Employing Artificial Neural Network as a Novel Method for De-noising of Partial Discharge Signals

摘要

Partial Discharge (PD) intensity assessment in the insulation of high voltage equipment is a very powerful approach for improving the stability of power systems. The evaluation of PD, however, has become complicated in the presence of noise, due to its nature, intrinsic limitations, much diversity, and low amplitude of the generated signals. As one of the most essential parts of the process, de-noising of PD signals can be referred to. Hence, an effective algorithm to do so, Wavelet Transform (WT) has long been utilized, reviewed and studied in the paper. However, that has demonstrated numerous limitations to achieve high performance. Therefore, the present paper seeks to introduce a new method using the curve fitting abilities of Artificial Neural Network (ANN) for de-noising of PD signals. The obtained results for different cases and noise levels, for simulated PD signals, prove the superiority of the proposed method compared to the conventional WT-based algorithms.