Development of a fault location device in rural distribution networks

The quality of electrical energy depends on its continuity and the reliability of the system. In this context, the rapid detection of the fault and fault location in networks become crucial. In the literature, various studies have been carried out using different methods for the detection of fault location in the transmission network using digital measuring devices. However, the branched structure of the distribution network, being affected by environmental factors (weather conditions, vegetation, etc.) and the difficulty in monitoring have prevented the studies in transmission network from yielding successful results in the distribution network. In this direction, in this thesis, a Random Forest-based fault location system that can be applied to multi-branched rural distribution networks has been proposed and 2 prototype device applications working with this algorithm have been developed. In the proposed algorithm, Discrete Wavelet transform and Fast Walsh Hadamard transform are applied to the 3-phase current-voltage data taken only from the sending end of the feeder to obtain features. The efficiency of the algorithm has been proven with data from Akyazı-Düzköy feeder in Trabzon modeled in the Digsilent PowerFactory program using real line parameters and IEEE-34 test bench. The device applications are real time and offline applications. In the real time application, the algorithm embedded in the F28379D with the help of MATLAB Coder reads the fault data in real time and the fault location result is obtained. In the offline application, results are obtained with the fault location detection system embedded in the Lattepanda development board. The studies were compared with other studies in the literature and their effectiveness was observed.