| UHV AC transmission lines as an important electrical equipment which put power supply and user together.Its failure rate was high due to the complex work environment and the varied natural conditions. Automatic reclosing can put lines into operation fast after instantaneous fault, thus its widely use plays a vital role in improving power system stability and power supply reliability. However, now automatic reclosing is unable to identify fault types, this defect lead automatic reclosing to act again after permanent fault, therefore circuit breaker is shocked Two times within short time or the unstability of the system oscillation is produced. So how to judge fault properties effectively is particularly important. Based on the research status of adaptive reclosing, and the deep research and analysis of arc fault features when single phase short circuit grounding fault happens, this paper put forward the reasonable way of fault feature extraction and the method of fault diagnosis model construction. Specific content as follows:(1) First, the mechanism of circuit ground fault arc generating and the dynamic characteristics of an arc and the secondary arc were analyzed, the instantaneous single-phase short-circuit ground fault model of1000kV UHV AC transmission lines was built in ATP-EMTP electromagnetic transient simulation software platform, and nonlinear arc grounding resistance model was constructed with TACS and MODELS in ATP-EMTP. Taking simulation results of fault phase voltage of the power supply current curve and the fault point voltage current curve as the research object, through the comparison and analysis of the waveform and the phenomenon of fault arc which was researched by domestic and foreign scholars, analysis results shows the arc fault simulation model was constructed in this paper can realistically simulate the grounding arc dynamic characteristics in actual fault cases.(2) By analyzing the causes and characteristics of high frequency transient components in the fault phase terminal voltage waveform, the transient signal time-frequency analysis method based on wavelet transform principle was studied in this paper. Finally, fault signal feature extraction method by wavelet packet decomposition and reconstruction algorithm and energy spectrum analysis principle was put forward. On the data basis of the fault phase terminal voltage, db15was selected as wavelet basis function. Eight continuous energy spectrums were extracted through three layer wavelet packet transform. Experimental results show that the spectrums of high frequency component caused by primary arc and the secondary arc could be accurately distinguished by this algorithm, and it could effectively extracts the fault characteristic information. (3) Finally,300set of the extracted fault signal characteristics were used as training sample of the fault diagnosis model. The three layers structure of BP neural network and the five layers structure of fuzzy neural network were strutted. The nature type of fault happened on lines was judged by it. According to the defects of falling into local minimum point and so slow iterative training process in the BP neural network, Levenberg-Marquard training algorithm was used for improving the BP neural network in this paper. At the same time, in view of the difficulty of establishing the fuzzy membership functions and structuring fault diagnosis fuzzy rule base, the improved way of compensation factor was adopted in this paper. It makes the fuzzy neural network adjust adaptively model parameters to the optimum state in the process of learning. Finally,78pairs of trials sample data were used for testing in MATLAB platform. Simulation experiments prove that in the aspect of training time, error convergence speed and fault type identification accuracy improved fuzzy neural network based on compensation factor is superior to traditional BP neural network model. It is feasible that the fault diagnosis model is applied to high speed transient protection in the high voltage transmission lines. |
PDF Full Text Request