Authors:
Antonios Karatzoglou
1
;
Julian Janßen
2
;
Vethiga Srikanthan
2
;
Christof Urbaczek
2
and
Michael Beigl
2
Affiliations:
1
Karlsruhe Institute of Technology (KIT) and Robert Bosch GmbH, Germany
;
2
Karlsruhe Institute of Technology (KIT), Germany
Keyword(s):
Smart Buildings, HVAC, Thermal Comfort, PMV, Energy efficiency, Model Predictive Control (MPC), Personalization.
Related
Ontology
Subjects/Areas/Topics:
Energy and Economy
;
Smart Grids
;
Smart Homes (Domotics)
Abstract:
There exist two ways of improving the climate conditions within a building; upgrading the building insulation and applying modern heating technology, whereby the combination of both would obviously yield the best result. Recent heating technologies lay high emphasis on forward-looking behavior in order to be capable of providing both more comfort and a higher energy efficiency. Some rely on outdoor and indoor temperature predictive models. Other utilize occupancy prediction. The majority and in particular the ones based on the Predicted Mean Vote (PMV), employ a PMV-driven fixed single temperature point, range (e.g. 22-24C) or curve as reference. In this paper, we introduce a hybrid, personalized heating control approach. It combines a probabilistic occupancy prediction model together with an energy- and subjectified comfort-aware model-based predictive controller (MPC), which can be tailored dynamically to the users’ preference of comfort. Starting with a default PMV and a correspon
ding first temperature set point, our system learns from the users’ interaction with the system’s comfort-driven UI and adapts online the MPC’s target comfort and thereby the MPC’s optimization function respectively. We conducted a user study in a real office environment and show that our dynamic customizable approach outperforms significantly the non-dynamic one in respect of both comfort and energy.
(More)