Timezone: »

 
What if We Enrich day-ahead Solar Irradiance Time Series Forecasting with Spatio-Temporal Context?
Oussama Boussif · Ghait Boukachab · Dan Assouline · Stefano Massaroli · Tianle Yuan · Loubna Benabbou · Yoshua Bengio

@
in The Synergy of Scientific and Machine Learning Modelling (SynS & ML) Workshop »
Event URL: https://openreview.net/forum?id=dL5BQsgSMh »
The global integration of solar power into the electrical grid could have a crucial impact on climate change mitigation, yet poses a challenge due to solar irradiance variability. We present a deep learning architecture which uses spatio-temporal context from satellite data for highly accurate day-ahead time-series forecasting, in particular Global Horizontal Irradiance (GHI). We provide a multi-quantile variant which outputs a prediction interval for each time-step, serving as a measure of forecasting uncertainty. In addition, we suggest a testing scheme that separates easy and difficult scenarios, which appears useful to evaluate model performance in varying cloud conditions. Our approach exhibits robust performance in solar irradiance forecasting, including zero-shot generalization tests at unobserved solar stations, and holds great promise in promoting the effective use of solar power and the resulting reduction of CO$_{2}$ emissions.
Keywords: context-enriched learning · multi-modal learning · Time series forecasting · solar irradiance

Author Information

Oussama Boussif (Mila, Quebec AI Institute)
Oussama Boussif

I am a visiting research student at Mila working with Professors Yoshua Bengio and Loubna Benabbou on Machine Learning Applied to Physics. Prior to this, I obtained my engineering degree in Applied Mathematics and Data Science from CentraleSupélec (2021) and a Msc. in “Mathematics, Vision & Learning” (2021) from ENS Paris-Saclay (commonly known as MVA). I am interested in discovering physical laws and processes in a data-driven way. I am also interested in finding physically informed algorithms for scientific applications such as material discovery.

Ghait Boukachab (Montreal Institute for Learning Algorithms, University of Montreal, Université du Québec à Rimouski)
Dan Assouline (Mila, Université de Montréal)
Stefano Massaroli (Mila)
Tianle Yuan (NASA GSFC)
Loubna Benabbou (University of Quebec UQAR)
Yoshua Bengio (Mila - Quebec AI Institute)

More from the Same Authors