master's thesis
OPTIMALIZATION HYDROTECHNIC SYSTEM BY NEURAL NETWORK

Antun Crnolatac (2015)
Josip Juraj Strossmayer University of Osijek
Faculty of Civil Engineering
Department for Hydrotechnics and Ecology
Metadata
TitleOPTIMALIZACIJA HIDROTEHNIČKOG SUSTAVA POMOĆU NEURALNIH MREŽA
AuthorAntun Crnolatac
Mentor(s)Marija Šperac (thesis advisor)
Abstract
U svrhu boljeg upravljanja hidrotehničkim sustavom potrebno je optimalizirati dobivene parametre hidrološke postaje Donji Miholjac C.S., primjenom neuralnih mreža. Neuralne mreže se mogu usporediti sa umjetnom inteligencijom. Umjetna inteligencija je termin koji se dodaje svakom neživom sustavu koji ima sposobnost snalaženja u novim situacijama. Za rad s neuralnim mrežama korišten je računalni program Weka. Programom je obrađena povezanost oborine i protoka na postaji Donji Miholjac. Ulazni podaci su bili izmjerene vrijednosti oborina od strane DHMZ – a, a izlazne vrijednosti su bili protoci hidrološke postaje Donji Miholjac, C.S. Računalni program Weka ostvaruje više različitih oblika algoritama korištenih za predviđanja, kao što su stabla odluke, pravila klasifikacije, neuronske mreže i slično. Nudi optimalno rješenje s pripadajućim modelom koji predstavlja izbor najpovoljnijeg upravljanja, a što ne znači da mora biti optimalno po svim kriterijima. Optimalizacijski model mora imati analitički definiranu funkciju cilja, u okviru sasvim jasno definiranih kriterija za vrednovanje svake upravljačke odluke. Rezultati koji se dobiju putem neuronskih mreža prikazuju koliko pojedini parametar utječe na uspješnost pojedine neuronske mreže. Predviđanje rezultata na dnevnoj bazi pruža znatno veće mogućnosti u svrhu zaštite od poplava. Upravo iz tog razloga je potrebno konstantno praćenje i mjerenje vodotoka, da se može pravilno intervenirati u slučaju takvih prirodnih katastrofa. Mogućnost predviđanja poplava se sve više povećava i olakšava s obzirom na to kako tehnologija i znanost sve više napreduje. Poplave su svuda u svijetu, pa tako i u Hrvatskoj sve učestalije, intenzivnije i opasnije. Ne mogu se spriječiti, ali se poduzimanjem učinkovitih, preventivnih i operativnih mjera njihove štetne posljedice mogu značajno ublažiti.
Keywordshydraulic system neural networks Weka
Parallel title (English)OPTIMALIZATION HYDROTECHNIC SYSTEM BY NEURAL NETWORK
Committee MembersLidija Tadić (committee chairperson)
Marija Šperac (committee member)
Brankica Malić (committee member)
GranterJosip Juraj Strossmayer University of Osijek
Faculty of Civil Engineering
Lower level organizational unitsDepartment for Hydrotechnics and Ecology
PlaceOsijek
StateCroatia
Scientific field, discipline, subdisciplineTECHNICAL SCIENCES
Civil Engineering
Hydrotechnology
Study programme typeuniversity
Study levelgraduate
Study programmeUniversity graduate study in civil engineering; specializations in: Hydraulic engineering, Supporting structures, Construction management and technology, Transportation engineering
Study specializationHydraulic engineering
Academic title abbreviationmag.ing.aedif.
Genremaster's thesis
Language Croatian
Defense date2015-09-25
Parallel abstract (English)
For the purpose of better management of water development projects it is necessary to optimize the parameters obtained by hydrological station Donji Miholjac CS, with the application of neural networks. Neural networks could compare with artificial intelligence. Artificial intelligence is a term that is added to each non-living system that has the ability to cope with new situations. To work with neural networks we used a computer program Weka. A computer program Weka achieves more different forms of algorithms used for prediction, such as decision trees, classification rules, neural networks, etc... It offers us the optimal solution with the corresponding model which is a selection of the best management, which does not mean it has to be optimal by any standards. The optimization model must be analytically defined objective function within clearly defined criteria for the evaluation of each management decision. The results obtained by neural networks give us display how individual parameter affects the performance of individual neural networks. Prediction results on a daily basis provides a significantly greater potential to protect against flooding. For this reason, it takes constant monitoring and measurement of water flow, that it can properly respond to such natural disasters. The ability to predict floods is on the increase and facilitate with regard to how technology and science are increasingly advancing. Floods are everywhere in the world, including Croatia and more frequent, more intense and more dangerous. It can not be prevented, but the taking of effective, preventive and operational measures of their harmful effects can be significantly mitigated.
Parallel keywords (Croatian)hidrotehnički sustav neuralne mreže Weka
Resource typetext
Access conditionOpen access
Terms of usehttp://rightsstatements.org/vocab/InC/1.0/
URN:NBNhttps://urn.nsk.hr/urn:nbn:hr:133:549946
CommitterVesna Zobundžija