Abstract
Tremendous growth of solar and wind capacity worldwide has led to rapid development of renewable energy forecasting techniques, and the performance of various forecasting models are usually compared by statistical metrics. However, two factors are not considered in current comparison. One is the amount of the useful renewable energy, and another is the energy consumption during forecasting. These two factors are considered in the proposed new method. Performances of the ANN and RNN models with three different sampling frequencies from literature are compared using this proposed method. Furthermore, the relationships between four statistical metrics-RMSE, (CV)RMSE, E and R 2 -and useful renewable energy are examined. Finally, the effect of adding a moving-window algorithm is tested. The use of the proposed method is expected to promote a virtuous cycle in which sufficient energy is available for our modern lives with fewer emissions of carbon dioxide.