Author(s): Ante Sikirica; Marta Alvir; Zoran Carija; Lado Kranjcevic
Linked Author(s):
Keywords: No Keywords
Abstract: The optimal design of hydropower turbines is of great importance as it affects performance, energy production and environmental sustainability. The traditional approach to design often involves complex simulations, which can be time-consuming and resource-intensive. This is suitable for new construction sites, but for existing turbines that are to be revitalised, this investment is often difficult to justify. In addition, revitalisation efforts are usually focused on a specific part of the turbine. In this context, machine learning techniques offer a promising way to simplify the design process. Our research focuses on the application of machine learning in conjunction with optimisation techniques to improve the design of the draft tube and other parts of turbines that need to be revitalised. We propose an optimisation workflow that leverages deep neural network surrogates, which can be complemented by in situ measurements.
Year: 2024