Studies On Intrusion Signal Diagnosis Of Oil Pipeline Based On Ensemble Empirical Mode Decomposition Method

Oil pipelines are one of the main transportation routes for oil.Most of the main pipelines are located in remote areas and are vulnerable to illegal invasions such as human activities,animal and plant movements,severe weather,and geological disasters.Routine maintenance and safety guarantee of oil pipelines are difficult.Under such circumstances,it is particularly important to use oil pipeline perimeter guard systems to protect oil pipelines from the harmful effects of illegal invasions.In the pipeline perimeter guard systems,the intrusion signal diagnosis algorithm is in the core position,and it is related to the ability to accurately diagnose and report the intrusion signals.This is also one of the research hot spots in recent years.The thesis takes the intrusion signal diagnosis algorithm as the research object.Under the background of the traditional diagnosis algorithm with low diagnostic accuracy and low alarm speed,the time-frequency characteristics of the intrusion signals are analyzed in depth,and the thesis studies the intrusion signal diagnosis algorithms based on two kinds of the oil pipeline perimeter guard systems.The achievements of the thesis are summarized as follows:(1)The thesis studies the theoretical basis of spectral centroid(SC)and introduces it into the intrusion signal diagnosis algorithm as the signal’s characteristic parameter.A threshold diagnosis method is proposed for the oil pipeline perimeter guard system with ground optical fiber networks.The method firstly decomposes the detection signals collected by the sensing optical fiber through the empirical mode decomposition(EMD)algorithm,then removes part of the intrinsic mode function(IMF)components to achieve denoising,and then combines the remaining IMF components to form the reconstructed signal.Then,the characteristic parameter SC is extracted,and finally the method determines the SC threshold so as to diagnose the intrusion signals,and further confirm the type of the intrusion signals and issue an alarm.(2)The thesis validates the threshold diagnosis method by analyzing the signals collectedby the oil pipeline perimeter guard system with ground optical fiber networks.The results show that the diagnostic accuracy of the method is higher than the traditional singular value decomposition(SVD)method,therefore reflecting the feasibility and accuracy of the method.(3)In the thesis,the least squares support vector machine(LS-SVM)is introduced into the intrusion signal diagnosis algorithm based on the threshold diagnosis method,and an LS-SVM classification method is proposed for the oil pipeline perimeter guard system with the buried optical fiber.The method firstly decomposes the detection signals collected by the sensor optical fiber through the ensemble empirical mode decomposition(EEMD)algorithm to avoid the possible modal aliasing phenomenon of the EMD algorithm,and then extracts the characteristic parameter SC.Finally,the method introduces the SC into the LS-SVM for learning and classification to diagnose intrusion signals,and further identify illegally invaded area and issue an alarm.(4)The thesis validates the LS-SVM classification method by analyzing the field data collected by the oil pipeline perimeter guard system with the buried optical fiber,and further compares the advantages with the traditional intrinsic mode singular value decomposition method.The results show that the diagnostic accuracy of the LS-SVM classification method is5.30% higher than the traditional method,reflecting the feasibility,accuracy and timeliness of the method.