Short-Term Prediction Of Wind Speed Based On Quantile Regression Neural Network

Quantile regression neural network method combines the advantages of quantile regression and neural network to provide new ideas for nonlinear data modeling and analysis.This paper uses the quantile regression neural network model to predict the short-term wind speed,which effectively improves the economic benefit and planning of the wind power energy market,and broadens the application field and method system of the quantile regression neural network.The construction of a sparse laplacian smoothing quantile regression bidirectional minimum gated memory network model has great theoretical significance and application value.Firstly,the public data set of Marx and Planck Institute of Biogeochemistry in Jena,Germany was selected,and position encoding is applied to numerical weather features,input to the multi-head attention mechanism for extraction features,and summarized with historical wind speed data.Then,the improved sparse laplacian smoothing term is applied to the loss function of quantile regression to train the quantile regression bidirectional minimum gated memory network to obtain the predicted wind speed under different condition quantiles.Finally,the kernel density estimation is used to calculate the probability density curve.To verify the validity and reliability of the composite framework,this paper tests the accuracy of the proposed model and the QRMGM,QRBi LSTM,QRLSTM,QRBi GRU,QRGRU,and QRNN.The point prediction analysis shows that the proposed model has a positive effect on the wind speed prediction ability,with a small deviation.The interval prediction results show that the coverage rate of the wind speed data exceeds comparative models while ensuring the certain interval width,which provides relevant technical support and theoretical basis for the stable operation of the power system,and has certain advantages.According to the probability prediction index,the probability prediction accuracy of the proposed model is high,and it has good adaptability to the fluctuation of wind speed data.For the reliability indicators,the prediction results of the model are reliable.In summary,a combination of sparse laplacian smoothing,quantile regression and bidirectional minimum gated memory network yields a conditional distribution of wind speeds,as well as the probability density curves,which can balance prediction performance and accuracy.At the same time,it helps to improve the effective management of the power generation market,and is of great significance to the long-term stability and safety of the power system.