Research On Photovoltaic Power Generation Prediction Based On Dimensionality Reduction Of Multiple Meteorological Element

At present,solar energy has been widely used as a green renewable energy.However,photovoltaic power generation is affected by environmental factors,and has greater instability and intermittence,which challenges the large-scale integration of photovoltaic power generation system into power grid and the development of energy storage system.Therefore,it is important to study how to improve the power prediction accuracy of photovoltaic generation to promote photovoltaic grid-connected and power system scheduling.This paper takes the photovoltaic power prediction as the research target,puts forward the algorithm of photovoltaic power prediction based on OVMD-t SSA-LSSVM and the method of dimension reduction based on PCC-GRA-PCA,and designs and develops a set of photovoltaic power prediction system based on OVMD-t SSA-LSSVM.The main work of this paper is as follows:Firstly,the cyclic neural network,limit learning machine and least squares support vector machine which are commonly used in the field of photovoltaic power prediction are studied,and the photovoltaic power model is established to predict and analyze.By comparing the prediction results under different weather types of each model,it is found that the least squares support vector machine has a higher prediction performance for photovoltaic power generation,a higher prediction accuracy and stability for different weather types,and a better dynamic performance.Secondly,in order to predict photovoltaic power more accurately,this paper presents a photovoltaic power prediction model based on optimal variational mode decomposition(OVMD),adaptive t-distribution sparrow search algorithm(t SSA),and least squares support vector machine(LSSVM).First,the input photovoltaic time series data is decomposed using OVMD to improve the accuracy of data decomposition and avoid information loss.To avoid the local optimum in the later stage of the algorithm,t SSA algorithm is introduced to optimize the parameters of LSSVM model.A photovoltaic power prediction model based on OVMD-t SSA-LSSVM algorithm is built,and the model performance is validated using the historical photovoltaic data and meteorological data provided by the in-school photovoltaic power generation system.The OVMD-t SSA-LSSVM model proposed in this paper has better prediction accuracy and fitting effect than the comparison model by using four indicators: the determination factor(R2),the mean absolute error(MAE),the root mean square error(RMSE),and the mean absolute percentage error(MAPE).The experimental data show that the MAPE of the model is less than 0.0789,RMSE is less than 0.4035 under different weather types,and the R2 of the model is more than 96%.Then,in order to remove irrelevant and redundant information from the dataset and reduce the complexity and computational load of the model,a dimension reduction method based on PCC-GRA-PCA is presented.This dimension reduction method uses Pearson correlation coefficient(PCC)and grey correlation analysis(GRA)to analyze the feature importance of a variety of meteorological features,and achieves the initial dimension reduction from the point of feature selection.Then,the primary dimension reduction data is processed by principal component analysis(PCA),and the second dimension reduction of meteorological data is achieved from the perspective of feature transformation.Finally,combined with the model created in Chapter IV and the meteorological data after dimension reduction,it is found that the prediction accuracy of the model has been further improved.The RMSE of the model has decreased by 0.1939,0.1207 and 0.1432 in sunny,cloudy and rainy weather,respectively.The validity of the dimension reduction method based on PCC-GRA-PCA is verified.Finally,based on the dimension reduction method proposed in this study and the optimized photovoltaic power prediction model,a photovoltaic power prediction system is designed and developed,which integrates data preprocessing,data dimension reduction,photovoltaic power prediction and photovoltaic component conversion efficiency monitoring.The prediction system is tested by using a campus photovoltaic site.In the 31-day experimental data,under different weather types,the total prediction error is 5.15%,and the average system prediction time is 35 ms,which verifies the stability and practicability of the photovoltaic power prediction system designed in this chapter.At the end of the paper,the research content is summarized,and the issues that need to be studied in the next step are prospected.