the system.
[This abstract was generated with the help of AI]
Documents