Graph Convolutional Neural Networks for Solar Irradiance/Weather Prediction

Predicting solar irradiance is a major new energies business problem. It is, however, extremely complex.

To be able to predict solar irradiance, weather and wind forecasts come into play. Sky-cameras (cameras looking skyward in specific areas) can be another source of information.

Now there are two types of information involved: temporal (time-series), and spatial (based on the location of weather station, cameras, etc.)

Graph neural networks (GNNs) allow for combining these spatio-temporal information into one problem setting and outperform traditional disjoint approaches.

There an open space here for innovation with the potential to have access to proprietary (hours of sky-camera) data.

Requirements

Experience with machine learning and statistical signal processing.

Contact

dr. Hadi Jamali-Rad

Signal Processing Systems Group

Department of Microelectronics

Last modified: 2021-03-29