Application Of Similarity Index And Its Confidence Intervals In Prediction Of Monsoon Wind Speed

Energy Internet,smart grid is the main form of current and future power utilization.Establishing a new energy system with power system as the core and new energy as the main primary energy,integrating other forms of energy such as gas,cold and heat has become the development trend of the energy industry.Wind energy is a renewable,clean,carbon-free energy source.Wind power has become an important part of the smart grid.However,the uncertainty and unpredictability of wind power is one of the main reasons for the difficulty of wind power consumption.From the "mechanism + identification" prediction strategy for spatial correlation wind speed prediction,the use of China's unique winter wind and summer wind can improve the accuracy and reliability of wind power prediction.In order to analyze the impact of similarity indexes on spatial correlation wind speed prediction.This article did the following work:(1)Sort the classification of some of the more commonly used similarity indexes.Based on this,it is determined which indexes can be used as similarity indexes for spatial correlation wind speed prediction.(2)The Pearson correlation coefficient's,the cosine similarity's and the Tanimoto similarity's confidence intervals are deduced respectively.The numerical simulation of the wind speed is simulated by Gaussian white noise,and the influence of different parameters on the confidence intervals is analyzed.(3)The k-nearest neighbor prediction algorithm is extended to the field of spatial correlation wind speed prediction.At the same time,linear regression and generalized neural network are used to compare wind speed prediction,and various regression models under different similarity indexes are used for error analysis.The main conclusions of this paper are:(1)Similarity indexes such as Pearson correlation coefficient,the cosine similarity and the Tanimoto similarity can reflect the future trend of wind speed time series,and it is more suitable for spatial correlation wind speed prediction than equidistance index such as Euclidean distance.(2)The confidence intervals of Pearson correlation coefficient,the cosine similarity's and the Tanimoto similarity's decrease with the increase of sample size;the confidence intervals of Pearson correlation coefficient and the cosine similarity's is decreased with the increase of the absolute value of vector correlation coefficient;The confidence intervals of the cosine similarity's and the Tanimoto similarity's decrease with the increase of the absolute value of the sample mean.The confidence intervals of the Pearson correlation coefficient are independent of the sample mean.The Tanimoto similarity has the largest confidence intervals when the correlation coefficient is about 0.8.In general,the Tanimoto similarity's effectiveness is higher than the cosine similarity's,and the cosine similarity is more effective than the Pearson correlation coefficient.(3)In regression models such as linear regression and generalized neural networks;the priority of the similarity index of k-nearest neighbor spatial correlation wind speed prediction is: the Tanimoto similarity,the cosine similarity,the Pearson correlation coefficient,and the prediction result is accurate and reliable.