Critical Clearing Time Prediction for Power Transmission Using an Adaptive Neuro-Fuzzy Inference System

An adaptive neuro-fuzzy inference system (ANFIS) is a hybrid algorithm composed of fuzzy logic and an artificial neural network. It takes advantage of fuzzy logic and artificial neural networks to solve complex problems. For power transmission, several dynamic parameters are ignored for conventional transient stability assessment due to the complexity of the equations, and due to the long computational time required. Certainly, it is very difficult to do the real-time assessment of large power systems by considering dynamic impacts in detail. In this paper, an approach is required to increase the accuracy of results. Herein, a method, namely ANFIS, was found to overcome the limitations. All significant effects of dynamics can be taken into account; not only the machine model but also the turbine governor model, automatic voltage regulator (AVR) model, and load characteristic model are carefully considered. In addition, the model used for each generator unit is varied to achieve real conditions. The ANFIS output is the critical clearing time (CCT). CCT values are very important to be correctly predicted for the stability of power systems after clearing the faults. When the faults are cleared by opening the circuit breakers within the CCT values, the power systems are still stable. If the faults are cleared after the CCT values, the systems become unstable. The modified IEEE 9-bus and IEEE 39-bus systems are applied in the implementation of this study. The locations of faults and the levels of loads (except the power system topology changes) are varied for each simulation. The results from the ANFIS application indicate that ANFIS (considering the dynamic effects of the machine models, AVR system, turbine governors, and load characteristics) can predict CCT values with high accuracy. Moreover, when ANFIS results are compared to the artificial neural network (ANN) results, which are generally used, it can be seen that ANFIS results are better and take a lower time for training and testing processes than ANN. ANFIS can be adapted, improved, and implemented in real practice.