Parameter tuning of the conventional power system stabilizer by artificial neural network

This paper presents parameter tuning of conventional power system stabilizer (CPSS) by artificial neural network (ANN). The ANN in the paper is radial basis function network (RBFN), whose parameters are chosen by adaptive orthogonal least squares (adaptive OLS) algorithm, to compensate error of linear model of power system where a fixed-parameter CPSS is analyzed. The adaptive OLS algorithm is developed from the orthogonal least squares (OLS) algorithm to reduce the neural network size more efficiently. When the system condition is changed, this makes the fixed-parameter CPSS less efficient than a varied-parameter CPSS by ANN. Moreover, the adjustment of damping coefficient using the gradient descent method improves the oscillation damping.