| The worldwide electricity business has expanded quickly in recent years,wind energy as a renewable energy,the growth and widespread use of wind energy not only support China’s energy sector’s sustainable development,but also an important means to eliminate energy depletion in all countries in the world.Large-scale wind energy integration into the grid will have a detrimental effect on the power system since wind energy is unpredictable,intermittent,and volatile.Accurate wind forecasts can help power operators plan ahead,improve the economics of grid operations,and mitigate the negative impact of wind energy uncertainty.It is important to study wind speed fluctuations in numerical weather prediction in detail.The accuracy of wind power forecasting in the future will be much improved by a thorough investigation of the variation of wind speed in numerical weather prediction.Firstly,a deep adaptive filtering framework is proposed,for the variational mode decomposition algorithm,the adaptive selection of its core parameters is realized by introducing Kullback–Leibler Divergence,and after multiple modal components are generated by decomposition,Non-Local Means denoising algorithm is introduced to filter out the noise components,and then it is reconstructed with the remaining effective components to obtain the noise reduction sequence.On this basis,the noise reduction sequence is used as input by season,and the most suitable prediction model for this season is selected from the alternative model base through the verification set for wind power prediction.The power prediction of large-scale wind power clusters is conducive to improving the power grid’s capacity to absorb wind power and ensuring the safe,stable and efficient generation of wind power.In order to consider the spatial correlation factors of wind turbines in the wind power cluster and improve the accuracy of short-term power prediction of the wind power cluster,based on single field prediction,the grey correlation degree sequence of selected reference wind farms and other target wind farms in the wind power cluster is obtained through grey correlation analysis for wind farms in multiple locations in the wind power cluster.Then,based on this sequence,a goaldirected deep learning approach,Deep Attentional Embedded Graph Clustering is introduced to cluster the wind power clusters,optimize the prediction process of wind power clusters,and then the deep learning based prediction framework is adopted to complete the power prediction of the whole cluster.For large-scale wind power clusters containing multi-location wind farms,analyzing the factors that affect the prediction errors and improving the links that cause errors are beneficial to improve the accuracy of wind power prediction.This paper considers the spatial correlation between wind farms and the similarity of wind power clusters in time series,the causes of wind power cluster prediction errors are analyzed from the perspectives of wind farm time series,spatial location,cluster prediction model,wind power size and numerical weather prediction wind speed,and the error analysis is carried out with the data of a large wind power cluster in northeast China,which provides the direction for realizing higher precision wind power prediction. |
