Research On The Monitoring And Fault Diagnosis For SF₆ Circuit Breakers

High voltage circuit breaker (HVCB) is one of the most important primaryequipment in power system, of which SF₆ HVCB is widely used because of itsgood quality of isolation and arcing, and so the malfunction of SF₆ HVCBdecreases the safety and reliability of power system directly. Gas failures andmechanical failuresconstitutequite a portion of all the failureswithin SF₆HVCB,the monitoring systems of which still have shortcomings. This paper makes greatefforts on moisture monitoring, vibration signal feature extraction, methods ofdiagnosis and their applications to improve the performance of monitoring andfaultdiagnosissystemforSF₆ HVCB.The precision of moisture monitoring is affected by the adsorption effectbetween circuit breaker and water, and no suitable method of quantify this effectcan be used. This paper describes adsorption effect by"adsorptionpotential-adsorption space"curve based on Polanyi adsorption potential theory.Formulas for conversion of moisture value between different temperatures havebeen developed. The validity of this method has been proved by experimentalapproaches.Time-frequency analysis is the important approach for extracting featuresfrom vibration signal, but no agreement has been made in this field. This papercompares the strength and shortcomings between short-time Fouriertransformation (STFT) and wavelet transformation (WT), and a new featureextraction method based on Hilbert-Huang transform (HHT) is presented. Thethree methods are employed to vibration signals of HVCB, and the clusteringquality of the features extracted is compared. Result shows that HHT is the mostidealmethodforanalyzingvibrationsignalofHVCB.Theexistingdiagnosis methods are lackofthe abilitytopursuethechange ofcertain mechanical state in state space. This paper presents a self-learningdiagnosis method for mechanical failures of HVCB on the basis of artificialimmune network memory classifier (AINMC). Comparison has been madebetween self-learning method and non self-learning method, and result shows thatself-learningmethodcanachievemoreprecisejudgmentofthemechanicalstateof HVCB.Finally, this paper presents a scheme of developing on-line monitoring anddiagnosis system for SF₆ HVCB. The monitoring principles, hardware structureand software flow are fullydiscussed. The system developed has been used in ChiFengbureauofelectricpowerandthemoisturemonitoringunitisgrantedapatent.