PMU-Based Fault Classification Using Artificial Neural Network for Power Systems Considering Data mismatches

Fast advancement in communication and measurement techniques accelerates the development of wide area protection, based on the wide area measurement system. In this study, global information of power system will be introdu ced into the backup protection system. By analyzing and computing real-time phasor measurement unit (PMU) measurements, based on artificial neural network fault classifier (ANNFC), fault type and fault location is determined. Proposed method is implemented on the IEEE ٣٩ buses sample network. Best inputs of ANNFC are the active powers of each phase and the zero sequence currents measured at the two terminals of the transmission line. For each type of faults on a transmission line, ANNFC has one output. The outputs of ANNFC for the fault test patterns, not presented in the training stage, show the accuracy of the ANNFC. The trained ANNFC is trained again by inputs that have measurement errors and data mismatches. The ANNFC outputs are accurate, even if the data are distorted by CT saturation or by data mismatchescaused by delays in the WAP data collection system or measurement errors.