Design of a Convolutional Neural Network and a Modified Genetic Algorithm for Power Grid Disturbance Classification

This paper discusses the use of genetic algorithms and convolutional neural networks for the classification of power flow disturbances in a power grid. The methodology relies on identifying singular and multiple disturbances of bus voltages of experimental power grid networks. The bus voltages have been recorded for an hour. First, any significant change on the voltage values could be identified using a genetic algorithm modified for detecting changes in millivolt ranges. Then, a convolutional neural network is designed to predict the overall changes in the network and classify grid level disturbances. The combined approach can identify the presence of singular or multiple disturbances in a power system and can also identify the areas that could be affected by various types of disturbances.