Serial Arc Fault Detection Through Wavelet Transform and Support Vector Machine

Electric arcs caused by damages in wiring and electrical connections of an installation may result in fire. Conventional protection devices, such as circuit breakers, fuses and residual-current circuit breakers (RCCBs) are not capable of detecting this disturbance. New equipment, such as the Arc Fault Devices Interrupter (AFDI) have appeared in the last two decades with the purpose of meeting such demand. However, their application and efficiency are still restricted due, mainly, to the masking deriving from disturbances generated by some types of charges present in the power grid. Making such detection more robust and accurate is not a trivial task and, therefore, new methodologies have been developed in recent years. Considering such context, the purpose of this work is to assess the disturbances caused by serial arc faults through experimental tests and use of Wavelet Transform (WT) on the detection of the phenomenon. Based on the results, it was possible to develop a sorting model that produced very satisfactory outcomes. The model was developed through concepts of the support vector machine (SVM) and presented average accuracy around 90% even in critical detection situations.