Renewable energy forecasts, driven by big data

Short version
I was part of a team of 4 that developed a web app that:

• Predicts future solar and wind energy generation for any point on the planet (that has weather forecasts).
• Uses big data - high-resolution time sequence data from 150+ different sources - to train regression models.
• Provides a user-friendly interface written in Angular javascript.

[code] [pdf]

Long version
For an MSc project, I worked together with Patricia Sauca, Albert Folch, and Xavier Liceras to develop an energy generation forecast application. Our goal was to allow users to see past and future solar and wind energy yields for any location on the planet. We correlated historical weather forecast data with energy yields in those locations and produced several regression models.

For training data, we used 161 sources, consisting of 110 solar farms and 51 wind farms. This data was combined with local weather forecasts and farm configurations (e.g. MWh capacity).

Our final system can be interpreted as a data processing pipeline. The system scrapes pages and queries APIs to collect data, trains models, and exposes them to a user interface.