Akademska digitalna zbirka SLovenije - logo
Univerzitetna knjižnica Maribor (UKM)
Knjižnica je med tednom odprta od 8. do 19. ure, ob sobotah od 9. do 13. ure. Čitalnica ČUK je odprta od ponedeljka do sobote od 9. do 24. ure, ob nedeljah od 16. do 24. ure. Informacije: 02 25 07 431, ukm@um.si
  • Simulacije obratovanja hidroelektrarn z uporabo globokih umetnih nevronskih mrež [Elektronski vir] = Simulations of hydropower plants operation using deep artificial neural networks
    Kranjčič, Dalibor
    This paper show practical use of deep artificial neural networks (ANN) for simulation of hydropower plants operated by the Slovenian Drava River Power Company. These networks represent the ... state-of-art trend in the field of artificial intelligence, and they are substantially more powerful than the previous generations of the so-called shallow artificial neural networks, mainly due to the more complex internal structure. The paper is focusing on the research of the effectiveness of these networks by using the black-box approach i.e. identification of the unknown model parameters based on the input/output measurement data, without deepening in the physical characteristics of the identified system. By giving the simulation results of the hydropower plants models in the Keras (Google) deep learning environment, the article describes the neural networks training procedures, with the emphasis on the data selection, extraction and pre-processing. In addition, a detailed description is given of the process information systems that have a crucial role in the preparation of training data and communication protocols, that allow connection with the mathematical/programming environments Matlab/Simulink and Python.
    Vrsta gradiva - prispevek na konferenci ; neleposlovje za odrasle
    Leto - 2019
    Jezik - slovenski
    COBISS.SI-ID - 37956099