Photovoltaic Generation Prediction Model Based On EEMD-Variable Weight Combination Forecasting And Application Research

Fossil energy has always played a very important role in the process of human society,but its own non-renewability and pollution to the environment have gradually failed to meet human needs.In contrast,solar energy has the characteristics of convenience,inexhaustible,clean and pollution-free,and has become one of the energy sources widely used by humans today.Photovoltaic power generation is rapidly expanding worldwide and has already accounted for a certain proportion of power generation methods in developed countries.In recent years,China has also vigorously developed photovoltaic power generation projects.Policies on photovoltaic development and pollution prevention have been proposed,and the importance of photovoltaic power generation has been emphasized from various aspects.However,since the photovoltaic power generation is greatly affected by various factors such as weather,the power output is not stable enough.In order to ensure the safe and stable operation of the photovoltaic power grid and the power dispatching is easier to carry out,it is imperative to accurately predict the photovoltaic power generation potential.In this paper,a method for predicting photovoltaic power generation based on ensemble empirical mode decomposition(EEMD)-variable weight combination forecasting is proposed.When combining EEMD with variable-weight combination forecasting model during forecasting period,the model prediction accuracy is improved.Firstly,EEMD is used to decompose the photovoltaic power data sequence,and the decomposed sequence components are combined into low frequency,intermediate frequency and high frequency according to certain rules.At the same time,three kinds of machine learning algorithms,including decision tree,support vector machine and integrated learning,are used as the basic prediction methods to form a variable-weight combination forecasting model,which can take advantage of each algorithm and make up for its shortcomings.The variable-weight combination forecasting model is used to predict the three parts of the low-frequency,intermediate-frequency and high-frequency sequences.The three-part prediction results are added to obtain the final prediction result.In the process of forecasting,the design of weights of combination forecasting methods is studied.Two methods of harmonic mean method and quadratic programming method are proposed,and the paper will explore which can achieve better results.The case study and model comparison results show that in the photovoltaic power generation prediction,the prediction results obtained by the EEMD decomposition are better than the direct prediction,and when the single prediction model becomes the variable-weight combination forecasting model,the prediction accuracy is further improved after the optimal weight is determined.After successfully validating the method of this paper,the model is optimized from the aspects of seasonal factors,input variables and similar day clustering,and the characteristics of photovoltaic power generation are studied,the accuracy of photovoltaic prediction is further improved.Finally,the practical application of photovoltaic forecasting in grid-connected photovoltaic power generation is studied and relevant recommendations are given.