Steady-state security analysis using artificial neural network

A new approach based on artificial neural network (ANN) is presented for single line contingency evaluation. The proposed method is able to yield system states consequent upon a contingency. Given the bus voltage magnitudes and angles, system conditions can be completely determined. Two suitable radial basis function (RBF) neural network architectures are proposed, one each for voltage magnitude and angle prediction following a contingency. These two networks are trained using the results obtained from off-line load flow calculations. Further, a line flow calculation interface is proposed to calculate active and reactive power flows. The proposed RBF network based approach is most suitable for studying all foreseeable contingencies at planning stage. The proposed scheme has been successfully implemented over sample 6-bus, IEEE 14-bus and IEEE 57-bus systems. The accuracy of the proposed networks is compared with a standard A C load flow algorithm. To demonstrate capability of the proposed RBF neural networks, the results are compared with feed forward error back propagation (FFBP) neural networks.