Multi-Class Classification of Power Network States Using Multi-Dimensional Neural Network

During power system operation, for the optimized operation and control of power networks, its operation is usually segregated into several different operational states. This works in both real-time and offline monitoring scenarios. These operational states are based on the variation of the levels of parameters like voltage magnitude, angles, system operational frequency, switching states, active and reactive power flow on the transmission lines. Initially, Tomas Liacco proposed the categorization of power system operation into 3 functional categories, which was later, further sustained by Lester Fink and Kjell Carlsen into 5 different functional categories. These categories classify the power system operation into 5 different states subject to the severity of line outages and overloading capacities. This categorization of power network states using original mathematical approaches can be very cumbersome and time-consuming. In this work, a simplified method for the segregation of power network states has been suggested based on the application of a multi-dimensional artificial neural network, that has been employed for the prompt multi-class classification of power network states based on Fink and Carlsen’s approach. The whole setup for this study, including the power network and the artificial neural network models, was developed on the Simulink environment of MATLAB (R2021a). The system was simulated on the RT Lab OP-5600 simulator in real-time. The results of this study prove that the application of the multi-dimensional neural networks will provide results in under 1s, which is faster than the operating time of many types of circuit breakers and relays, hence proving the practicality of the method.