Study On Ultrasonic-based Method To Detect Partial Discharge In Transformer

In the dissertation, the main points of the author's study are ultrasonic detection of transformer partial discharge. It finds the way of real discharging magnitude estimation from partial discharging model. The equivalent circuit for partial discharge-generated ultrasonic waves is found. The rule of the ultrasonic wave propagating from its source to the sensor is gained. An ultrasonic-ultrasonic locating method with great practical value is put forward in the dissertation. The dissertation resorts to such nonlinear and non-stationary signal processing technologies as chaotic, fractal, and wavelets analysis to come into being characteristic vectors, then, the characteristic vectors are sorted by means of artificial neural network for pattern recognition. It bring forward the way of real discharging magnitude estimation from transformer. An ultrasonic measuring system of transformer partial discharge has been developed.A nanosecond-level equivalent circuit for partial discharge measuring is built according to actual cases. The formula for real discharging magnitude estimation is gained, and the validity of the model and formula are ratified by digital simulation and relevant experiments. Resorting to the method of electrical-mechanical-ultrasonical analogy, the dissertation offers the relevant model-of-gap-based equivalent circuit for partial discharge-generated ultrasonic waves. It makes clear three key problems of ultrasonic detection. The dissertation obtains the rules of the ultrasonic wave propagating from its source to the sensor. Also, through laboratory testing has been made clear damping and propagating time constants of ultrasonic wave and wave form changes as well. At the same time the relationship between the waveform and the propagating distance has been recognized.In the dissertation, three aspects affecting partial discharge location are addressed, relative time differences among the sensors, algorithm of location, and propagation path of ultrasonic wave. From a comprehensive view, an ultrasonic-ultrasonic locating method with great practical value is put forward in the dissertation. The method mentioned in the dissertation improves the precision of the location and effectively constrains divergence of discharge detection. In the laboratory, for the mutual correlation function of multi-source waveform is denser than that of the individual one. The dissertation resorts to such nonlinear and non-stationary signal processing technologies as chaotic, fractal, and wavelets analysis to come into being characteristic vectors. 11 characteristic data have been worked out, including the maximum Lyapunov index λ, fractional dimension D, lacunarity Λ and eight energies of various frequency bands. After being normalized, they can function as the characteristic vectors. Then, they are transmitted into an artificial neural network to perform pattern recognition.Starting from the mechanism of ultrasonic generating, initial ultrasonic pulses are calibrated. A stylish discharge source is directly coupled with the ultrasonic sensors, and work out actual discharge magnitude. The ultrasonic signal waveforms reaching the sensors are supposed to be recorded for calculation of magnitude, frequency and attenuation time, etc. Thus the basic calibrated information of the ultrasonic signals is obtained. Then, according to the Green function, the waveform from the sensors can be inversed to estimate the ultrasonic waveform at the actual discharge source. Finally, the characteristics of the resultant waveform should be compared with the initial calibrated information to calculate the actual discharge magnitude. The laboratory data has demonstrated the validity of this quantity method.An ultrasonic measuring system of transformer partial discharge, on Labview platform, has been developed. It consists of sensors, data acquisition board and computer. The system achieved welcoming results in terms of locationing, pattern recognizing and quantify magnitude. The experiments has positively the system can use on transform...