Generating Power Forecasting Of Photovoltaic System Based On GRA And LS-SVM

As the global energy shortage and security issues become increasingly prominent, the development of new energy and renewable resources has become the consensus around the whole world. However, for the power output of photovoltaic system is affected by different weather conditions, it is uncertain and cyclical. Accurately predict the PV system output power is of great significance for ensuring the stability of the grid and promoting the development of large-scale PV systems.I On the basis of reading a large number of domestic and foreign literatures, this paper studied photovoltaic grid power prediction from four different aspects. By analyzing the related factors influencing the photovoltaic grid power and data mining technology, the paper selected similar characteristics of meteorological data from a large amount of data and using the grey correlation analysis(GRA) theory to forecast the photovoltaic power. We choose the main factors include irradiance, temperature, humidity which influence PV generation heavily as the input variables of least squares support vector machine(LS-SVM) forecasting model to predict the output power of PV systems on head of 24 hours. Considering the merits and demerits of GRA method and LS-SVM method, the paper propose a method of GRA and LS-SVM in parallel and a method of GRA and LS-SVM in series.In this paper, all the data are from the grid-connected PV monitoring system of Tianjin University. Based on the data, we establish four prediction models to predict for sunny, cloudy, rain and fog days respectively. The experiments show that the prediction results of the methods of GRA and LS-SVM in parallel or in series are better than single GRA method and LS-SVM method, and the precision of the method of GRA and LS-SVM in series is the highest.