Transfer learning in a Solar Radiation Forecasting context

This project is part of a larger effort towards the decarbonization of our energy by focussing more on renewable energy (eg. solar, wind). Hydro-Quebec is invested in this effort as it is a major provider of electricity throughout Northeastern America. Many renewable energy sources are highly variable; it is therefore important to be able to predict the behavior of these energy sources ahead of time. The goal of this project is to automate the prediction of solar radiation on the ground, which is a proxy for the amount of photovoltaic energy that can be produced. This will be done using satellite images and deep learning methods.

Faculty Supervisor:

Ioannis Mitliagkas

Student:

Partner:

Hydro-Quebec (Varennes, QC)

Discipline:

Computer science

Sector:

Utilities

University:

Université de Montréal

Program: