Use of artificial neural networks in fault location determination in energy systems

In power systems today, the continuity and reliability of power is of utmost importance for repair and maintenance, this has caused a growth in the use of different approaches to try and optimize this process. Some of the conventional techniques such as impedance-based and Traveling wave are widely adopted but face a myriad of challenges. These limitations have given rise to the growing interest in Artificial Neural Networks which can offer the potential for quick and robust fault identification and location through pattern recognition. This thesis presents the use of Artificial Neural Networks for the rapid and accurate identification and localization of a variety of faults in electrical transmission system. The symmetrical components specifically the positive and zero sequence voltage and currents were extracted from a 132 kV Transmission line model that was simulated in MATLAB Simulink, these time-domain values were used as input for the feedforward ANN. The result is a Trained Network that is designed to simultaneously classify the fault type and estimate its location along the 200 km transmission line, achieving high accuracy even after randomly varying fault resistance and fault inception angle values. Compared to conventional methods the ANN-based approach displayed lower sensitivity to noise and model parameters and generalization over many fault types. The testing confirmed that the average accuracy of the Neural network was 99.5% for fault identification and over 92% for fault location. The results shown, support the claim that ANN-based methods could be used alongside existing protection systems to ensure accurate fault identification and location.