A Comparative Study in the Design of an ANN Based Controller

The power system stabilizer is based on artificial neural networks and is designed to cope with changing operational circumstances in real-world applications. The parameters of a conventional or supplementary power system stabilizer are generated with changing operating circumstances using a Multi-layer feed-forward with back propagation (MLP) Artificial neural network. As a result, the power system stabilizer's utility becomes more stable. Many researchers included this artificial neural network in their designs; however, information on network creation functions, hidden layer size, non-linear (active) functions employed in output and hidden layers, and Back Propagation (BP) training functions were not studied in detail. In this paper a study is carried out on these aspects to explore the effect of these aspects on the functionality of power system stabilizers.