Study On Influencing Factors And Forecasting Methods Of Medium And Long-term Power Generation Of Distributed PV Power Stations

At present,the construction of distributed photovoltaic power stations in China lacks systematic and comprehensive preliminary planning,and the prediction of project power generation involves the interests of investors.At the same time,accurate estimation of the power generation of photovoltaic power stations in the medium and long term is also of great significance for the improvement of grid planning,scheduling optimization and management development.However,due to the large time scale,low accuracy of weather forecast,limited historical power generation data,and significant difference between medium and long-term power generation prediction and short-term power prediction,short-term power prediction technology cannot be directly applied to medium and long-term power prediction.Therefore,it is of great significance to establish an effective forecasting method for medium-and long-term distributed photovoltaic power generation.This paper focuses on the analysis and prediction method of influencing factors of medium and long-term power generation of distributed photovoltaic power stations,and the specific research results are as follows:Aiming at the problems of improper selection of main impact factors and low data quality,a correlation analysis method model based on copula function was established,and the nonlinear effect and trend correlation measure of copula function were used to effectively extract the key impact factors of photovoltaic power generation.In addition,the data preprocessing technology is used to deal with the missing values and outliers in the data samples,so as to obtain complete and high quality input data.In order to select distributed photovoltaic power station system efficiency(PR)impact factors,analysis of the meteorological factors and the connection degree of PR,analysis the loss model of PV modules and photovoltaic modules is given 5 parameter calculation model,established a distributed photovoltaic power station,power transmission equipment such as inverters,ac/dc cable,and junction box and other physical loss model.This model provides input parameters for PR prediction model of distributed optical power station.Based on the traditional random forest(RF)model,a PR prediction model for distributed photovoltaic power stations based on fuzzy c-means clustering and random forest(FCM-RF)was established.By comparing the prediction performance of the two models with the actual distributed photovoltaic power station data and the local meteorological data,it is concluded that the improved stochastic forest model has better prediction performance.Based on LSTM neural network,a medium-and long-term irradiance prediction model is established.The prediction effects of LSTM neural network model,multiple regression model,persistent prediction model and support vector machine model are compared with the actual meteorological data.The results show that LSTM neural network has wide applicability in the case of limited data samples.On the basis of PR prediction model and medium and long term irradiation prediction model,an indirect prediction model is established.Based on the actual distributed photovoltaic power station data,the error between the actual power generation and the predicted power generation is analyzed and compared,and the method is proved to be feasible.