Vinsamlegast notið þetta auðkenni þegar þið vitnið til verksins eða tengið í það: http://hdl.handle.net/1946/7764
This Master Thesis investigates the possibility of using the surrogate modeling techniques in the complex energy system modeling problems. The different mathematical (nearest-neighbor, linear, spline, cubic Hermite and polynomial interpolation as well as polynomial fitting) and artificial intelligence (neural network) methods were introduced and implemented.
Prepared paper includes information from many different fields of science: energy science, mathematics and computer science. All of them were used to prepare multi-science analysis of surrogate modeling problem.
Several different surrogate models were created for two different energy systems. First of them was gas turbine with recirculation, syngas production and CO2 capture; second system was the steam network. The energy system descriptions were provided in the Thesis.
Implemented models were analyzed and their errors were found. The results helped in generalizing the features of surrogate models of each type. The conclusion are focused on the following topics: methods of choosing the best initial sets of points, method of elimination errors in the training process, validity space of the model and possibilities of connecting two algorithms: genetic optimization and creation of surrogate model. Gathered information will be used in the second part of the project – implementing the generic tool to create surrogate models for any energy system model. The second part of the project will be realized in Poland at Jagiellonian University in cooperation with École Polytechnique Fédérale De Lausanne. In the Thesis, the process of creating the automotive computer tool is designed. Moreover the possible positive influence, not only for Poland, but for any problem examined by this tool, is described.