Causal Inference for Windmill-Based Electricity Production

Description model for windmill electricity production based on causal inference
Students: Preben Bruntse Nielsen
Supervisor: Philippe Bonnet
Level: MSc, Semester: Spring 2020
Tags: Causal Inference, Energinet, Electricity Forecasting

We are in the middle of the transition to more renewable energy. A lot of the energy in Denmark comes from windmills, whose electricity production varies depending on the weather. To be able to properly incorporate the electricity from windmills in the energy markets it is important to be able to forecast the electricity generated from parks of windmills. Energinet has developed a model for describing the electricity production on a single windmill based on historical data. Forecasting the electricity production of a single windmill based on weather data, image data, or other data sets is an open issue. The goal of this project is to design, implement and evaluate a model for forecasting electricity production that uses additional datasets to improve the prediction accuracy. To find which types of data is having an effect on the energy production, causal inference will be used.

The intended learning outcomes of this project are to: Explore the Energinet data set used for windmill electricity production

Define and test a causal model describing the relationship between the windmill electricity production and external factors

Design, implement and evaluate a description model for windmill electricity production based on causal inference.

Reflect on how such a descriptive model can be used to derive predictive models.