A Study On The Model Of Forcasting For Photovoltaic Power Generation Considering Meterologic Elements

In recent years, the photovoltaic power generation technology has grown rapidly, the generating costs started to fall, photovoltaic grid connected capacity has grown gradually. However, the output power is uncertain and unstable because the output power of photovoltaic system is affected by various meteorological factors, photovoltaic grid connected to the power brings certain influence to electricity quality and power grid safety. It requires to predict the photovoltaic system and daily cumulative power accurately to schedule the grid optimally and maintain the safety of electric power. The choice of photovoltaic predictive model as the research content, combined with meteorological data as input has certain academic value and practical significance.This paper first explores the impact of several meteorological elements of PV output power, analyzes the relationship between the intensity of solar radiation, air temperature, relative humidity, wind, weather type and photovoltaic power generation through the historical generation data and meteorological data. Then this paper uses the method of correlation coefficient to do correlation analysis of meteorological elements and the PV output power, then gets the weights for influence of meteorological elements on the PV output power. This paper selects the intensity of solar radiation, atmospheric temperature as the main meteorological factors, relative humidity as the minor meteorological elements, which has provided the theory and data base for the establishment of prediction model.This paper has expounded the basic theory of multiple linear regression, support vector regression, grey theory and BP neural network, detailedly describes and analyzes the parameters and establishment process of prediction model. On the basis of correlation analysis, this paper utilizes the factors of historical output data and meteorological data as input. The prediction models based on four algorithms were established in Matlab 2012 b respectively.Then this paper does the error analysis of the four prediction models with photovoltaic power station historical data. The predicted results show that besides the grey theory model, the rest three models can predict the PV output power accurately, and the support vector machine model has the highest prediction accuracy and the greatest stability. At last, this paper uses support vector machine model to predict the PV output power in sunny days, cloudy days, overcast days and rainy days. The results indicate that the model has better forcast ability and has achieved higher forecast accuracy.