Nonlinear Model Identification for Optimal Operation of an Industrial Cryogenic Air Separation Plant
- Process Systems Engineering
- Focus/Key Topic:
- as of now
During the past decades, the percentage of renewable energy among the total electricity production has significantly increased. However, the electricity generation is thus subject to stronger fluctuations (hourly, daily and seasonally), leading to volatile electricity prices. With so-called Demand Side Management (DSM), large industrial energy consumers try to adjust their electricity consumption and hence their production to fluctuating prices. In particular, the use of DSM poses great challenges for chemical industries.
Common approaches to process operation in chemical industry are based on linear model predictive control (MPC), i.e., a linear process model is identified from plant data, such that the process inputs can be computed by solving an optimization problem. However, linearizations are in general only valid within a certain range. Consequently, this approach seems to be not suitable for significant fast load changes, which are required for an energy price-adapted production. In the Kopernikus-SynErgie project FlexASU, we therefore investigate in cooperation with Linde Engineering, if the use of nonlinear process models could overcome this limitation. Here, the selection of suitable nonlinear models as well as the estimation of respective model parameters from plant data is of particular importance.
This will be the main topic of your thesis. You will start by doing a literature research concerning nonlinear model identification methods. Based on your research, you will select some promising approaches for further investigation. Plant data for estimating model parameters can be generated using a „virtual plant“, that Linde provides. Various tools are possible to assist you, e.g., the MATLAB System Identification Toolbox. Finally, nonlinear model predictive control (NMPC) strategies based on the identified process models are implemented and evaluated.