Developing diffusion models for wind downscaling

This research project aims to develop a ready-to-use model for predicting detailed wind patterns at wind farms to help the industry better plan turbine placements and forecast energy output. Wind speed data is usually available at a large, low-detail scale, which makes it hard to predict specific site conditions accurately. Generating high-resolution wind data using physical models is accurate but is not always practical due to high financial cost, while simpler approaches can lack accuracy. Our project will develop a model by using advanced machine learning models, particularly diffusion models, trained on high-quality simulated data. This approach will make it easier and cheaper to produce detailed wind data at a finer scale. The intern will work on developing these models by including relevant weather variables, using advanced metrics to ensure accuracy, and testing the models’ ability to generate realistic wind patterns.

Faculty Supervisor:

Adam Monahan;Slim Ibrahim

Student:

Partner:

Veer Renewables

Discipline:

Mathematics

Sector:

Professional, scientific and technical services

University:

University of Victoria

Program: