Detection of Series AC Arc Fault Based on Continuous Wavelet Transform and Artificial Neural Network Under Voltage Variation Disturbances

Series arc fault in electrical power system can cause electrical fires. In distribution system, power quality disturbances include voltage variation disturbances often occurs. Voltage variation disturbances can lead to non-stationary waveform, which affect accuracy arc fault detection. To overcome this problem, this paper proposes combination methods of continuous wavelet transform and artificial neural network for detection. Continuous wavelet transform is superior to analyze of non-stationary waveform from series arc fault under voltage variation disturbances. Continuous wavelet transform recognizes series arc fault under voltage variation disturbances by transform of signal to time-frequency domain. Artificial neural network using type of feed forward neural network with Levenberg Marquardt Algorithm for series arc fault identification by data obtained from continuous wavelet transform. Data is trained and tested by artificial neural network. Several waveform model of series arc fault under voltage variation disturbances are selected to tested include normal system, series arc fault under normal condition, series arc fault under voltage sag condition and series arc fault under voltage swell condition. The result show that the algorithm of continuous wavelet transform and artificial neural network have good accuracy for series arc fault detection under normal system, series arc fault, voltage sag condition and voltage swell condition with the accuracy of 99.69 %, 99.2%, 99.97% and 99.9%, respectively.