SECOND SYMPOSIUM ON EXPERT SYSTEMS APPLlCATlON

In this paper we are proposing the use of Artificial Neural Networks (ANN'S) as an operator aid in the dynamic security assessment of power systems. The basic role of ANN's in this study is to provide assessment of the system's security based on training examples from off-line analysis. The ANN'S in this study are trained either by 1) randomly distributed data which facilitate the utilization of both real measurements and offline simulations; or by 2) data obtained by the interval halving method for more accurate results around the boundaries. The training algorithm is based on the Back-Error Ropagation method which provides a high degree of accuracy.