Study On Partial Discharge Detect And Signal Process In Power Transformer

Insulation fault is the most important part in the fault of the large scale power transformer, there are imitate relationship between the partial discharge and insulation in the power transformer, so detect the insulation situation by detecting the partial discharge in the power transformer has significant meaning. The traditional detect methods of detecting partial discharge have their limitations, ultra high frequency (UHF) detect method is a method developed in recent years which receive the high frequency electromagnetic generated in the partial discharge by the high frequency sensor, it can avoid the periodical interference signals, and has higher detect accuracy.The generating mechanism of the partial discharge and the propagation of electromagnetic wave generated by partial discharge are learned in the paper. Relay on the character that the partial discharge can radiate electromagnetic the frequency of which can up to 3GHz, a detect system which contained high frequency sensor and preposing pretreatment circuit in the paper. First design an internal plane sensor which has a simple structure, small size, conformal with the internal wall of the power transformer case, suit to install from the bleed hole in the transformer; Next choose a low noise amplifier in the pretreatment circuit based on the performance of the designed sensor, and design a cross coupling filter which can well solve the shortcoming of the high order filter in the partial discharge, the high order filter has high request of the machining accuracy and difficult to tune itself; Then create the 3D model of the designed sensor and cross coupling filter and simulate by the high frequency simulation software HFSS, transformer the antenna and filter to equivalent circuit and simulate the whole detect system.Although the UHF detection method can avoid the low frequency periodic interference signals, the white noise cause by the hot noise of the device can not be avoided. So a signal processing method of white noise based on wavelet packet is improved, a method choose optima wavelet based on similar between is used, a least entropy metric is used to choose the optima decomposition tree of the wavelet packet and a threshold function is developed, last a simulation is given to detect the effect of the method.