Research On Short Term Photovoltaic Power Prediction Based On Similar Day Theory And EGA-VMD-GRU

In the context of global environmental pollution and energy crisis caused by modern industries relying on fossil fuels,it is urgent to find clean and pollution-free renewable energy to change the existing energy structure.Photovoltaic power generation has the advantages of low pollution,low noise,and simple energy conversion process.As a clean energy power generation method,its installed capacity proportion in the power industry is constantly increasing.However,photovoltaic power generation has the characteristics of both randomness and high volatility in power generation,which makes the integration of photovoltaic power generation into the power system interfere with the safe and stable operation of the power system.Predicting the power of photovoltaic power generation can provide reference for the output scheduling of the power system,allowing the power grid to develop corresponding scheduling plans in advance for the predicted power generation,and to suppress the impact of intermittent and fluctuating characteristics of photovoltaic power generation connected to the power grid on the power system,promote large-scale photovoltaic power generation integration and consumption.This article conducts research on short-term photovoltaic power prediction methods,and the main work content is as follows:Firstly,by comparing the photovoltaic power generation curve with the meteorological factor curve,the impact of various meteorological factors on photovoltaic power generation was analyzed,and the Pearson correlation coefficient between the photovoltaic power generation curve and the meteorological factor curve was calculated to evaluate the correlation between various meteorological factors and photovoltaic power generation.The meteorological factor data with higher correlation degree was selected as the input data for the subsequent photovoltaic power prediction model.Select the GRU neural network that performs well in processing time series data as the backbone of the prediction model.Secondly,by observing the photovoltaic power generation curve under different weather conditions,it was found that the photovoltaic power generation curve under non sunny weather conditions has the characteristics of large fluctuations and instability,which is not conducive to the prediction of photovoltaic power.To solve this problem,Variational Mode Decomposition(VMD)is used to decompose the photovoltaic power series to improve prediction accuracy.Due to the significant impact of the number of modal decompositions and the selection of penalty factors on the noise level and information content of the decomposed modes in VMD,the elite preserving genetic algorithm(EGA)combined with minimum envelope entropy was selected to optimize the parameters of VMD.In order to further improve prediction accuracy,the theory of similar days was introduced to search for historical days similar to the predicted day,and the data of historical days was used as training data for the prediction model.The Grey Relational Analysis(GRA)algorithm is used as a means to evaluate similar days,and a resolution coefficient value strategy and TOPSIS method are introduced on the basis of GRA to improve the accuracy of selecting similar days.Finally,establish a short-term photovoltaic power prediction model based on the theory of similar days and EGA-VMD-GRU.Test the predictive effect of the prediction model using examples.The test results show that the short-term photovoltaic power prediction model based on the similar day theory and EGA-VMD-GRU effectively improves the prediction accuracy.