Research On Short-term Wind Power Forecasting Method Based On WDRI-RF And Sensitivity Error Feedback

Vigorously developing green energy power generation is one of the feasible ways to achieve the goal of "carbon neutrality",and wind power is an important part of the construction of renewable energy power generation.At present,the total installed capacity of wind power in my country continues to increase,and the proportion of wind power in the power grid is gradually increasing.Wind power forecasting technology of different time scales solve this problem.At the same time,the application of wind power prediction technology in distribution network dispatch control has a positive effect on the demand side response of new energy participation.In view of the deficiencies of the existing forecasting methods,corresponding researches are carried out in this paper,the specific work is as follows:First,for the input side of the model,In-depth mining and processing of hidden features in wind direction data is studied in this paper.Short-term wind power forecast method based on wavelet packet and decision tree is proposed.Extract the wind direction variation characteristics and improve its relevance with wind power.The problem that the effectiveness of wind direction factor cannot be reflected in the prediction model due to its low relevance with wind power is solved.Further,Wind Direction Relevance Improvement Algorithm is proposed in this paper.Wind power data is divided into outlier and normal data sets and power trends are used to further refine data set characteristics.For the model operation side,Wind Direction Relevance Improvement-Random Forest is proposed.The decision tree is trained for multiple data sets,and finally WDRI-RF is constructed.Finally,based on the WDRI-RF,on the model output side,Nonstandard Third-order Polynomial Normal Transformation Method is introduced and based on sensitivity analysis,the relationship between the input of clustered data set and model parameters and prediction error is constructed.Optimize the model at the data input and model parameters.For the forecasting method proposed above,wind farm data is used as an example to verify it.The results show that the correlation of wind direction is improved by continuously optimizing the prediction method proposed in this paper.The model is optimized by using sensitivity error feedback,the forecast error is reduced gradually,and the accuracy of the forecast model is improved.