Last month Modelon participated at the 2015 Modelica Conference in Versailles, France, for presenting libraries, tools and scientific papers, as well as for hosting the tutorial “Optimal Control and State Estimation with Modelica and Optimica”.
In this blogpost I write about the content and outcomes of our tutorial.
On September 21 I was one of the experts from Modelon within optimization technology and power generation systems, tutoring our tutorial session at this occasion and supporting attendees in learning our tools.
Compared to other optimization tutorials we have held at previous Modelica conferences, this year we focused more on applications rather than methods, meaning more focus on industrial optimization problems.
We were glad to observe an increased interest from the community, with about 40 attendees learning our optimization tools, which is more than ever before.
The topics of our tutorial included optimal control, both offline and on-line non-linear model predictive control (NMPC), and moving horizon estimation (MHE) of a continuously stirred tank reactor and of a heat recovery steam generator.
This year we put quite a lot of focus on exercises which we developed specially for this purpose. The exercises were carried out using our tool for optimization, the OPTIMICA Compiler Toolkit (OCT), which I also posted about in September.
At the beginning we gave a short introduction to OCT, for which the attendees received an installation package and a temporary license.
After the short introduction, our session's attendees began working with the tutorial examples, including NMPC and MHE, built upon our optimization friendly Modelica library for power plant modeling.
Together with my expert colleagues, we offered instructions and support on how to set up the optimization problems, solve them and plot the results.
One of the exercises concerned the start-up optimization of a heat recovery steam generator (HRSG), containing a boiler, an evaporator, re-heaters and superheaters - see Figure 1.
The challenge for the attendees was to formulate and solve an optimization problem with OCT that drives the HRSG, using boiler and valve controls, to a state where turbines for electricity production could be connected.
A resulting optimized start-up from the tutorial is found in Figure 2.
This shows how the re-heater and superheater pressure and temperature follow their reference values, while temperature gradient constraints are held within limits which must not shorten component lifespans.
In a mid-tutorial demonstration I went through the different steps of this exercise in order to enhance the take-away for our audience.
During the tutorial we received a lot of questions from the attendees, and learned that some of them already have used JModelica.org, which is our open-source optimization tool - a limited version of OPTIMICA Compiler Toolkit.
This led to interesting discussions outside the scope of the tutorial, concerning both optimization and simulations using FMI standard.
For instance, questions on parameter sensitivity analysis were brought up, a type of analysis which is also supported by our tools.
In general, the feedback from our attendees showed that the tutorial was both relevant and useful for their work. We are excited to have contributed to building knowledge for our audience.
Furthermore, the interactive parts in the tutorial were well received and we will look into ways for extending that into future training.
In summary, we can say that this year tutorial was much appreciated by the participants, who learnt that adding optimization to their model-based workflow can improve their result further.
Thank you all who attended our OCT tutorial at Modelica Conference! You all contributed to helping us in the development of an improved course for this tool.
We will be happy to continue together this journey, and to support your further interest in applying OPTIMICA Compiler Toolkit for your engineering applications. Let’s stay in touch!
Per-Ola Larsson holds a Ph.D. in Automatic Control and an M.Sc. in Electrical Engineering, both from Lund University and with focus on optimization, process control and signal processing. At Modelon he is a simulation consultant and project manager in several consultancy and research projects within the field of thermodynamics, containing both steady-state and dynamic models as well as several types of optimizations.