A Graph Neural Network Approach to Short-Term Electric Load Forecasting

Short-term electric load forecasting for local areas is of interest to Hydro-Québec to efficiently respond to demand at the distribution level. Any significant forecasting error can result in reliability issues, loss of opportunity, or additional costs to the business. Improving short-term load forecasting can benefit both consumers and utility
companies to optimize resources and costs. With the deployment of its Advanced Measurement Infrastructure, Hydro-Québec now has a significant amount of new consumption data that can be used to refine load predictions. In this project, we plan to use recent progress in machine learning to improve the short-term load forecasting.
Specifically, we will use graph neural networks for short-term electric load forecasting at the distribution level to leverage uncovered similarities and latent dependencies between buildings and neighborhoods.

Faculty Supervisor:

Benoit Boulet;Di Wu

Student:

Partner:

Hydro-Quebec

Discipline:

Engineering

Sector:

Professional, scientific and technical services; Utilities

University:

McGill University

Program: