Active power contingency ranking using a radial basis function network

This paper presents a new method for ranking active power contingencies in a large scale power system using a Radial basis function neural network (RBF) model. A list of contingencies whose probability of occurrence is high and critical for system security is prepared using previous experience. It is assumed that the effect of the contingency is limited to a localized sub network around the contingency. This sub network contains an area within two tiers from the contingency. Each sub network is modeled by a RBF network. The inputs to the RBF networks are the pre-contingency active power line flows within the sub network and an error term representing the effect of the complete power system. Each RBF network computes a scalar performance index as the output. These indices are then presented in a descending order. The networks can be trained off-line and training time is in the range of msec. Performance of the RBF network is evaluated using the IEEE 118 bus system. The advantage of this method is fast and accurate contingency screening promising future on-line use.