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.

  

Stochastic Nonlinear Model Predictive Control for Building Climate Control

Author(s):

M. Hofer
Conference/Journal:

Semester Thesis, FS14 (10347)
Abstract:

A novel Stochastic Nonlinear Model Predictive Control algorithm is implemented and applied in the eld of building climate control. Buildings are subject to uncertain disturbances, such as the inuence of the weather or internal gains caused by occupants and equipment. Minimizing the energy consumption, while satisfying comfort level constraints with a certain (time averaged) probability, can be formulated as a chance constrained Finite Horizon Optimal Control Problem. The ccFHOCP is in general computationally intractable and therefore needs to be approximated by the scenario approach. The uncertainty is replaced by a nite number of sampled disturbances, without making any assumption on the underlying distribution. Dierent to scenario-based Model Predictive Control, a bilinear model is incorporated for this study, leading to non-convex optimization problems. The Randomized Nonlinear MPC algorithm (RNMPC) requires the construction of a convex hull to ensure the optimal input to be probabilistically feasible. We investigate and benchmark in open-and closed-loop simulations during a representative month in winter and summer. The results are compared to existing Randomized Linear MPC (RLMPC) approaches, which are based on linearization techniques. In the open-loop case, the RNMPC algorithm leads to a higher control cost during winter and a slightly lower cost in summer compared to the RLMPC approach. However, the closed-loop cost is signicantly reduced during winter when incorporating the bilinear model (RNMPC), while it remains comparable in the summer period. The RNMPC algorithm achieves a higher and hence less conservative constraint violation probability, while remaining below the predened threshold. iii

Year:

2014
Type of Publication:

(13)Semester/Bachelor Thesis
Supervisor:



File Download:

Request a copy of this publication.
(Uses JavaScript)
% Autogenerated BibTeX entry
@PhdThesis { Xxx:2014:IFA_4959
}
Permanent link