Research On Dancing Signal Recognition Method Of Optical Fiber Composite Overhead Ground Wire

Optical fiber composite overhead ground wire is an important device to transmit electric energy,as the carrier to transmit electric energy,its operation safety is the basis and premise of the reliable supply of electric energy.The extrinsic factor such as weather may cause OPGW to swing with unusual high amplitude and low frequency,lead the optical cable to trip,damage and even break,which can not only cause large-scale power-off,but also serous economic loss.Therefore,identify the OPGW galloping signal,so it is important to discover the potential safety hazard of OPGW in time,in order to make sure that the normal electric energy transmission.In the paper,we mainly studies the following content:As to the random noise problem made by the galloping monitoring data mixes with external environment,the paper proposes a depressing noise method of OPGW galloping signal based on CEEMD algorithm and improved KSVD algorithm.First of all,the collected galloping signal data are subjected to signal differential processing,so as to facilitate signal analysis.Then,it make use of CEEMD algorithm to decompose OPGW galloping signal to get an eigenmode function set,introduce the structural similarity parameter,and design a method of redesign of signal,50 as to obtain a K singular value decomposition learming dictionary.Finally,the sparsity adaptive matching pursuit algorithm is used to denoise the signal sparsity,and complete the preprocessing of galloping signal monitoring data.Compared with the traditional denoise method,the method proposed in the paper can effectively random noise in the OPGW galloping signal,and get clear galloping signal information.In the feature extraction stage of the OPGW galloping signal,the paper proposes a feature extraction method based on multi-angle signal.As to the current problem,when the galloping signal feature is extracted,often based on single parameter extraction method such as signal time domain or frequency domain features,the method only can represent part information problem of the signal,analyze signal tilue domain and frequency domain features in the normal signal and abnormal of OPGW galloping monitoring data,extract effective features of signals respectively from multiple angles of the signal time domain,the signal frequency domain characteristies,the signal of time domain and the linear prediction coding coefficients characteristics of the signal,to fully reflect the characteristics of OPGW galloping monitoring data,and select the extracted characteristics based on entropy principal component analysis method,and obtain the ranking of the characteristic importance by using the value instead of taking the variance contribution rate of principal components as the weight.In the contribution stage of OPGW galloping signal identifieation model,galloping signal identification model based on optimizing random forest algorithm is designed.In view of the features of forest algorithm,such as suitable for large quantity,unbalanced data and good classification,it introduces random forest algorithm to identify OPGW galloping signal,propose a method based on the combination of particle swarm optimization and improved grid search,to optimize the parameters of forest algorithm,the method can improve the problem of subjective parameter values in forest algorithm and lead to the reduction of classifier precision,and speed up and optimize the practicing of model.Compared with traditional optical fiber signal recognition model,OPGW galloping signal identification model based on optimizing random forest algorithm can better identify the galloping situation of optical fiber composite overhead ground wire,enhance identification accuracy.