Separation of partial discharges from pulse shaped noise signals with the help of neural networks

The paper introduces a method to separate Partial Discharges (PD) from pulse shaped noise signals using a neural network. After a short introduction of the applied neural network, its ability to classify PD and pulse shaped noise is demonstrated on simulated PD and noise pulses. Furthermore, with the help of signals injected at different points of a transformer coil, the possibilities and limits of the neural network used for a distinction of pulses is shown. The paper ends with a presentation of the first results of the neural network in the classification of signals measured on a power transformer.<>