Research On Forecasting Of Grid-connected Photovoltaic System Based On Three-way Decisions Theory

Photovoltaic(PV)power generation has many advantages,such as clean,non-pollution and renewable etc.But confined by many factors such as meteorology,equipment and design etc,the PV power generation has the characteristics of instability and periodicity.So it is important to make an accurate prediction.It can reduce the negative effects on power grid dispatching and security,which are caused by the uncertainty of PV generation.The previous operation data was collected by the 5.6kW grid-connected PV power generation in Beijing University of Civil Engineering and Architecture.Depending on these data,the power forecasting model for PV power generation system based on three-way decisions theory has been established.And in this model,solar irradiance,ambient temperature and plate temperature are the main influencing factors.Then the historical meteorological data and the PV power generation data are conducted a series of operations with covering algorithm,which consist of classification processing,data normalization,vector projection in the same length and the random selection of samples centers.After that,a group of covering sets will be formed on the basis of the historical data of PV generation power.Besides,the three regions including positive region,negative region and boundary region will be established on the basis of different types of covering sets.What's more,the approximate covering samples will be obtained from the positive region by the Euclidean distance method,after the prediction moment is classified into a certain coverage through three-way decisions rules.Eventually,the PV power generation of the prediction moment will be obtained according to the regression analysis method.The forecasting model of PV system and error evaluation simulation program have been completed with Visual C++ 6.0 language.The predicted results are compared with the measured results.Finally,the results show that the maximum RMSE between the actual and the predicted value is below 6.5%.It is less than 15% stipulated by the standard.What's more,the percent of pass is 100%.Meanwhile,it is analyzed that the influence of various weather types on the prediction of photovoltaic power generation.Particularly,the prediction accuracy of hazy days and snowy days are studied.The results show that the prediction accuracy of this model is still less than 2%,even if the PV output power is very low and unstable in extreme weather.In this thesis,three-way decisions theory is used in the field of forecasting of grid-connected PV for the first time.And covering algorithm is used to calculate the forecasting of grid-connected PV generation.At last,it is the high simulation precision that proves that the prediction model is correct and accurate.