Artificial Neural Networks for Online Voltage Stability Assessment of Saudi Power Grid

This paper proposes and compares various techniques based on artificial neural networks for the on-line assessment of voltage stability of Saudi power grid in terms of a voltage collapse proximity index computed from the knowledge of the real and reactive power injections at various buses in the system. Various ANNs investigated include MLFFNN), RBFN and RNN-LSTM. The input-output training patterns required for the respective learning algorithms are generated by performing conventional Newton-Raphson load flow analysis (NRLF) of the power grid for various load conditions. The proposed models are assessed in terms of solution accuracy and computation time. The simulation results show that the ANN based methods can offer fast assessment of voltage stability as accurate as the NRLF method but without repeated iterative load flow calculations. Among the investigated models, the RBFN shows better performance in terms of both solution accuracy and computation time for on-line voltage stability assessment.