Data compression and pattern recognition for partial discharge ultrasonic signal based on fractal theory

According to the complex, nonstationary and nonderivative characteristics of partial discharge (PD) ultrasonic signal, the piecewise self-affine iterated function system (IFS) model of PD ultrasonic signals is established. The IFS parameters of this model are calculated to realize data compression. The IFS fractal dimension is extracted based on this model which is provided for partial discharge pattern recognition. The novel route is that both the tasks of compression and feature extraction in a single step, affords a new tool for on-line monitoring of partial discharge in a transformer.