Detection of serial arc fault on low-voltage indoor power lines by using radial basis function neural network

• On electric power lines, characteristics of current waveforms can be analyzed by using DWT and signal energy.
• Test results show that the performance of DRBFNN is better than DSE and HFDWT to detect serial arc faults.
• A low voltage experimental circuit is designed and fabricated.

Dangerous serial electric arc faults on low voltage power lines must be detected before fire hazards occur. The detection technology is requested to have high accuracy. However, the characteristics of line current waveform during serial arc faults are complicated. This paper uses the approach of a radial basis function neural network (DRBFNN) to identify the occurrence of serial arc faults. At first, the discrete wavelet transform (DWT) is employed to obtain the time–frequency domain characteristics of line current waveforms to reflect the serial arc fault patterns. Then some measured data are used to train the DRBFNN. Finally, this study compares the detection results under different loading conditions and operation conditions. It also compares the detection results with other two methods, detection of sub-spectrum energy (DSE) and high frequency detection by wavelet transform (HFDWT). It can be observed that DRBFNN has better ability than DSE and HFDWT to detect serial arc faults.