Optimization Of Spatial Correlation Wind Speed Prediction Regression Method

Regression prediction can find the implied relationship between things,and predict the situation or development of things.So it have a wide range of applications in the social sciences and natural sciences and other fields.Wind power prediction is an important application of regression prediction.Due to the tight supply of energy and the increasing of environmental pollution,more and more attention has been paid to the research of clean energy.Wind energy is a kind of high-quality clean energy,however,because of its instability,the application of wind power is subject to certain restrictions.Wind power prediction is an important way to solve the problem that wind power is difficult to parallel in the grid.In order to improve the reliability of wind speed prediction used in ultra-short-term wind power prediction,the main work of this paper is as follows:(1)A spatial correlation method based on current and recent historical observations,including wind speed and wind direction,of anemometer towers in the surrounding areas is adopted.Firstly,the lag similarities between the wind speed of the predicted site and the wind speed of anemometer tower are calculated,if the wind direction of anemometer tower is closer to the predicted site enough.If its similarities are higher than the threshold value,this anemometer tower is selected as a wind speed upstream site of the predicted site.Secondly,taken account of the optimal lag time,a prediction model is trained.Lastly,current historical observations of each upstream wind speed are inputted to the model,and the predicted wind speeds of the predicted site are obtained.(2)Taking Huibertgat in Netherlands and Tianjin as the predicted sites,and the partial least squares regression as the prediction model,numerical experiments provide the relations among the future wind speed prediction errors,the order of the model and the historical sample size of the training sets.Linear regression,least squares support vector regression,generalized regression neural network and random forest regression are used as the contrast prediction models.(3)The method of weighted least squares is used to optimize the regression prediction method in order to improve the effect of spatial correlation wind speed prediction,and the influence of the weight parameter value on the spatial correlation wind speed prediction error is studied.The main conclusions are as follows:(1)The simulating results show that the prediction errors are not sensitive to the order of the model and the sample size during the winter monsoon,if the sample size is large enough.These results show that the spatial correlation is a reliable approach to predict future ultra-short-term wind speeds.(2)Partial least squares regression and least squares support vector regression are superior to other regression methods in accuracy,but least squares support vector regression requires more parameters and runtimes is longer than partial least squares regression.Considering the prediction error and the computation time,partial least squares regression is an ideal method to predict the wind speed with spatial correlation.(3)Using the weighted least squares method can improve the effect of spatial correlation wind speed prediction.