Research On Photovoltaic Power Forecasting Based On SVM And ABC

With the acceleration of China's modernization process,The exploitation and use of fossil energy leads to the shortage of energy resources.It emits a lot of pollutants,and brings great pressure to environmental protection.Therefore,the development of green energy,such as wind energy,water energy,solar energy and other sustainable development and utilization of energy,has been paid more and more attention by the country.Among these energies,solar energy,with its clean,pollution-free,sustainable development and low development cost,has become a major alternative energy,which converts solar energy into electricity through photovoltaic power generation system.However,as photovoltaic power generation is greatly affected by weather conditions,its power generation has a certain range fluctuation,which will have a negative impact on the stability andsecurityoperation of the power grid.Therefore,the accurate prediction of photovoltaic power generation is great for making a reasonable scheduling plan and stability operation of the power grid.With the development of big data,the in-depth research on intelligent algorithm,machine learning,neural network and other methods has laid a foundation for the research on photovoltaic power generation prediction.Using these methods can effectively improve the accuracy of photovoltaic power generation prediction.Based on the research of domestic and foreign scholars on the prediction of photovoltaic power generation,this paper proposes an improved intelligent algorithm combined with machine learning to predict photovoltaic power generation.By comparing error of different prediction methods through experimental analysis,it is proved that the proposed method can effectively improve the prediction accuracy of photovoltaic power generation.The main research contents of this paper are as follows:(1)Based on the obtained data of daily radiation on the slope,temperature,humidity and wind speed,the factors affecting photovoltaic power were analyzed,and finally,the daily radiation on the slope,temperature and humidity were selected as the meteorological factors affecting photovoltaic power.(2)Choose the forecast date for three different weather conditions: sunny,cloudy and rainy.According to the selected meteorological factors,GRA and Euclidean distance are used to select similar day.In order to test the accuracy of two similar day selection methods,the mean value of photovoltaic power of two groups of different similar days is taken as the power prediction value of prediction day.The result shows that theprediction error of power prediction based on GRA is smaller than that based on European distance,so that the similar day selection based on GRA is more accurate.(3)An improved artificial swarm algorithm was proposed to optimize the kernel function and penalty parameters of SVM,and a prediction model based on SVMoptimized by improved ABC was established.Then,predicting photovoltaic power by using the data of similar days and prediction model based on SVMoptimized by improved ABC.Finally,through the experimental analysis and comparison of the prediction errors of different methods,it is proved that the prediction method based on GRA and SVMoptimized by improved ABC is effective and accurate.