Artificial neural network based fault location for overhead distribution lines

In this thesis, a fault location and classification technique based on artificial neural network (ANN) for 34.5 kV medium voltages (MV) overhead power distribution line is presented. Feedforward networks have been employed along with back-propagation algorithm for fault location and classification process. The separate ANNs are used for classifying and locating the shunt faults on the distribution system. The technique utilizes voltage and current pre- and post-fault data at one line end only. These values are stored as waveform samples by a digital fault recorder (DFR) in the substations. Spontaneous three phase voltages and currents acquired from the fault locator at different frequencies are employed for testing and training the ANNs. The power distribution line is simulated using MATLABR2009b/Simulink. In the development of the DFR, the important aspects of the practical fault recorders such as voltage transformer (VT) and current transformer (CT) responses, analogue interface effects and quantization errors are taken into account. This is made to ensure that the performance obtained is considerably close to a real-life situation. The ANNs which have been developed for different fault types are implemented for 34.5 kV medium voltage overhead power distribution line between Kütahya and Enne circuit breaker measurement cabin. The accuracy of the technique is evaluated for the fault type and the location, the presence of remote-end in-feed and the fault inception angle and the corresponding results are given as tables.