Author(s): Arthur Favrel; Ghofril Kahwati; Quentin Dollon

Linked Author(s):

Keywords: Cavitation surge; Hydropower units; Machine learning; Data clustering

Abstract: In this paper, a novel approach for identifying and modelling the onset of cavitation surge in hydropower units is proposed. For the considered test case, three regimes of cavitation surge are identified through data clustering based on experts’ criteria. This clustering makes it possible to predict the onset conditions of each regime as a function of the unit’s operating parameters by using a Machine Learning classifier algorithm. The influence of the surge frequency and amplitude on the response of both the shaft line and generator is also highlighted. Finally, the potential for automatizing this approach through data dimension reduction and clustering algorithms is briefly explored and discussed.

DOI: https://doi.org/10.1088/1755-1315/1483/1/012005

Year: 2023