Research On Diagnosis Method Of Series Arc Fault In Low-voltage Distribution System

With the rapid development of advanced technology of electrical engineering, variouskinds equipments are constantly emerging, and the electricity consumption of productionand household is gradually increasing. All electrical fires arising therefore are seriouslythreatening the personal safety, the reliability and the stability of electrical equipments. Inmany causes of electrical fires, arc fault can not be ignored. Arc fault is difficult to find inlow voltage distribution system, it is mainly caused by poor electrical connection, wireinsulation aging and the other reasons. Traditional protection circuit breakers can notidentify the arc fault, so it is very important to study arc fault to reduce the occurrence ofan electrical fire.In this thesis, the series arc faults in low voltage distribution systems are consideredas the research object. Firstly, the series arc fault mathematical model is established basedon the SIMULINK system of MATLAB7.1, and then the corresponding features of seriesarc fault are obtained.Secondly, the experimental data acquisition and analysis of the series arc fault. Basedon the established experiment platform, different typical loads are selected to test undersingle load and combined loads. And the oscilloscope is used to sample current under bothnormal condition and arc fault condition. Moreover, the sampled data are analyzed in thetime domain and frequency domain, and the characteristics are found to distinguish theseries arc fault current and the normal working current.Thirdly, the extraction and the dimension reduction of the feature of series arc faultcurrent signal. The obtained experimental data are analyzed by combining the wavelettransform and information entropy, the wavelet packet energy-entropy has been extractedto describe the energy distribution of arc fault current signal in different frequency bands.And then the principal components have been extracted from the feature by principalcomponents analysis(PCA) to realize the dimension reduction. On the basis of the above study, three layers BP neural network is established and theextracted fault feature are used to train and test the network. The simulation results showthat the above method has high accurate rate of fault diagnosis and can effectively identifythe series arc fault in low voltage alternating current distribution system.