Skip to Main Content
Distribution utilities have been employing efforts for the automation of distribution networks in order to restore power faster, to result in better continuity levels and increasing consumer satisfaction. Despite digital relays provides an indication of the possible fault location, based on the circuit impedance, it is not accurate for distribution systems because there are radial networks with lots of branches, different wire gauges, etc. This work proposes a combined treatment of information from the monitored protection devices and computer analysis. A self-organizing map (SOM) is trained with simulated variables (short-circuit current, voltage, demand, etc.) to represent a segment/block of the distribution system in each neuron. So when a set of input data from monitored equipment is presented to the SOM it indicates the faulted section of the network. This tool is being tested in a pilot area of the Nova Palma Energia, an electric utility from Brazil's southern state.