PhD position in Machine learning for turbulence modelling in wind farms

Updated: over 2 years ago
Job Type: Temporary
Deadline: 16 Oct 2021

TU Delft is a top tier university and is exceedingly active in the field of Artificial Intelligence. The AIFluids lab was recently established to foster the use of AI in the Aerospace Sciences. Designing more efficient aircrafts and wind farms requires a deeper understanding of complex flows. The AIFluids Lab is focused on two major challenges of fluid mechanics: the prediction and the control of complex, transitional and turbulent flows.

New experimental techniques and high-fidelity flow simulations are providing larger and more detailed datasets. Using AI algorithms, the AIFluids Lab leverages numerical and experimental data to build interpretable models of transition and turbulence. Our objective is to combine human and machine insights to get to the essence of complex physical flows.

This research project focuses on the modelling and control of flows found in wind farms, including effects of atmospheric boundary-layer stratification, wake development and interaction with downstream turbines, and the effect of power production and loads. Only with a good understand of these flows, including the ability to predict them reliably, can effective control strategies to be developed to achieve target power production and minimize loads (extending lifetime and reducing maintenance costs). This is necessary to scale wind-energy to a dominant position in energy supply.

LES is an accurate model for these flows, but is too costly to be usable within a design or control loop. In this project we will look at ways of achieving similar physical fidelity at reduced cost, by enhancing the accuracy of physical models with partially or completely modelled turbulence – e.g. RANS, URANS and PANS. In particular, we will use symbolic supervised machine learning methods to develop novel turbulence closures for these models, that outperform general purpose models for these flows. These we will couple with structural models, and use for development of wind farm control strategies.



Similar Positions