Research On Nonlinear Prediction Method Of Wind Field Based On Laser Weathering Radar

The change of low-altitude atmosphere is complicated,and it has great influence on human life.The laser method has the characteristics of fast measurement and high precision in low-altitude atmospheric measurement.At low altitude,the wind shear rate is short,and the response time to make the right decision is particularly short and difficult.It is necessary to predict the future trend of the wind field.Doppler laser radar is used to detect the wind velocity in the radial wind field,and then use the DBS(Doppler Beam Swinging)inversion algorithm to obtain the three-dimensional wind field.For how to obtain the future wind field change,this paper needs to be solved in the scientific research problem.In this paper,the wind field of a wind field in northeast China is taken as the research object,and the wind field prediction of time series is carried out.Firstly,the Doppler laser radar is introduced,and the real-time wind speed is calculated according to the frequency shift of the transmitted pulse and the received echo signal.The Doppler beam swing model is used to invert the wind field.Wind shear is based on radar data using differential filtering or least squares recognition.The Brownian third-order exponential smoothing prediction algorithm,the Bock-Jenkins prediction algorithm,the BP neural network prediction algorithm and the gray prediction algorithm are studied.The working principle of each prediction algorithm is analyzed.The error curve and the precision curve are used to describe each the actual effect of the prediction algorithm.The properties of the Brownian third-order exponential smoothing algorithm are polynomials,which depend on the primary and secondary smoothing coefficients.The Bock Jenkins algorithm needs to preprocess the predicted time series and select the appropriate prediction according to the correlation coefficient of the time series and the autocorrelation coefficient The BP neural network is used to adjust the network coefficients automatically by the preset time series according to the residuals of the output by the preset neural network structure.The BP neural network is used to update the coefficients and output the predicted values.The residuals of the gray algorithm are Order differential equation,the predictive value has an exponential characteristic in the time series,and often a few predictive data can be completed.In this paper,the software part is in the visual studio2010 programming platform using C ++ language for software development,the Doppler laser wind radar radar data obtained by the wind field prediction related research work.According to the comparison between the wind speed curve and the actual wind speed curve and the comparison of the error and the precision,a new method based on the gray modified BP neural network residual is proposed.The wind speed curve is compared with the actual wind speed curve.Prediction algorithm.Finally,the spatial wind field data measured by Doppler laser wind radar with time-space resolution of 2min and 50 m are used to predict the gray field.The prediction algorithm of gray corrected BP neural network residual prediction is used to predict the wind field.Which indicates that the predicted wind field effect is different from the actual wind field,and the error and accuracy are better than the traditional prediction algorithm.Experiments show that the gray prediction BP neural network residual prediction algorithm is feasible in wind field wind speed prediction.