Study On Partial Discharge Identification And Harmfulness Assessment For Gas Insulated Switchgear

High-voltage switch gear is the most critical equipment for direct power supply to the distribution network or general large consumer,and has been widely applied in power supply and distribution systems.The failure of high-voltage switch gear will directly affect the quality of power supply services,leading to the most direct economic losses.Switchgear in the operation of voltage,current,heat,force as well as in the production,transportation,installation and maintenance process,will inevitably produce or make a variety of insulation defects,which results in the decrease of electrical strength and the device failure.Partial discharge detection is an important technical method to prevent the sudden power failure caused by switchboard insulation fault.Accurate detection,positioning and identification of insulation defects,can improve the reliability of the switchgear operation,reduce the interruption maintenance and prevent the power failure.Under the development of detection technology,the multi-detection information fusion has become an inevitable trend of the development of fault detection technology.This paper breaks the information barrier between different detection methods and establishes the multi-sensor information fusion detection model of switchgear partial discharge based on DS evidence theory.The accuracy and validity of the proposed method are demonstrated in the field fault cases.Based on the experiment,a partial discharge multi-sensor information diagnosis system is established by applying partial discharge transient voltage detection method,ultrasonic detection method and UHF detection method.Through a large number of laboratory experiments,the metal protrusion in the switchgear,metal suspensions,internal defects of insulation,air gap discharge characteristics and signal characteristics,are obtained,and therefore a multi-information-source database is established.Multi-sensor detection can reflect the regularity of partial discharge in view of multi-feature quantity,and therefore can identify the defect attributes.In order to overcome the shortcomings of the detection sensitivity in the most of defects,the transverse and longitudinal comparison information is applied to provide evidence for the existence of suspended or airborne air gap discharge.The ultrasonic method is applied to extract the model information of amplitude detection,and identify metal protrusions,metal suspended solids and air gap defects.In the information fusion,the insensitivity to insulation inner defect is applied,to prevent the defect misjudgment.By taking advantage of high sensitivity of UHF detection,and clearanc of defect discharge characteristics in PRPD mode,the contour characteristics of the discharge distribution are extracted,and then the various types of insulation defects are identified effectively.Based on the BPNN model,and the characteristics of ultrasonic amplitude signal,the recognition model is established with the recognition rate 93.18%.The statistical characteristic of the UHF PRPD spectrum is taken as input and the PCA method is applied to optimize the input used to optimize the input,to prevent huge computational cost and information redundancy caused by high dimension of PRPD statistical characteristic,to improve the efficiency of UHF diagnosis system,with the recognition rate 94.79%.Finally,based on the three kinds of identification information and the DS evidence theory,a partial discharge multi-source information fusion diagnosis system is established.Through the testing of various kinds of samples,it is proved that the system has better stability and fault tolerance and higher recognition rate,than the single sensor.Through two cases of 35 k V switchgear insulation defect diagnosis,the whole process of information acquisition,feature extraction,and fusion calculation in decision diagnosis is explained,and the reliability of multi-information fusion diagnosis system is demonstrated.Compared to the traditional detection and analysis model,the proposed multi-information fusion detection can make full use of the favorable information of field detection,enhance the recognition conclusion,and obtain a higher reliability of diagnosis result.