Download files
Abstract
Large numbers of installed medium or high voltage power cables are now of advanced age and have gradual insulation deterioration problems. On-line Partial Discharge (PD) measurement has inherent advantages over the conventional off-line measurement, but it also suffers from a very significant Electro-Magnetic Interference (EMI) problem due to the small PD signal levels being monitored. The large magnitude EMI signal often completely swamps the smaller magnitude PD signal, making it difficult to monitor anything but extremely large PD activity. To monitor cable condition and to be able to assess insulation degradation trends requires advanced techniques.
In this thesis, the issues of sensor selection, digital filtering, software based on-line differential technique and wavelet Transform (WT) de-noising techniques are studied. Their applicability, advantages and limitations are discussed with the simulation and high voltage measurement results. The High Frequency Current Transformer (HFCT) type PD sensor is selected after being compared with other sensor in term of sensitivity, universal applicability, frequency response and the installation difficulty.
The sinusoidal noise frequency identification and application of the conventional digital filtering are studied. The Least Mean Square (LMS) and Recursive Least Square (RLS) algorithm digital adaptive filters are compared in detail and RLS adaptive filter is selected.
Wavelet transform de-noising technique for on-line PD measurement is carefully studied and the applications of noise reduction are developed. A novel WT threshold value selection algorithm is presented in this thesis. The new WT algorithm is compared with the existing wavelet techniques using numerical simulation and laboratory high voltage testing data on cables. The results show that this new fully automatic WT de-noising method has achieved great progress with the capability of detection 30 pC PD signal during on-site on-line measurement where typical noise level is ten times higher in magnitude.
On the basis of traditional off-line differential or balanced detection circuit, a software based on-line differential technique is proposed in this thesis. The novel method developed has the capability to process one whole AC cycle of PD data. It enables traditional q-Φ and n-Φ distributions to be obtained as well as PD repetition rates and the usual integrated PD parameters.
These techniques are developed for the on-site on-line PD measurement in power cables, but they are not limited to cables. They can also be applied to other high voltage equipment with minor or without modification in the data acquisition procedure.