Neural-networks for predicting the operation of an under-frequency load shedding system

Dynamic security assessment is of special importance to island power systems. The CPU time required in order to apply conventional methods for those calculations does not allow real-time application. The fast calculation time is, therefore, an important advantage of artificial neural networks compared to other methods. This paper presents two neural network models that were designed to calculate the minimal frequency and the load shedding system operation during a forced outage of a generating unit. The minimal frequency and the extent of the load shedding are strong indications of the severity of the fault. Hence, it is a significant part of the dynamic security assessment procedure.