Research On Defect Feature Information Extraction And Mechanical Fault Diagnosis Of High-voltage Circuit Breaker Operating Mechanism Based On Multi-source Data Fusion

Since the 21st century,with the new energy revolution becoming increasingly fierce,our power grid is transforming into intelligent and comprehensive direction.The access of new energy has expanded the distribution range of the power grid,making the operation of the power grid increasingly complex.The input of switching appliances such as circuit breakers,isolation switches and load switches has increased.As the switch with the most complete functions and the widest application range,the operating environment of circuit breakers has become increasingly stringent.Circuit breaker in the power system mainly plays the role of protecting motor and other power equipment,when the system occurs short circuit and other faults,the circuit breaker can immediately cut off the fault line to protect the rest of the line from damage;When the circuit breaker fails to cut off the line normally,the fault scope will be expanded,and the entire power system will be paralyzed in serious cases.The fault of circuit breaker often occurs in spring operating mechanism,and the fault of spring operating mechanism is mainly mechanical fault.At present,many scholars at home and abroad have carried out a lot of research on the diagnosis technology of mechanical fault of operating mechanism,but the type of signal used for fault diagnosis is too simple,and the fault diagnosis method is not targeted.Based on the working principle of CT26 type spring operating mechanism of LW30-72.5kV high voltage SF6 circuit breaker,this thesis analyzes the characteristics of current signal of closing coil and surface vibration signal of operating mechanism,and builds the "electricvibration-sound" signal acquisition platform of operating mechanism.In order to improve the accuracy of fault diagnosis,the feature extraction method of each signal is studied and the characteristic parameters of each signal are fused.The mechanical fault diagnosis algorithm of operating mechanism is proposed,and the mechanical fault diagnosis of operating mechanism is realized.The research content of this thesis is as follows:(1)Research on Characteristic Signal Acquisition Technology of Circuit breaker Spring operating Mechanism.Firstly,the working principle of the spring operating mechanism is analyzed,the difference of the vibration signal of the operating mechanism in different working conditions is expounded,and the characteristics of the vibration signal are expounded.The characteristics of the current signal of the switching coil are analyzed and the internal relation between the current signal and each stage of the operating mechanism is studied.In order to realize effective collection of "electric-vibration-acoustic" mixed signals,a real-time signal acquisition platform was built in this thesis.Current oscilloscope was selected to monitor coil current signals,acceleration sensors to monitor vibration signals,and acoustic sensors with preamplifiers to monitor sound signals.Then the above mixed signals were stored in the database to establish the characteristic signal database of operating mechanism.Three typical faults,such as screw loosening,abnormal output of opening spring and loosening of closing electromagnet,were simulated manually.The characteristic signal database was expanded and the comprehensive database of mechanical state of operating mechanism was established.(2)Research on Signal preprocessing and feature Extraction of Circuit breaker Spring operating Mechanism.After obtaining the original data of the "electric-vibration-acoustic"mixed signal of the operating mechanism,in order to eliminate the influence of environmental noise and electromagnetic interference of the surrounding equipment on the original signal,the mathematical morphological filtering method is used to preprocess the original signal,eliminate the burrs and furrows in the original signal curve,remove the high-frequency noise component,and provide guarantee for the subsequent feature extraction.After signal pretreatment,the characteristics of current signal in time domain and vibration signal in time and frequency domain are analyzed,and the differences of current signal in time domain waveform and vibration signal in time and frequency domain waveform in different mechanical faults are compared,which provides a theoretical basis for fault feature extraction for diagnosis.In this thesis,different feature extraction methods are adopted according to the characteristics of each signal.For coil current signal,the time and amplitude corresponding to its characteristic data point are selected as the local feature quantity,and the mean value,standard deviation,root mean square value and kurtosis of each segment after the subsection of the characteristic data point are taken as the subsection feature quantity to construct the characteristic database of coil current signal.For vibration signal,FDJI-VMD(Frequency Domain Judgment IndexVariational Mode)algorithm is proposed to decompose the vibration signal and then extract the characteristic parameters of each mode component.In order to solve the problem of selecting mode decomposition layers in VMD algorithm,a frequency domain judgment index function including envelope entropy,correlation coefficient and center frequency interval was constructed.By observing the variation trend of indicator function,the decomposition layers of VMD algorithm were selected as four layers,and then the energy entropy and sample entropy of modes at each layer were calculated as the characteristic parameters of vibration signals.Construction of vibration signal characteristic database;For sound signal,its generation mechanism is similar to vibration signal,but due to the different propagation path,the energy of vibration signal in the frequency band.(3)Research on Fault diagnosis Technology of Circuit breaker Spring operating Mechanism.After the mixed characteristic parameters of "electric-vibration-acoustic" signals are obtained,principal component analysis is used to reduce the dimension of the mixed characteristic parameters to obtain the main components that can reflect the mechanical state.The key characteristic parameters that can characterize the mechanical state are selected according to the principle of principal component analysis.In this thesis,the Support Vector Machine(SVM)is used as the basic classifier,and the whale optimization algorithm with adaptive function is adopted to optimize the parameters g and C of SVM.Adaptive Whale Optimization Algorithm(AWOA)can intelligently jump out of iteration,save calculation time,and prevent falling into local optimal.The operating conditions of circuit breakers are complex and the fault types are variable,so the accuracy of basic classifier is insufficient.Ten SVM basic classifiers are fused and strengthened by Adaboost fusion weighting algorithm,and the diagnostic accuracy of the strengthened classifier is improved by nearly 10%.The AdaboostAWOASVM fault diagnosis method based on "electric-vibration-acoustic" multi-source signal proposed in this thesis has a correct rate of 95%,which has certain practical significance.In this thesis,the whole process of "electric-vibration-acoustic" signal acquisition,preprocessing,feature extraction,principal component analysis and fault diagnosis of the circuit breaker spring operating mechanism is realized.The experimental verification shows that the mechanical fault diagnosis method of the high voltage circuit breaker spring operating mechanism based on "electric-vibration-acoustic" multi-source signal feature fusion analysis is reliable and effective,and has certain reference value.