Transient stability assessment for single-machine power systems using neural networks

The use of back-propagation neural networks for fast and efficient determination of the stable and unstable modes of operation of power systems is proposed. The transient response quantities selected as input features for training the neural network and for carrying out the assessment using the neural network are quantities which can either be calculated easily or be measured. The training process for the neural network requires very low computing time. When applied to a single-generator system, the neural network can assess the transient stability of the system accurately for a symmetrical fault on any point along the external circuit and for the range of normal prefault loading conditions. The authors also developed a back-propagation neural network for the estimation of critical fault clearing time. The authors describe the neural network configurations adopted and their use in stability assessment. Results obtained by applying the neural networks to a single-machine system are presented