Mikkel Lindstrøm Sørensen: Forecasting of renewable energy production
The consequences of global warming is gradually becoming more and more apparent, and hence the world is increasingly looking to renewable energy sources to minimise the societal impacts of e.g. flooding and increased sea level. The main renewable energy sources being implemented around the world is wind and solar power. However, integrating these into the power grid poses a number of problems since these sources are stochastic in nature, i.e. they only produce power as the wind blows, or as the sun shines. This uncertainty being introduced into the grid needs to be reduced and quantified to simplify the green transition and provide more information for the end user.
Here methods such as probabilistic forecasting can come into play. Most methods currently in use for forecasting the power production provide point forecasts, i.e. the forecast only provides a single value for the expected power production. By instead using probabilistic forecasting methods, we will be able to provide more reliable information than the single point value, such as uncertainties and the distribution of the forecast.
Evaluation of probabilistic forecasts will thus also be an important topic. When evaluating the quality of a forecast one can apply a number of different methods and measures that focus on different aspects of the forecast, but the real value will also lie in the financial gain from applying the forecast. Hence, a mapping from the probabilistic forecast to the financial gain based on (predictions of) market prices will be addressed.
Lastly, the world of today is also ever-changing, and information therefore needs to be available in real-time. Hence methods for real-time forecasting needs to be dynamic and able to adapt to the changing conditions imposed by e.g. dirtiness of PV panels, changes in the roughness of the surrounding terrain, or changes in meteorological models.
This project aims at addressing these issues as well as bridging the gap between the academic work described above and methods suited for real-life implementations.