Research On New Fault Feature Information In Transformer Oil Based On THz-TDS

Oil immersed power transformer is widely used and plays an important role in power system.Accurate and timely diagnosis of transformer fault is of great significance to the safe and reliable operation of transformer and power grid.At present,the most commonly used,sensitive,accurate and on-line diagnostic method is dissolved gas analysis(DGA),which can diagnose the existing or latent faults in transformer and their severity,but also has some shortcomings such as limited criteria.Therefore,it is of forward-looking significance to seek new technical means to extract new fault feature information of transformer.Based on the original time domain spectrum data of transformer insulating oil in different aging stages,this paper attempts to use terahertz(THz)time domain spectrum technology to extract and summarize the corresponding new fault feature information.The main research contents and achievements are as followsFirstly,the generation mechanism of gas in transformer oil and the corresponding fault types are described in detail,and the feature extraction method of transformer heating and discharge fault based on DGA results is analyzed,as well as the application,characteristics and limitations of conventional diagnosis methods.Secondly,the generation and propagation characteristics of THz electromagnetic wave,the principle of THz time domain spectroscopy,and the composition of time domain spectroscopy detection system are analyzed.This paper focuses on the production and requirements of the oil sample,the characteristics of the time domain spectrum signal output by the system and its influencing factors.Thirdly,for THz time-domain spectrum is easily affected by noise,the sparse decomposition method is adopted,and Gabor atom library is used to denoise THz time-domain spectrum signal.According to SNR,MSE,NCC and JBL,the denoising effects of sparse decomposition,traditional wavelet analysis and empirical mode decomposition are compared.The results show that the sparse decomposition method is more suitable for THz time domain spectral signal denoising.Finally,based on the time domain spectral data of transformer oil samples,the time domain and frequency domain characteristics corresponding to different aging stages are extracted and summarized.A total of 12 time-domain waveform characteristic quantities,including three troughs,three peaks and their amplitude attenuation degree,are constructed to characterize the time-domain spectral characteristics.The variation of the above characteristic quantities in different aging stages is analyzed.The results show that these characteristic quantities are different and distinguishable in different aging stages,and can be used to identify different aging stages.The best frequency bands of absorption coefficient spectrum,refractive index spectrum and extinction coefficient spectrum are selected by using the self defined convergence factor FCT.The results show that there is a regular relationship between the change of spectrum amplitude of the above three coefficients and different aging stages in their optimal frequency bands.The time domain spectra of transformer oil in different operation years are analyzed,and the results confirm the availability of the time domain and frequency domain characteristics.The research in this paper has reference value for the application of time domain spectroscopy in transformer fault diagnosis show that these characteristic quantities are different and distinguishable in different aging stages,and can be used to identify different aging stages.The best frequency bands of absorption coefficient spectrum,refractive index spectrum and extinction coefficient spectrum are selected by using the self defined convergence factor FCT.The results show that there is a regular relationship between the change of spectrum amplitude of the above three coefficients and different aging stages in their optimal frequency bands.The time domain spectra of transformer oil in different operation years are analyzed,and the results confirm the availability of the time domain and frequency domain characteristics.The research in this paper has reference value for the application of time domain spectroscopy in transformer fault diagnosis.