Selection of Suitable Feedforward Neural Network Based Power System Stabilizer for Robust Excitation Control System

A reliable and economical power system is the key requirement for the development of any country's power sector. Electrical power system is quite complex due to its non-linear behaviour and changing load conditions. In an excitation system, the Automatic voltage regulator (AVR) and load frequency controller (LFC) are connected to maintain the terminal voltage and load frequency. The high gain of AVR produces negative damping on rotor of generator. This behavior of AVR can be easily controlled by the introduction of power system stabilizer (PSS). The major function of PSS is to inject supplementary stabilizing signals to minimize the effect produced by AVR. For this purpose, various controllers as power system stabilizers have been proposed by researchers to tune their parameters. In this research work, three types of controllers i.e. probabilistic feedforwad neural network (PFFNN), multi-layer perceptron feedforwad neural network (MLPFFNN) and radial basis function feedforwad neural network (RBFFN) are designed in Matlab / Simulink to tune the parameters of PSS. These controllers are implemented as single machine connected with infinite bus (SMIB). The simulation results for terminal voltage (vt) and load frequency (f) of these PSSs are compared. The simulation results for RBFFN based power system stabilizer show the promising improvement in stability of synchronous generator than other types of FFNN controllers.