| On-line monitoring and recognition of partial discharge (PD) occurred in HVswitchgear is an effective way to evaluate internal insulation condition and status. It canhelp to make pre-fault decisions in a timely manner to ensure the switchgear’s safety.Traditional ultra high frequency (UHF) detection acquires complete waveform of PDsignal and diagnoses discharge types by analyzing the features of PD signal, thus it setsdemanding requirements on sampling rate, transmission bandwidth and signalprocessing speed of the detection system. Novel detection method and diagnose theoryneeds to be refined and proposed. Since energy features of UHF PD signal is capablefor fault analysis and diagnosis, designing one kind of UHF antenna with largebandwidth to collect UHF PD signal, developing on-line monitoring system with signalpre-processing circuit by extracting sub-band signal energy to lower the demandingrequirements of traditional UHF method, and proposing an pattern recognition methodfor PD fault analysis using cluster algorithm, is of theoretic significant and applicationvalue for PD on-line monitoring and recognition.Based on the traditional UHF detection, this thesis analyzes the transmissioncharacteristics of electromagnetic wave occurred in switchgear, proposes an novelpattern recognition method for PD fault analysis by extracting features of PD UHFsignal energy. The whole thesis is summarized as:①Depending on size and material parameters, a real model of HV switchgear isestablished. Impact of four major components in the switchgear on electromagneticwave propagation characteristics is analyzed and summed up. Considering issues suchas overall detection performance, volatility of test result of six different PD sources andsafe distance, optimal installation location of antenna sensor is determined bycomprehensive comparison of detection performance under different locations. Thispoint is validated to be of the best performance by measured results obtained from thetesting platform.②Based on traditional microstrip antenna, a fractal antenna with large bandwidthis designed. Combined with the spread fractal theory, a novel UHF fractal antennasensor is developed by analysis and simulation and adjustment of the antenna structuresize. The sensor promotes the signal receiving performance largely with the receiving0.4-1GHz and average gain1.7dB. Full range of electromagnetic signals can be received. Moreover, it has compact size and can be easily installed. The receptivity ofnew UHF fractal antenna achieves the overall ascension, which is a kind of sciencebreakthrough.③For the first time, based on mixing theory and power envelope detectiontechnique, on-line monitoring system for UHF PD signal with signal pre-processingcircuit by extracting sub-band signal energy is developed. White noise level,out-of-band noise suppression, I/O response characteristics, and gain-frequencycharacteristics are tested and results show that the circuit is stable and reliable, and canrealize sub-frequency output and energy extraction, which is a new signal dealing wayfor UHF signal.④Four kinds of typical insulation defects are set in real test platform to simulatedifferent PD type. Aiming at various testing results under different testing voltage anddifferent insulation, the proportions of three sub-band energy to whole band energy arecalculated as features for PD pattern recognition. By establishing a fingerprint database,the energy features of UHF signals with cluster characteristics is validated. For the firsttime, Mahalanobis distance with frequency divide theory is introduced as criterion forenergy based PD pattern recognition. Back sentenced algorithm shows that themisjudgment rate was0.875%.In summary, the developed UHF fractal antenna sensor with large bandwidth hasexcellent performance, and the signal energy extraction circuit can effectively lower thedemanding requirements on sampling rate, transmission bandwidth and signalprocessing speed of the traditional UHF detection system. Energy feature based PDpattern recognition has become the latest hot spot and theoretical breakthrough of PDmonitoring, which will have broad academic value and application prospect. |
PDF Full Text Request