Research On Generating Power Forecasting Of Photovoltaic System

As China’s vigorous development of renewable energy, the installed capacity ofgrid-connected PV systems is increasing. Grid-connected Photovoltaic system isaffected by meteorological factors, its output is intermittent and volatility. A largenumber of grid-connected PV systems will increase the complexity of the powersystem, changing the existing grid system margin and power generation plan, causingthe system to face the risk of collapse. Improving the prediction accuracy can helppower sector develop a detailed scheduling program and improve the power systemstability.In this paper, a lot of reading at home and abroad on the basis of the literature hadbeen conducted and three different aspects of photovoltaic power generationprojections were studied. By analyzing the impact of grid-connected PV power relatedfactors, we select the main factors as neural network input variables to buildphotovoltaic power generation based on neural network prediction model. Throughdata mining techniques, we screened from large amounts of data and forecast periodswith similar meteorological characteristics data sequence, using gray-correlationtheory to predict the photovoltaic power generation. By using theoretical weightcalculation techniques of the combined forecast, neural network prediction and greyprediction are given different weights for the combination of photovoltaic powerforecasting technology.In this paper, all the data is based on the grid-connected PV monitoring system ofTianjin University. Neural network-based prediction of photovoltaic power generationtechnology, gray relational analysis of photovoltaic PV power forecasting techniquesand combined forecast techniques are analysed under the same weather types forsimulation and data validation. And these three predictions prediction techniques wereassessed through the use of five different error indexes. Assessment indicates threeprediction methods have reached a higher prediction accuracy.