ABSTRACT
Transmission lines suffer from unexpected failures due to various random causes which can lead to instability. The functions of protective systems are to detect, classify faults, locate and sent trip signal to circuit breaker for isolation. The main objective of this task is to study the available techniques and algorithms to develop improved relaying algorithm based on adaptive neuro-fuzzy inference system (ANFIS) which could have 100% accuracy and operate with minimum delay. The training , testing and validation data samples to be used by the ANFIS models were generated using sequence current components and line voltages under normal and fault conditions at various locations on a 400kv,50Hz, 100km transmission line. Simulations were performed using EMTDC/PSCAD, on a sample three-phase power system. The lines current were first processed using FFT algorithm and then the sequence components were derived from the same fundamental frequency. Various fault scenarios are considered in this work. The ANFIS’s were trained and tested using the various sets of data in which six inputs were used in different combinations. The inputs to ANFIS’s are rms line voltages and ratio of sequence currents was used. Different membership functions were used with different number of epochs and ‘gbell’ membership function was found to be the best in performance for both training and testing. Data was extended and the ANFIS was tested with ‘gbell’ membership function. The result obtained show 100% accuracy with lesser number of epochs than needed with ANN.
