Parallel radial basis function neural network based fast voltage estimation for contingency analysis

Abstract

Estimation of bus voltage magnitudes under normal and contingency cases are very important for secure operation of power system. A novel parallel radial basis function neural network (PRBFN) is suggested to predict bus voltage magnitudes in an efficient manner. Proposed parallel radial basis function neural network is a multistage network, in which stages operate in parallel rather than in series during testing. The nonlinear mapping capability of radial basis function has been exploited along with forward-backward training. The input features for PRBFN are selected using entropy concept to reduce the dimension of inputs as well as size of the neural network. The proposed method is used to predict bus voltage magnitudes at different loading as well as generating conditions and for single line outages of IEEE-14 bus system. A single PRBFN has been trained to predict voltage at all the PQ buses for the base case as well as for the line outages.