Abstract
Predicting short-term solar irradiance is important to introduce solar power in an electric power system. Considering that large quantities of solar power will be introduced across wide areas in the future, this paper addresses multi-point predictions of solar irradiance. Support vector machines (SVMs) perform well for predicting solar irradiance, but involve relatively high computational complexity. In addition, prediction models should be updated regularly to adapt to the season. Here, we introduce a multi-point prediction system that reduces the amount of calculation. Rather than constructing prediction models at all grid points, this system creates certain clusters based on the dynamic time warping (DTW) distance, and then constructs a small number of models representing each cluster. Mathematically, this consists of three algorithms, DTW, cluster analysis and SVM.
