期刊名称：International Journal of Energy and Environment
印刷版ISSN：2076-2895
电子版ISSN：2076-2909
出版年度：2017
卷号：8
期号：2
页码：195-208
出版社：International Energy and Environment Foundation (IEEF)
摘要：The requirement of the in advance knowledge of the future photovoltaic (PV) production in the domestic field for a better allocation of the on-site PV generation to the local load demand and the available storage facilities is more and more emerging. In this study two different methods were applied so as to forecast the next hour PV power using artificial neural networks (ANN). In the first case the weather parameters of solar irradiance and ambient temperature were predicted, the output was fed to the developed model of the PV installation and the next hour PV power was computed. In the second case it was attempted to predict directly the PV power. The performance of the applied ANNs was compared with the respective outcomes from the persistence models. In each case the applied ANN outperforms the persistence model. In addition, during the evaluation phase the extracted annual energy results were compared with the respective registered data from the installed meters. Again in both cases the results approximated the reality, though in the first case the difficulty in identification and representation of malfunctions in operation of the PV plants due to snow accumulation on the panels caused minor deviations.
