2019 Solar Energy Systems Conference

Solar Farm Power Forecasting Using Deep Learning

Authors

Arnie de Castro - Presenter, SAS Institute Inc.
Tae Yoon Lee, SAS Institute Inc.
Yun Tae Hwang, Georgia State University
Hye Yeon Lee, Tmoney Co. Ltd.
Solar energy is the third largest renewable energy source in the world after hydro and wind power. Its role is expected to become even more significant as technology allows the production of cheaper and higher efficiency units, resulting in increased adoption and tremendous impacts on grid operations. Forecasting solar energy is therefore becoming more and more important.

Deep Learning models using RNN, LSTM, GRU are gaining popularity for electrical Load forecasting. The accuracy obtained with these techniques vary depending on the forecasting horizon. Different Deep Learning models work for Very Short Term (VST), Short Term (ST), and Long Term (LT) forecasting. In this research, we test various Deep Learning models at different time steps for solar farm power forecasting for each of these time frames, comparing with Machine Learning models of Random Forest, Gradient Boosting and Neural Networks.