Research On Feature Extraction And Identification Of Power System Operating Overvoltage

The electric power system can not avoid various operation on power system components during its daily operation. At the same time, there are many lighting, grounding or resonance phenomenons in power system. All the above phenomenons will lead to overvoltages, and overvoltages will endanger the whole power system. With the continuous development of detection technology and materials science, overvoltage monitoring technology has a great improvement. The power system overvoltage waveform data that contained lots of time domain and frequency domain informations and directly reflecting power system operation state is more and more accurate. So the feature extraction and recognition algorithm of overvoltage are significant. In view of the different overvoltages which mainly considered the operation overvoltages, a set of hierarchical stepped identification framework is designed.This framework mainly considers three factors: Firstly, a large number of references related the research. Secondly, the current status of overvoltages classification both at home and abroad. Thirdly, the dependency relationship between different overvoltages and the recognition system expansibility. The proposed framework can effectively solve the problem of low efficiency and low expansibility of the program using single feature and complex algorithm in the traditional way. In order to solve the problem of overvoltage feature extraction, three kinds of techniques(time domain analysis, FFT transform and wavelet analysis) are used to find the most simple and effective distinguishing feature among different overvoltages.In the end, this paper uses different features to distinguish different overvoltages:Using the time domain characteristic of the wave front time to distinguish the lightning overvoltage, using the characteristic of only containing fundamental component after add window FFT transform to distinguish the power frequency voltage rising, using the characteristic of energy entropy of ninth detail layer after 1.5 period after wavelet transform to distinguish the intermittent arc grounding overvoltage, using the characteristics of energy entropy of fifth, sixth, eighth detail layer after wavelet transform to distinguish the closing capacitors overvoltage, closing unload transformer overvoltage and closing/opening unload line overvoltage. In order to solve the problem of overvoltage identification algorithm, three kinds of techniques(set the threshold, multifactor fuzzy identification, fuzzy cluster analysis)are used to quickly and effectively complete the specific overvoltage identification. In the end,this paper uses different ways to distinguish different overvoltages:Using setting the threshold to distinguish the lightning overvoltage, the power frequency voltage rising and the intermittent arc grounding overvoltage, using two different ways(multifactor fuzzy identification, fuzzy cluster analysis) to distinguish the closing capacitors overvoltage, closing unload transformer overvoltage and closing/opening unload line overvoltage. At the same time, the advantages and disadvantages between multifactor fuzzy identification and fuzzycluster analysis are compared. Finally, the calculation and analysis of the measured data and simulation data verify the effectiveness of the recognition framework, feature extraction and recognition algorithm. In order to apply the system to the practical engineering, the shortcomings of the system and the improvement directions of the system are discussed.