Note: This content is accessible to all versions of every browser. However, this browser does not seem to support current Web standards, preventing the display of our site's design details.

  

Use of weather and occupancy forecasts for optimal building climate control (OptiControl): Two Years Progress Report – Main Report

Author(s):

D. Gyalistras, M. Gwerder
Conference/Journal:

Terrestrial Systems Ecology ETH Zurich R&D HVAC Products, Building Technologies Division, Siemens Switzerland Ltd., Zug, Switzerland, pp. 158
Abstract:

This report presents the work conducted during the first two years of the OptiControl project (www.opticontrol.ethz.ch), an interdisciplinary project dedicated to the development of predictive control technologies for buildings. The overall aim is to minimize the energy usage of buildings whilst maintaining or even improving occupant comfort and reducing peak electricity demand. Here we report on project Phases I and II that dealt with the development of methods, tools and control strategies, the assessment of the potential of predictive control, and the in-depth analysis of selected cases. This work was based entirely on computer simulations. In Phase III the newly developped control approaches shall be tested in a demonstrator building. Integrated Room Automation (IRA) for office buildings was identified as a promising candidate for the investigation of predictive control strategies. IRA deals with the automated control of blinds, electric lighting, heating, cooling, and ventilation of an individual building zone or room. Most of the work undertaken within the OptiControl project so far has focused on this application. A comprehensive set of criteria covering all aspects of IRA control solutions was defined to provide guidance for the research and development work. The criteria ranged from control performance and requirements of different users to marketing potential. Based on these criteria and on further considerations Rule-Based Control (RBC) and Model Predictive Control (MPC) were identified as the most promising control approaches. A generic framework for the assessment of control performance was developed that uses the so-called Performance Bound (PB) as an absolute benchmark. The PB is a theoretical value and presents the lowest achievable control cost (in terms of energy or money) for a given building, cost function, disturbances (weather, internal gains), and set of comfort requirements. It is estimated by assuming perfect knowledge of the building system and all disturbances acting upon it. The difference in control cost between the best currently known non-predictive controller and the PB presents the theoretical savings potential of predictive control. The realizable potential in practice will always be smaller since every real controller will show higher costs than the PB. The IRA control task was formally defined in all its facets: Relevant building types, types of heating, cooling, ventilation, blind and lighting subsystems, control operation types, and representative building locations were identified, the subsystems were sized properly, and meaningful energy usages/costs were specified. The hierarchical architecture of modern Building Automation and Control systems was considered from begin on in order to ensure that the solutions developed could be easily integrated therein later on. Four non-predictive RBC strategies plus associated procedures for the automated tuning of their control parameters were identified or newly designed and implemented for use in simulations: RBC-1, a state-of-the-art strategy; RBC-2: same as RBC-1, but allowing for continuous blind transmission values and for time-continuous (rather than event-triggered) repositioning of blinds; RBC-3: an entirely new strategy that, instead of working with threshold values, uses historical heat and cold demand signals and historical room temperature data, and that also allows for maximum freedom in blind movement; and RBC-4: same as RBC-3, but with blinds repositioning restricted to once per hour. A new family of Model Predictive Control (MPC) strategies was developed that was tailored to the needs of building control: PB, an algorithm that estimates the PB with the aid of MPC; Certainty Equivalence (CE), a controller that uses imperfect models and/or disturbance predictions but treats them as if they were correct (i.e. equal to certain); and Chance Constrained Stochastic MPC (SMPC), an enhanced approach that may or may not involve perfect models but takes the uncertainty in the disturbances predictions, in particular of weather predictions, into account. A new model for the coupled thermal, light and air quality dynamics of a single building zone was developed, tested and validated. The model was used to simulate the behavior of real buildings under different control strategies, and to describe the