A HYBRID DECISION TREE-ARTIFICIAL NEURAL NETWORK TECHNIQUE FOR REAL-TIME TRANSIENT STABILITY PREDICTION

Abstract

Fast transient stability assessment is one of the most important needs in control centers during power system operation. Transient stability is concerned with system behavior following any large disturbance in system operating conditions. Short-circuits, in particular, can give rise to large electromechanical oscillations among generator rotors, which may lead to system collapse. In this thesis, two artificial intelligence based techniques have been developed for predicting transient stability in real time operation. The first approach is based on the decision tree (DT). It is suggested to extend the application of this approach to the field of feature extraction of main system parameters. The second approach is based on the artificial neural network (ANN). Through utilizing a proposed neural network whose inputs are specified from the DT approach, a simple linear expression has been suggested for determining the critical fault clearing time (CCT) and consequently predicting the system stability. The method introduces different observations for studying the sensitivities of the . CCT to system parameters. The proposed method has been demonstrated on two power system models: (i) the 4-machine, 6 busbars system model, and (ii) the 7-machine, 10-busbars CIGRE system model. The computer simulation results confirm the validity of the developed approach.