About the project or challenge area

Wind-flow models are key tools in describing how a wind resource varies across a wind farm project area. Existing commercially popular models are all physics-based. Models include linear models, such as WAsP, and more complex CFD models (Computational Fluid Dynamics). Linear models have the advantage of easy setup, quick calculation time of estimates, and low cost of software and computation. The physics of the atmospherics flow have been simplified in order to obtain these advantages. Unfortunately, linear models are only effective in simulating flows in simple terrain. Due to the physic modelling assumptions, they struggle in complex environments. CFD have the ability to generate accurate wind flow estimates in complex terrain, but can be difficult to properly initialize, have long computational run times and are expensive. These factors can make CFD models commercially prohibitive.

We propose an alternative machine learning (ML) data-based model. The ML model will provide the wind industry with a tool that is computationally quick, easy to use, opensource based and delivers reliably accurate wind speed and energy estimates. Utilizing advances in deep learning and long short-term memory (LSTM) neural networks, the algorithm will be trained on adjusted and validated WAsP runs from various on-shore and off-shore wind project sites. The model will be implemented in open source software, primarily Python and TensorFlow. The resulting flow model will require the traditional wind resource modelling inputs: roughness orography and wind frequency. The model will output a wind resource grid. Results validation comparisons of the ML model with WAsP and the NREL opensource model will be computed for a variety of publicly available reference sites including Askervein, Bolund and others.

Why choose this project?

The project will allow you to conduct state-of-the art machine learning research in partnership with DNV GL, a world leading renewable energy consultancy. 

About you

This project would suit a candidate with strong interests in wind energy, machine learning and flow modelling.

How to apply

All students can apply using the button below, following the Admissions Statement (PDF, 188kB). Please note that this is an advertised project, which means you only have to complete Section A of the Research Statement.

This project is not funded, for further details please use this link.

Before applying, we recommend getting in touch with the project's supervisors. If you are interested in this project and would like to learn more about the research you will be undertaking, please use the contact details on this page.