building system’s dynamics for MPC. It was a 12th order multiple-input-multiple-output bilinear model with a resistance-capacitance network representation of thermal energy fluxes. Analysis of the approximations employed and a comparison with a detailed radiative-convective model suggested that the model delivers accurate and reliable results. Hourly weather data from 10 representative European SYNOP measurement sites for the years 2001 (or later) to 2007, predictions by the COSMO-7 numerical weather prediction model for the years 2006 and 2007, and various Design Reference Year datasets (special, representative datasets compiled from long term climate observations) were prepared as an input for building simulations. Algorithms for the disaggregation of hourly global radiation into the direct and diffuse part, and for the derivation of global radiation components on vertical oriented surfaces were implemented. Novel statistical post-processing methods were developed and applied to improve local predictions of the most important weather variables for the building control applications under investigation. In general, forecast biases could be successfully removed on a seasonal basis. The root mean square error of local temperature predictions for the first 24 hours ahead was reduced by 20–30%. For wetbulb temperature the reduction was 35–45%. For the radiation components no reductions or slight increases were obtained for winter and summer, but reductions of 10–60% were achieved for spring and autumn. These improved data sets were made available for building simulations. The theoretical savings potential of predictive control was assessed by comparing the performance of the controllers RBC-1 to RBC-4 with the PB. The found savings potentials were put into context by comparing them with possible energy savings due to the following low-cost measures related to control: a) a reduction of the thermal comfort when the building is not used, by allowing for room tem-perature set-backs during nights and weekends (base case: no set-backs allowed); b) a general reduction of thermal comfort due to a widening of the room temperature comfort range by ~1.5 degC (base case: narrow comfort range); c) the use of Indoor Air Quality controlled ventilation (base case: application of a constant minimum fresh air supply rate according to a fixed occupancy schedule); d) the adjustment of the control such that it optimized control actions for energetic rather than monetary cost (base case: optimization of control for money). Conducted was a large-scale factorial simulation experiment (~23’500 whole-year, hourly time step dynamic simulations) considering 64 building/room types (differing in façade orientation, construction type, building standard etc.), 5 building systems (S1–S5, employing different heating, cooling, and ventilation subsystems), 2 “cost” functions (Non-Renewable Primary Energy [NRPE] usage, and monetary costs), 4 different building sites, 4 thermal comfort definitions, and 2 ventilation strategies. Annual total costs and annual comfort indices were analyzed by building system, building standard (PA–“Passive House”, or SA–“Swiss average”), and building class (I–“very frequent”, II – “less frequent”, III–“exotic” building case). RBC-3 (time-continuous repositioning of blinds and perfect luminance control via blind operation) proved clearly as the best performing non-predictive controller and in many cases it came very close to the PB. The average absolute (relative) theoretical NRPE savings potential over the building classes I and II was 2.6 kWh/m2/a (9.2%) for the PA building standard and building system variants S1-S5, and 3.8 kWh/m2/a (9.7%) for the SA building standard and building system variants S1-S3. Much larger theoretical savings potentials were obtained for the RBC-1 and RBC-4 controllers that employed much more realistic assumptions on repositioning of blinds as compared to RBC-3: average maximum possible NRPE savings (building classes I and II, building system variants S1–S5) for these two strategies were 34% and 33% for the PA, and 30% and 23% for the SA building standard, respectively.

Further Information
Year:

2010
Type of Publication:

(04)Technical Report
Supervisor:



File Download:

Request a copy of this publication.
(Uses JavaScript)
% Autogenerated BibTeX entry
@TechReport { GyaGwe:2010:IFA_4001,
    author={D. Gyalistras and M. Gwerder},
    title={{Use of weather and occupancy forecasts for optimal building
	  climate control (OptiControl): Two Years Progress Report –
	  Main Report}},
    institution={},
    year={2010},
    number={},
    address={Terrestrial Systems Ecology ETH Zurich R\&D HVAC
	  Products, Building Technologies Division, Siemens
	  Switzerland Ltd., Zug, Switzerland},
    url={http://control.ee.ethz.ch/index.cgi?page=publications;action=details;id=4001}
}
Permanent link