Study On Measurement Of Partial Discharge In Large Generator Stator Winding Insulation And Its Aging Characteristics

Large generator is the key apparatus in power system, which has been paid great attention for its operation reliability attributes to the stability of power grid. Insulation system is one of threats to its safe operation. Stator winding is often subjected to combined action of electrical, thermal, mechanical, and chemical factors which results in insulation performance deterioration. When the degradation becomes worse enough, insulation fault comes out. Therefore a diagnosis of insulation condition to stator winding is an extremely important means to the assessment of electrical machine, especially for large one's insulation system. Research on relationship between stator insulation characteristics variations under multi-factor aging condition with insulation, measurement technology of partial discharge (PD), assessment of stator insulation based on signal process technology are of great significance.This paper establishes multi-factor and function partial discharge measurement system. Detecting impedance is replaced by current transducer for PD measurement, and a current response transducer is put forward for signal sensing which largely improve the performances of amplitude and phase characteristics versus frequency respectively. Based on complex programmable logic device (CPLD), a scheme for PD data acquisition system with multi-channel, variable sampling frequency and large storage capacity is put forward for the first time. Advanced electric technology and software are introduced to acquire effective PD signals with a high speed and accuracy, and low speed for invalid signals. The validity of data acquisition is achieved, and the goals of improving efficiency of memory and prolonging continuous acquisition time are also fulfilled. Synchronous dynamic random access memory (SDRAM) substitutes static random access memory (SRAM) for buffer memory, which enlarges capacity of data acquisition system and reduces the cost of hardware.The measured PD signal contains many noises. White noise and periodic narrowband noise are two important interferences. Based on existed wavelets method for white noise depression, the different Lipschitz exponent between white noise and PD signal, and their different characteristic in the time-scale plane, this dissertation bring forward a pragmatic method for PD signal processing by which there are no need to get transcendental knowledge about signal and proper external parameter setting. It is a good way to eliminate white noise and can preserve even the slight PD signal.Distortion of PD after noise eliminating operation become small, and can be used in PD signal processing. Wavelet package has better decomposition characters than wavelet, and can demarcate signal frequency space more effectively. Thus, it is used for eliminate periodic narrowband interference. According to the analysis of its decomposed characters in PD signal processing and research on periodic interference, this paper put forward a periodic interference denosing method based on optimal wavelet package basis by choosing minimum non-zero coefficient in wavelet package decomposing as object function. Many examples indicate that it is a good way and has no need to set external parameters.This dissertation have done multi-factor aging experiment under electrical, thermal, hot and cool circular circumstance to 200MW generator stator bar, studied on the variation of PD parameters with aging time, the phase distribution of PD aging parameters, and the changing rules of PD statistical factors. A series of conclusion are achieved: Discharge gradient of largest discharge quantity and number of discharge increase as the insulation aging time prolonged; Discharge phase can partially reflect aging degree, whose width extend also with the prolongation of aging time. The skewness and kurtosis have close relation with aging process, which skewness decrease and kurtosis increase with the aging time.