Using The Wavelet Transform To Extract The Partial Discharge Signal

The partial discharge (PD) is produced because there is uneven electric field in the part of insulator, and when there is high partial electric field, which will lead to dielectric breakdown within the local area of the insulator or the electrical discharge phenomena. In order to monitor online and diagnose insulation condition of transformer,foreign and domestic scholars have done a lot of work to study the electrical characteristics which reflect the state of transformer insulation such as:leakage current, partial discharge and dielectric loss, and they found that the parameter of partial discharge had much better sensitivity and more comprehensively reflect the information than the other parameters. PD can lead to the insulation breakdown of electrical equipment, and is the harbinger of insulation deterioration. When electrical equipment is in the actual producing process and operation process, some weaknesses of insulating material will affect the inherent quality and reliability when running.Therefore, monitoring partial discharge signal online to judge its development trend, the type and intensity is very necessary, and can achieve maintenance of the state, which can prevent and reduce the occurrence of sudden. Meanwhile, it is very important for the safe operation of the whole system and will bring huge economic and social benefits.However, the partial discharge signal is particularly weak.If it has the strong electric field interference, it will be very difficult for us to extract the amount of electrical characteristics and do the pattern recognition. In result, monitoring online partial discharge is very difficult to achieve practical application level. Partial discharge signal has the strong mutation and strong singularity, and it is not suitable to use the traditional method of Fourier transform. The wavelet transform has the excellent localization qualities in the time and frequency domain, and it is very suitable for us to deal the signal of mutations. Because the wavelet transform has superior function of noise cancellation and can capture transient signals, the wavelet analysis can be regarded as a powerful tool to eliminate noise and extract PD signals. In this paper, the wavelet transform is used to process and analyse the simulated signals and experimental signals of partial discharge, and the main work and the related conclusions are as follows:(1) The reasons for the formation of partial discharge and the types of partial discharge are introduced simply. According to the occurred location of partial discharge, formation mechanism and the discharge process, it can be divided into the internal partial discharge, the surface partial discharge and corona discharge. The time of partial discharge is rather short, so we can make the partial discharge signal be equivalent to four models, and namely the single exponential decay model, the double exponential decay model, the single exponential decay surging model and the double exponential decay surging model.(2) The origin and role of the wavelet transform is described, and the wavelet transform is evolved by the fourier transform, which experienced the fourier transform of window. However, the wavelet analysis theory and the construction of wavelet function are from fourier analysis, so they are closely linked. And then the definition of continuous wavelet transform and its properties are elaborated, including the linear, the time scale theorems, differential operators and energy conservation. Finally, in order to facilitate to do comparison and analysis of several kinds of wavelet, a nature list of the brief description is given.(3) The theory and the step of wavelet denoising is illustrated briefly. However, the signal processing which used by the wavelet transform is directly affected by many factors, such as the wavelet function, the threshold selection rule, the way of threshold processing, the wavelet decomposition scale and so on. In this paper, the correlation coefficient is used to determine the result of signal processing, and the signal reconstruction is best when the bior wavelet has the soft threshold method, the vanishing moment of 6.8 and scale parameter of 7. Finally, the corresponding parameters are used to process the simulated signals of partial discharge and experimental signals, and the effect of noise elimination are very obvious.The wavelet function is not determined and has diversities, so how to select the appropriate wavelet basis function is more difficult in practical engineering applications, and it is a particularly important issue because there will be different results produced when we use different wavelet basis functions to process the same problem.So far, we mainly use the correlation of the results after the wavelet analysis and the original signal to judge the good or bad of the wavelet function, and the wavelet base function is choosed finally. In this paper, according to the theoretical analysis, the corresponding parameters of combination to the best coefficients of the bior wavelet is gotten, and the results are used to process the experiment signals and bring more obvious effect of noise elimination.Theoretical and experimental results show that:when the appropriate vanishing moment parameter, scale parameter and the threshold approach are selected, the wavelet transform can effectively eliminate the noise in the partial discharge signal.