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.


Embedding Predictive Control in Hierarchical Integrated Room Automation Systems (Part 1/2)


T. Baltensperger

Semester/Bachelor Thesis, FS 09

This thesis was written in the framework of the OptiControl project1, whose main goal is to assess the benet of using weather and occupancy forecasts for building climate control.
The tasks of this thesis were to (i) further develop the existing model predictive control (MPC) strategies for the integrated room automation (IRA) such that they are suitable for the use in practical application, (ii) to design new predictive rule-based con- trol (P-RBC) strategies for the IRA and (iii) to assess these strategies by comparison, e. g., to the performance bound (PB) or reference strategies.
Present-day building automation and control (BAC) systems are realized by a hierar- chical control structure with so-called high-level (HLC) and low-level controllers (LLC). Communication between the controllers is based on operation modes (OMs). For that reason an interface was designed which translates the output of a MPC algorithm by rules into OMs. Ideal OM-based low-level control was assumed.
The design of the translation rules was done in an iterative procedure: An initial set of rules was dened and then assessed by simulations. The performance of the new OM- based controller (MMPC) could be measured by comparing it to the original algorithm (CE-MPC). The additional comparison to a rule-based control (RBC) algorithm gave a measure for the potential loss. The main result of the iterative procedure was a set of translation rules as part of the new (high-level) controller MMPC.
Simulations showed that for cases without restrictions on blind movement and with unlimited power for "high-cost" actions the presented set of rules translates the output of MPC to OMs with little cost increase compared to the performance bound (PB). The potential loss of MMPC compared to PB resulting from the translation is negligible. This result allows MPC to t into a conventional BAC setup. Further investigations were executed in order to assess the robustness of these rules.
The translation rules proved to be robust against perturbed weather predictions. How- ever, the performance of the rules depends on how the building parameters are chosen: It was found that in some cases MMPC performs even better with perturbed parameters than with perfect knowledge, whereas in some cases the performance is worse.
In case of power limitations for "high-cost" actions and restrictions on blind move- ment MMPC performs better than CE-MPC in terms of violations but worse in terms of costs. This is because the translation to OMs avoids a part of the violations caused by the MPC algorithm. This aligns with the above results: The translation to OMs adds robustness to the system, however, it also increases the costs.
Approaches for the design of P-RBC strategies are discussed. The presented ideas can easily be assessed with the existing resources. While tasks (i) and (iii) were treated in-depth in this thesis, task (ii) needs to be further explored in order to provide signicant results and conclusions.


Type of Publication:

(13)Semester/Bachelor Thesis

M. Morari

No Files for download available.
% Autogenerated BibTeX entry
@PhdThesis { Xxx:2009:IFA_3435
Permanent link