Novel Formulation using Artificial Neural Networks for Fault Diagnosis in Electric Power Systems - A Modular Approach

This work proposes a new method for fault diagnosis in electric power transportation systems based on neural modules. With this method, the diagnosis is performed by assigning a generic neural module for each type of element conforming a transportation system, whether it be a transportation line, bar or transformer. A total of three generic neural modules are designed, one for each type of element. These neural modules are placed in repeating groups in accordance with the element to be diagnosed, taking into consideration its circuit breakers and relays, both internal and backup. For the diagnosis of a transportation line, this method is further reinforced by taking into consideration the corresponding waveforms of fault voltages and currents as well as the frequency spectrums of these waveforms, through a neural structure, in order to verify if the line had in fact been subjected to a fault, and at the same time to determine which type of fault (LT, LLT, LL, LLL, LLLT). The most important and innovative aspect of this method is that only three neural modules will be used, one for each type of element, and these can be employed for a diagnosis as one function, the instant any change of status is detected in the internal and/or backup relays relating to the element subjected to diagnosis.