Research On The Type Identification Method Of Partial Discharge Signals For Electrical Equipment

The deterioration of the insulation for power equipment will directly affect the normal operation of the equipment,which is harmful to the safe and stable operation of the power system and has a negative effect on the power grid.Partial discharge is an important performance of insulation degradation.Identifying the type of partial discharge can assist the localization of partial discharge,which can repair the faulty equipment timely and ensure the safe operation of the power system.The pattern recognition of the signals can be realized by extracting the relevant characteristics and using its inherent characteristic attributes.Based on the partial discharge data generated by the insulation defect model built in the laboratory,the classification of the discharge signal is studied.The main contributions of this paper are as follows:A new"phase-gradient"spectrogram mode for gradient evaluation of discharge information was constructed.Firstly,used the PRPS sequence of each discharge to generate the corresponding PRPD spectrum.Secondly,φ-φandφ-u spectra were generated using theφ-q_maxspectrum in the PRPD spectrum.Then,the gradient information was obtained for theφ-φ,φ-u,φ-q_max,andφ-n spectra respectively.Finally,statistical parameter features were extracted fromφ-Δφ,φ-Δu,φ-Δq_max,φ-Δn andφ-n spectra.The experimental results show that the discharge characteristics obtained by the"phase-gradient"spectrogram mode have greater differences between different types,and the data distribution within the same type is more concentrated.A method of partial discharge signal identification based on variable predictive model and Tanimoto similarity（VPM-Tanimoto）method was proposed.Firstly,each type of VPM₁～_pwas trained using each type of discharge sample in the database.Secondly,the feature vectors of a certain sample were input into various VPM₁～_prespectively to obtain the feature prediction matrix of various models for the sample.Then,measured the similarity between the prediction vector and the sample to be identified with Tamimoto similarity.Finally,the magnitude of each similarity was compared,and the model category where the maximum similarity was located was taken as the sample category.Aiming at the problem of non-known discharge-type signal recognition in the discharge signal,an unknown type sample classification method based on VPM-Tanimoto principle was proposed.In order to determine the presence or absence of an unknown type of signal,firstly,all signals were identified by the ordinary VPM-Tanimoto method.Secondly,the reliability integrator discriminant analysis rule（IDAR）of the recognition results was calculated.Then,the IDAR indicator was divided into two thresholds,and appropriate recognition rules were selected for each interval.Finally,the Tanimoto similarity and the IDAR of each region were used to double filter all signals to separate unknown samples.The experimental results show that this method can determine whether there are unknown signals in the sample set,and separate them,which has a certain recognition effect.