The Study On Patter Recognition And Discharge Quantity Estimation Of Partial Discharge Based On Optical Method

Gas Insulated Substation (GIS) is widely used in urban power network because ofits advantages of small occupying area and stable operation. However, there areinevitable insulation defects which may cause major insulation accidents in GIS. Aspartial discharge (PD) is an important parameter which reflects insulation condition ofGIS, PD on-line monitoring in GIS becomes hot research. As the optical method of PDhas the advantages of high sensitivity、strong anti-interference ability and the sensorswhich can place into complex insulation structure, PD on-line monitoring in GIS hasbroad application prospects.Based on the research results of PD recognition and technique of the opticalmethod of PD, the article studies on PD recognition based on the optical method and itssignal corresponds to the charge quantity of PD to provide a basis for PD on-linemonitoring and fault diagnosis, the main work and achievements are as follows.(1) Depending on the common types of insulation defects in GIS, the authordesigns four insulation defect model. The fluorescence fiber sensor system for detectingPD is adopted to obtain experiment data of PD four insulation defect model underdifferent strength. Moreover, the φ-u-n spectral and gray images based on the opticalmethod are built under the PD induced by different defects. Experimental results showthat the signals of the optical method have different characteristics under PD induced bydifferent insulation defects, and the PD recognition based on the optical method isfeasibility.(2) Based on the theory of multi-fractal and definition multi-fractal spectrum, themethod of calculating the multi-fractal spectrum of gray images formed by PD for thefirst time. In the analysis of the physical meaning of the multi-fractal spectrumcharacteristics of PD gray images,16features of the multi-fractal spectrum of PD grayimages formed are extracted. And the16characteristic features are used as the featuresof PD recognition.(3) Based on the study of theory of artificial neural network, with the input of thecharacteristic features proposed in the article, a back propagation neural networkclassifier which is optimized by algorithm of modified conjugate gradient. The resultshows that its recognition rate is higher than the recognition rate which uses differencebox counting and information dimension as the characteristic features. It indicates thatthe characteristic features proposed more reflect characteristics of PD under different insulation defects. At the same time, it also proves the signal of PD based on the opticalmethod can identify the different types of discharge.(4) The integral value of the signal of the optical method reflects naturedly theenergy of light released by PD from the principle of the optical method, then it isdeduced there is linear relationship between the integral value of the signal of theoptical method and the charge quantity of PD, to provide guidance of determiningcharge quantity of PD in GIS based on the optical method.(5) The PD experiment under four types of insulation defects designed in thearticle is done once again, and applying the optical method and current method detectthe PD at the same time to study the relationship between the signal of the opticalmethod and the apparent charge quantity of PD. The experiment result shows therelationship between the average internal value of signal of the optical and the averageapparent charge quantity is linear, but the linear ship between the average internal valueof signal of the optical and the average apparent charge quantity is different because thenature of PD under different insulation defects is distinct. In addition, the distance fromthe needle to the plate influences the linear ship between the average internal value ofsignal of the optical and the average apparent charge quantity. The distance is greater;the slope of the linear ship is greater.The research result shows the type and level of PD can be determined based on theoptical method, so the PD online monitoring and fault diagnosis based on the opticalmethod is feasible.