Decision tree classification of daily total photovoltaic energy generation data considering meteorological conditions

Abstract

With the increasing penetration rate of photovoltaic systems in the power grid, there has been a remarkable interest in photovoltaic power prediction. Since, accurate predictions handle the volatile characteristics of solar energy, and prevent reliability issues and economic losses in the power system. To this end, in this paper, daily total energy generation data of a photovoltaic system is classified for the purpose of evaluating the effects of different meteorological inputs. Daily measurements of total global solar radiation, average temperature, average relative humidity and average wind velocity are processed in the decision tree classifiers developed. The classification results not only show the promising usage of daily total global solar radiation variable within a 3-tupled input structure, but also illustrate its capability in improving the classification accuracy.