Probabilistic Forecasting of Solar Power: An Ensemble Approach

Abstract

Probabilistic forecasts account for the uncertainty in the prediction that arise due to errors in real-time measurements and the inherent uncertainty of a prediction model. With the emergence of renewable technologies and variable power generation depending on the weather conditions, probabilistic forecasts have gained popularity. Wind power is a mature technology and is in place for decades now, various probabilistic forecasting techniques are used in this domain. On the other hand solar power is an emerging technology and as the technology matures there will be a need for forecasting the power generated days ahead. In this study, we use state-of-the-art machine learning algorithms to generate month ahead solar power forecasts. It is observed that having multiple models gives exceedingly better results when compared to individual models. Also running separate models on the data belonging to the same hour vastly improved the results. Getting accurate forecasts will help the decision makers take optimal decisions when large scale solar farms are integrated into the power grid.

Date of Award	May 2015
Original language	American English
Supervisor	U Zeyar Aung (Supervisor)