Characteristics Analysis And Identification Of Wind Farm Clutter On The Condition Of The Motion Platform

Ground-based radar will be affected by wind farms during target surveillance and weather observation,and the performance of the radar system will be severely disrupted.For airborne radar,this impact of wind farms is more complicated and difficult to deal with.In order to explore the impact of wind farm clutter on the airborne radar of the moving platform,it is particularly important to study the wind turbine echo simulation technology and how to realize wind farm clutter identification for the airborne radar.Firstly,the array antenna echo model is combined with the wind turbine scattering point superposition model,and wind turbine echoes model at any observation point for the airborne radar is established.It provides data source for the subsequent identification of wind farm clutter for airborne array radar.Secondly,the characteristics of the wind turbine echo are analyzed in the range doppler domain,time-frequency domain and space-time domain to explore the influence of the different blade angle,pitch angle and azimuth angle on the characteristics of the wind turbine echo.At the same time,the simulation analysis results are compared with the theoretical analysis results to verify the validity of the echoes model.Finally,in view of the prior information related to wind farm clutter for the airborne radar is unknown,the unsupervised machine learning of autoencoder and K-means are combined to realize the identification of the wind farm clutter and target in the radar scanning scene.For the radar scanning scene without target,the time-frequency(or spectrum)data of each scene range resolution cell is used as input sample,and the dimensionality of the input data can be reduced based on autoencoder to obtain low-dimensional feature data.Identification of the wind farm clutter can be realized based on the K-means clustering method.For a radar scanning scene with target,a balanced sample set is firstly constructed,and the autoencoder and K-means are used to cluster the balanced sample set in reduced dimensions.Then,the autoencoder network and K-means class center point are extracted to form the identification model.The identification model can be used to realize the identification of wind farm clutter and target in the scanning scene.