Master's Thesis Adrian Marx

 

Automated Model Generation for Model Predictive Exergy-Based Control of HVAC Systems

In the Master’s Thesis at hand, a method for automated model parameterization in Heating, Ventilation
and Air Conditioning (HVAC) Systems is developed and validated.
Firstly, I implement standardized models of components HVAC systems typically comprise of in
Modelica. For those models, a MatLab script, which is capable of automatically conducting identification
experiments, is developed and implemented. Data records obtained during these experiments
are used for parameterization of the designed models.
Together, models and script can be integrated into an existing exergy-based control algorithm. This
enhanced algorithm can utilize the parameterized models to control the considered HVAC system.
In order to prove the functionality of the algorithm, software-in-the-loop (SIL) experiments are conducted,
using a partial model of the HVAC system as a source for the training data. These experiments
show that the algorithm is generally capable of finding the parameters that are used in the
partial model and fitting the models to the training data.
After validation, identification experiments are conducted for all the designed models, using the real
HVAC system as data source. Using the obtained data records, the models are fitted to the behavior
of the real HVAC system. The results show that the parameterized models are not only capable of
reproducing the training data, but also predicting the behavior of the systemfor other data records.
On the other hand, a strong dependency of the parameterization on the quality of the data records
is observed.
The work concludes with a generalization of the results and a discussion of the future prospects.