AMS Solar Energy Prediction: A Comparative Study of Regression Models

The prediction of renewable energies production such as solar energy could improve their integration in the energy mix which is heavily penalised by their intermittent nature. Accurately predicting the total renewable power available helps to control the electricity flows in the grid and therefore guarantee the balance between production and consumption and avoid power overloads or outages. It is in this light that the American Meteorological Society organised a solar energy prediction competition aiming at predicting the total daily solar energy received at 98 solar farms based on the outputs of various weather forecast models. In this paper, a methodology to reach this goal has been explained and the performance of the most used regression models in the field of solar energy prediction has been compared under the same experimental setup. The results have shown that the best performance is obtained using the Ridge regression with a Mean Absolute Error of 2.215E+6.