Ann-based faster indexing with training-error compensation for mw security assessment of power system

Abstract

A simplified faster method for contingency analysis and screening is proposed for both on-line and off-line applications in electrical power networks. The suggested method uses the transmission line power flows to develop artificial neural network models which are then used for monitoring the transmission lines in real-time and provide binary output that signifies the state of the network. The outputs of the neural net are then used to calculate an index to determine the state of the whole power network grid. An additional term for the misclassification data has also been included to compensate for the errors in the classification of states, while using the neural networks. The proposed approach was applied to a test bus system and a state-owned utility. The results testify that the proposed method will provide faster results in shorter response time. The whole process of ANN-based security assessment is completed within 8 min.