Experimental Study On AC Series Fault Arc Location Of Low Voltage Distribution System

In the low voltage distribution lines,the series fault arc often occurs due to the aging of the lines,environmental pollution or bad contact.After the occurrence of series fault arc,it will not only affect the power supply quality and bring serious impact to people's production and life,but also even cause electrical fire and seriously threaten people's life and property safety.At present,the researches on series fault arc by domestic and foreign scholars are mainly focused on the arc characteristics and arc identification,but there are few researches on the location of low-voltage series fault arc.Therefore,the study of AC series fault arc location method for low-voltage distribution system has important theory significance and application prospect for improving the maintenance efficiency of low-voltage distribution fault lines.Using the self-developed series fault arc positioning experiment platform,the AC series fault arc positioning experiment was carried out,and the voltage and current at the power supply and the load end of the experimental loop were collected.The collected voltage and current data were analyzed in the time domain and frequency domain respectively.The amplitude,mean value,standard deviation and waveform factor of the voltage and current signals at the power supply end and the load end were extracted in the time domain analysis.It was found that the amplitude of the voltage at the load end and the location of the fault point had obvious regularity.The characteristics of the odd harmonic components and the total harmonic distortion rate of the voltage,current signal at the load end and the current signal at the supply end after noise reduction by the wavelet threshold are extracted by frequency domain analysis,found that the load voltage of 3 times,5 times of the harmonic component and the power of the current total harmonic distortion rate and the total harmonic distortion of load terminal voltage and current rate and fault point location presents obvious regularity.The feature quantities with significant regularity with the location of the fault point are screened out as the feature vectors of the fault location.Finally,the BP neural network optimized by levenberg-marquardt algorithm is selected as a tool for fault arc location.According to the data characteristics,the network training is divided into two training methods: single-group positioning and multi-group positioning for comparative analysis.At the same time,the network parameters are determined by the method of gradual experimental optimization.After the network training is completed,the test sample data are input,and the positioning results show that the proposed fault positioning method based on LM-BP neural network can effectively locate the series fault arc: in single-group positioning,the average positioning error is about 0.4m,and in multi-group positioning,the average positioning error is about 5m.