| This thesis summarizes the research efforts on localizing short-circuit faults in power distribution networks. Specifically, an experimental testbed is constructed to emulate realistic power distribution networks; various short-circuit faults are artificially created in the testbed; voltages/currents are measured over the testbed; and finally, fault localization is achieved by analyzing the measurement data.;Two types of power distribution networks are analyzed: serial/radial and grid/mesh networks. In the serial network, a scheme based on terminal voltage and current measurements is developed. Two voltage profiles, namely "forward voltage profile" and "backward voltage profile," are plotted and their intersection is estimated as the fault location. This algorithm is found effective to provide accurate results for different fault impedances. In the case of grid/mesh network, two novel fault localization algorithms, "signature pattern recognition" and "sparse sensing," are implemented. They are both validated by extensive experimental data. The accuracy and robustness of the two algorithms are compared. It is observed that, standard voltage measurements at a small number of pre-selected nodes suffice to localize the faults precisely; and hence, our fault localization scheme is of low cost. In addition, the algorithms are found to be highly efficient: typically, fault localization is completed in less than 50 ms. |
