Thyristor controlled series compensated transmission lines protection using artificial neural networks based symmetrical components analyzer

This dissertation reports the development, implementation and evaluation of fault detection, classification and location modules for protecting power transmission lines with thyristor controlled series compensation (TCSC). The developed modules make use of the main capabilities of Artificial Neural Networks (ANN) to achieve their highly accurate decision. In this study, multilayer perceptron neural networks are used because of their capability of recognizing and mapping any nonlinear relationship from a body of information representing the problem at hand. Prior knowledge on the fault characteristics and features are embedded in the training data sets and used as precursors for developing the ANN based protection scheme.;Two different novel techniques for addressing the problems of fault detection/direction identification, fault classification, fault location and loss of potential issue detection are considered. One of them makes use of a modular artificial neural network, where the main problem is divided into subtasks and each task is handled by its own individual ANN. The second technique is established on ANN based unified protection scheme with the adequate input information and optimal number of outputs. Both techniques make use of instantaneous magnitude and phase angle values of phasors, positive, negative and zero sequence components of the three-phase voltage and current signals at the relay location.;The first scheme is characterized by four parallel neural network modules: fault detection and direction module, fault classification module, fault location module and additional novel ANN based module for detecting loss of potential issue (LOP). The developed ANN modules utilize a quarter cycle sliding data window of the instantaneous magnitude and phase angle values of phasors, positive, negative and zero sequence components of the three-phase voltage and current signals. Once the fault is detected, the information about the fault type and location is available for post-fault processing. The developed modules have been tested to evaluate their performance under severe operating conditions like high fault resistance, close-in faults and voltage and current inversion conditions.;The developed modules have been trained and verified through computer simulation studies for a typical two machine power system model implemented in PSCAD. Simulation studies have also been considered for different operating conditions, including high fault resistance, fault location, degree of compensation and pre-fault power flow. This scheme has been found to be superior in terms of fast response and accuracy.;A different novel ANN based unified protection scheme has also been developed and examined for fault detection/direction identification, fault classification, fault location and loss of potential issue detection problems. The developed ANN based unified protection scheme is programmed on LabVIEWRTM and verified in real time. Furthermore, a graphic user interface (GUI) is developed to monitor and interact with the power and the implemented ANN.;An experimental lab scale model has been developed, which consists of two-machine transmission system with TCSC located on one end of the transmission line, for evaluating and testing the developed protection schemes. The performance of the developed ANN based unified protection scheme has been experimentally verified in real time using Lab-Volt 9062-00 data acquisition in addition to NI-PCI6221 interface to LabVIEWRTM software environment.. Experimental results confirm the superiority of the developed technique for detecting, classifying and locating faults on TCSC compensated power transmission lines through using only local measurements.