Artificial neural networks: a new approach for treating electrical engineering problems

Abstract

In this paper we present an approach to introduce electric power engineering students to new topics of research. Publications show how some classes of power systems and electromagnetic problems are solved using artificial neural networks. Some students at the American University of Beirut, willing to further their careers along academic lines or in research and development, are encouraged to take special projects to investigate and laydown the foundation for more serious research. Two such developments are reported in this paper. The first is a neural network-based automatic mesh generation method capable of developing meshes that preserve major properties such as Delauney triangulation and Dirichlet tessellation. In the second application, we demonstrated the possibility of creating a single artificial neural network to estimate the critical clearing time of faults in a realistic power system with differing load levels, fault locations, and network topologies.