| Partial discharge (PD) is one of the important reasons causing power transformer insulation degradation and breakdown. It is meaningful to detect and diagnose partial discharge in power transformer. The propagation principle of PD pulse along the winding is studied and it is used for locating PD in power transformer. A novel method to reject field interference and an effective method to realize PD pattern recognition are also investigated. Based on multi-conductor transmission lines (MTLs) theory, a new model of partial discharge pulse propagating in single phase transformer is established. The parameter matrices C in the model are calculated based on static–steady electromagnetic field theory, and other parameters R, L, G result from the relationship between the parameters. The transformation of mode voltages and mode currents is applied in order to obtain MTL general solution, and the transfer functions of transformer winding are obtained. The pulse method and scattering parameter method for measuring the transfer characteristics of transformers winding are presented. The measuring results show that scattering parameter method has higher precision, and can eliminate the effects of reflection and noise. The experimental results tested on 180-turn single winding and 400kV transformer winding consistent with the simulation results derived from their MTL model. The basic idea of transfer function-based PD localization is analyzed. Sectional Winding Transfer Functions (SWTFs) can be obtained through direct measurement and MTL model simulation. Two applied electrical methods to locate PDs in transformer are presented. One is the method of distance function, the other is the method of energy-ratio curve-line. Based on the MTL model, the first method is firstly the transfer function ratios with PDs at different locations are acquired, then the distance of transfer function ratios between transformer test and simulated results are calculated. The location of PD is determined by the value of distance function. Single winding and 400kV transformer winding test results show the validity of the method. The second method depends on the energy of the signal in the frequency from 100 kHz to 4MHz, and the relationship between energy ratio and location is described by energy-ratio curve-line. 400kV and 500kV transformer windings test results show their location error is less than 6.25%. A novel integrated morphology-wavelet filter is used for suppressing PD field noise. At first, a kind of morphological open and closing operators combination filter is designed to filter periodic narrow bandwidth noise. Then wavelet singularity analysis is applied to remove white noise. Simulations and field measuring results show that their NNR are respectively 85.21dB and 37.66dB. So it is effective in suppressing fielding noise and applicable in PD detection. PD pattern recognition is an effective method to evaluate insulation condition of power transformer. As morphological image tool, granulometry is useful for estimating object size and granularity in grayscale image, or characterizing textures based on their pattern spectrum. Based on mathematical morphological multi-scale opening operation, PD pattern spectrums are extracted. The pattern spectrum is used as feature of PD grayscale image. Six kinds of typical discharge models are designed and their pattern spectrums are calculated, then a double hidden-layer neural network is applied for PD pattern recognition. The analytical results show that the proposed method is effective in PD pattern recognition.
|
PDF Full Text Request