Research On AC Series Arc Fault Recognition Method Under Complicated Harmonic Conditions

Arc fault often generates in power supply system due to aging of cable insulation and loosing of electrical contact point caused by mechanical vibration or electro-dynamic repulsion force.In severe cases,it will produce a series arc fault and even lead to electrical fire accidents.Some industrial sites,such as coal mines,have lots of harmonic contents in power supply system.The non-linear loads used in the circuit such as soft starters,frequency converters,switching power supplies,rectifiers and so on will also increase the harmonic contents.When a series arc fault generates,these harmonic contents and load noise will reduce the recognition accuracy.There is no detection method for series arc fault in the existing literature under harmonic interferences.Therefore,it is necessary to study new methods to accurately detect arc fault.In view of the serious interference of power supply harmonics and load noise in the industrial site,the identification of series arc faults has been studied.First,series arc fault experiments under complicated harmonic conditions were carried out.Secondly,the collected experimental data was analyzed,and two arc fault feature extraction methods were proposed.The first method used kernel principal component analysis to process the power supply voltage signal,fault phase current signal,and normal phase current signal,and the kurtosis and skewness of the fifth and sixth principal elements were used as the fault feature vectors.The second method used variational modal decomposition to decompose the fault phase current signal,and the permutation entropy and energy entropy of the low and intermediate frequency decomposition components were used as the fault feature vectors.Finally,an arc fault recognition model based on support vector machine was established,and the penalty parameter C and kernel function parameter g of the support vector machine were optimized by the firefly algorithm to realize the recognition of arc fault.The influence of different training sample numbers,data sampling frequency,data processing period and fault characteristics on arc fault recognition rate and recognition time were studied,and the universality of the two fault feature extraction methods was verified by using different fault feature extraction methods,harmonic experimental power sources,load types and arc generation methods.There are 30 figures,26 tables and 70 references in this paper.