Noise suppression and data reduction for partial discharge measurements using orthogonal transformations

The measurement of partial discharges (PD) at power transformers as a basis for a monitoring system gains an increasing importance. Unfortunately, the PD measured on site are often superposed with disturbances. Further problems are caused by the processing of the huge amount of data results from the usually long periods of supervising. This paper presents effective methods for noise reduction as well as for data reduction via orthogonal transformations. On synthetically generated signals it is demonstrated that the choice of a suitable orthogonal transformation concentrates the energy of the signal in a small area of the obtained spectrum. With the help of measured signals the qualification of orthogonal transformations for the PD measurement is approved. One of the methods used for noise reduction limits the disturbances via a transformation to a few spectral components. After erasing these spectral components and inverse transforming, the noises in the time-signal are suppressed. Another method which is approved here, limits the PD parts of the measured signals to a few spectral components via a transformation. Then, during the inverse transformation, only these components are considered. The second method mentioned additionally gives the user the possibility of an effective data reduction, whereby the error can be limited by a useful choice of the used spectral components for the inverse transformation. A comparison of the data reduction via orthogonal transformations with other conventional methods at the end of the paper shows that this method is much more effective and advantageous because the time signal remains neatly unchanged