Research On Partial Discharge UHF Detection Of Large Power Transformer Based On Em Wave Leaking From GAP

Partial Discharge(PD)is an electrical discharge process in which insulation is partially broken down between conductors.In the insulation system of electrical equipment,PD is similar to "Insulation Tumors".If it cannot be discovered and eliminated in time,the long-term discharge will lead to further deterioration of the insulation,which will eventually lead to the failure of the entire insulation system.Therefore,it is of great significance to carry out technical research on the detection of PD insulation defects of power transformers,especially large power transformers.The PD detection method based on Ultra High Frequency(UHF)electromagnetic pulse coupling technology has been widely used in the field of PD insulation detection of power equipment due to its strong anti-interference ability and non-contact measurement.This paper focuses on the topic of "Research on Partial Discharge UHF Detection of Large Power Transformer Based on EM Wave Leaking from Gap",using the characteristics of large-scale power transformer non-metallic insulation gasket gap to leak electromagnetic waves.The conducted detailed researchs are:(1)The characteristics of leaked electromagnetic wave gap propagation are studied.The vibrator antenna is used as the carrier of electromagnetic wave signal radiation.Finite difference time domain method is used to study the propagation characteristics of PD source radiated electromagnetic wave signals in the transformer gap in different directions,as well as the gap height and depth,and carry out experimental research on the real transformer tank.It is found that the electromagnetic wave intensity in the gap is much larger than the electromagnetic wave intensity outside the tank,and the electromagnetic wave signals radiated by the PD source in different directions are quite different in the gap.As the gap depth decreases,the electromagnetic wave intensity in the gap shows a gradual increase,and there is a tendency for the strength to increase at the tail(junction between the gap and the free space outside the transformer)of the gap.(2)The research on gap UHF antenna sensor is carried out.Based on the unidirectional propagation characteristics of leaked electromagnetic wave,electromagnetic shielding and easy installation of antenna sensor,the two-dimensional Vivaldi antenna sensor with metal shielded casing is optimized by finite element method.The linear gradient microstrip balun effectively improved the return loss of the antenna sensor UHF band,its working frequency range is 807MHz-3GHz,the maximum gain is 9.4dB,and the double-layer metal box shielding shell is designed by the resonant cavity principle.The receiving performance of the proposed Vivaldi antenna was tested on the transformer PD experimental platform.It was found that the amplitude of the response of the Vivaldi gap antenna to the UHF signal is much larger than that of the externally used helical antenna.(3)The research on PD source location based on gap leaked electromagnetic wave detection is carried out.It is difficult to eliminate the singular value in three-dimensional space and the three-dimensional singular value in space,which will cause the waste of effective data,and proposes method of singular value elimination.The proposed method maps the three-dimensional coordinate sample set of the PD source to the one-dimensional coordinates,and uses the improved box-line diagram method to eliminate the singular values on the basis of the one-dimensional coordinate set,and uses the cyclic progressive least squares method to find the best PD source location,improving the location accuracy of the planar layout UHF antenna array from the dm level to the cm level.(4)Aiming at the problem that the z-coordinate of the PD source result of the planar layout UHF antenna array has large fluctuation error,the reason of z-coordinate error of the PD source result by the time difference method is analyzed.It is found that the layout of UHF antenna array leads to the conclusion that the proportion of variable z decreases in the optimization equation.Based on this,a stepwise calculation method of transformer PD source z coordinate based on planar layout UHF antenna array is proposed,which converts the acquisition of PD source z coordinate into univariate.The results show that the stepwise calculation method can reduce the fluctuation variance of the z-coordinate of the PD source from 0.6 to 0.2.When the time difference sample is ?10,the average absolute error of the z-coordinate is reduced from about 0.4m to less than 0.1m.(5)The z-coordinate step calculation method of transformer PD source based on planar layout UHF antenna array for content(4)has the problem that the z-coordinate calculation effectiveness is not good when the time difference sample set is small.The PD source based on golden section tree optimization algorithm is proposed.The positioning method introduces the one-dimensional golden section algorithm into the three-dimensional space,realizes the global-local alternate optimization in the global scope,and finally analyzes the influence of the golden tree depth on the positioning result,and finds that the preferred depth of the golden section tree is N,N>8,if the computational efficiency of the device allows.(6)The research on transformer PD pattern recognition work is carried out.The problem of the statistical feature quantity for PD pattern recognition need the expert experience and lacks certain generalization.The CNN convolutional neural network mode pattern recognition algorithm lacks the time series feature information of PD map.The PD PRPD pattern pattern recognition algorithm of CNN-LSTM which combines the advantages of CNN network to exploit the local spatial information of PRPD map and the advantages of LSTM long and short memory network to exploit the time series feature information of PD map is proposed.The feature vector output by CNN pooling layer is used as the input of LSTM.The spatial characteristics and timing characteristics are model-recognized by Softmax classifier.The performance of CNN-LSTM is compared with CNN network and LSTM network,which shows that the overall recognition performance of CNN-LSTM network is better than CNN and LSTM.