Abstract
Estimation of the solar radiation level reaching a specific zone on the surface of earth is a crucial step in the design and planning of solar energy systems. The large number of parameters affecting the estimation and prediction processes mandates dimension reduction of the input feature space. In this paper, we address this problem for a prediction system in which uncertainties play a major role. We propose an adaptive memory programming-based approach to optimize the input feature space of a solar radiation predictor. The fitness values of reducts are calculated using granular computing. The attribute reduction concept in the rough set theory is invoked and the dependency degree is used as a fitness function. The proposed methodology is evaluated using a large environmental temporal dataset collected for regions that exhibit diverse climate conditions.