Artificial neural network approach to transmission line relaying

This thesis deals with the design of an Artificial Neural Network (ANN) based relay for transmission line protection. A novel feedforward neural network that indicates whether a fault is within or outside the protection zone (fault indication) of a transmission line is presented. This method has been extended to locate the distance of the fault (fault location). The proposed scheme utilizes the frequency components of the voltages and currents to make a decision. -- The first part of the work employed frequency components of one cycle of post- fault data as the inputs to the ANN. The results obtained were promising, thus forming the basis to improve the speed of the relaying decision. This is achieved by using the frequency components of half cycle of pre-fault and half-cycle post-fault data as the inputs to the ANN. -- The neural network employed is small in size, fast and robust. Data obtained from the Electromagnetic Transients Program (EMTP) for single-line-to-ground faults and three-phase faults have been used for testing and the results are found to be accurate. The performance of the trained neural network is good and the proposed ANN has the potential for implementation in a digital relay for transmission line protection. The results of the proposed ANN methodology are found to be accurate under the conditions of different fault location, fault inception angle and fault resistance.