| Summary: | Partial Discharges are symptoms of insulation degradation that continually promotes the deterioration of the insulation condition, eventually leading to permanent failures. The PD event itself is not dangerous but it is the state of the discharge activity that can lead to unforeseen failures. Hence continuous monitoring of PD enables the prevention of these unforeseen failures that result in economic losses. High voltage equipment that has been installed more than forty years ago is highly susceptible to PD since it is reaching the insulation end life. For the case of underground cables, continuous PD monitoring will help avoid unplanned outages and the need for the immediate replacement of faulty cable sections that will incur large costs. Traditionally, PD diagnostics have been carried out through offline methods, whereby the cable specimen is removed from service during the diagnostic tests. On the other hand, online PD diagnostics are preferable since the services are not disrupted. However, there are some major challenges that come with the online approach. For instance, heavy noise interference, the detection and location of PD through the interpretation of measured signals. Presented in this Thesis are presented novel contributions to the area of online PD detection for underground HV cables. The work encompasses aspects of physical data acquisition procedures and the post-processing signal processing algorithms. A new online PD data acquisition unit equipped with pre-processing and signal conditioning is presented. The system has other features that take into account the difficulties of logistics as well as issues related to the regulations and protocols of the utility. The primary aim of this developed system is to produce a PD database that will be used for research purposes. Since applied wavelets for the field of PD diagnostics have been popular amongst researchers, it was investigated further in this Thesis. Following from the data acquisition system, large data sizes required intensive processing. A new wavelet-based algorithm combined with the use of Higher Order Statistics is presented. This algorithm enables the simplification of data signals to highlight potential PD activity resulting in the reduction of manual examination of wavelet algorithms. The conventional wavelet algorithms applied in the literature generally referred to a specific approach of the wavelet implementation i. e. the decimated approach. However, it was found that the non-decimated approach has several advantages with regards to PD signature detection and PD location. The application of both approaches and their comparisons is applied to simulated data as well as online field data. Finally, the analysis of online PD data acquired from the new data acquisition system is presented. Several PD characterisation processes are applied and positive results were generated. Aspects relating to the physical environments of test site are also included. The challenges and experiences gathered in this research project are described.
|
|---|
