Electrical Device Failure Diagnosis Research Based On The Analysis Of The SF₆Gas Decomposition

Sulfur hexafluoride (SF6) gas is widely used in high-voltage electrical devicesfor its excellent insulating and arc-extinguishing properties.When electrical deviceoccurs in discharge failure, SF6gas would decompose.Those decomposed gases willreact chemically with the impurities in the gas chamber and then generate a series ofoxyfluorid. The traditional electrical experimental methods have lots of limitations infault diagnosis. So applying the decomposed gases components of SF6to diagnosefaults and assess the condition of the devices plays a very important role in the earlydiagnosis and prediction of device’s failures.This paper mainly consists of four parts of research:(1)Clarify the decompositionmechanism of SF6gas,influencing factors, relative contents and trends of SF6decomposition components.(2)Establish discharge failure model and insulation defectmodel of SF6electrical device. Select appropriate gas composition or the ratio as thefeature parameters for these two types of failure.Develop fault diagnosis procedure inMatlab based on the support vector machine multi-class classification.(3)Selectmonitoring gases and fit the gases’ curves to predict failure trends and assess device’scondition.(4)Develop ‘failure analysis system’ in Excel VBA to provide a platformfor the fault diagnosis and condition assessment.This paper mainly has the following three results:(1) Successfully developed thefault diagnosis procedure. The test results show that the fault diagnosis procedure canmake correct diagnosis of discharge fault types and insulation defect types accordingto the needs.(2) Selected SOF2gas and SO2F2gas as the monitoring gases. The fittingfunction of the gases curves is very ideal. The fitting function can forecast the contentof SO2F2gas or SOF2gas at any time in the future of the faults.(3) Successfullydeveloped the "fault analysis system". It provides a good platform for operatingpersonnel to realize fault diagnosis and prediction.