This paper presents a new method for power system state forecasting using arti cial neural networks (ANN). The state forecasting problem has been solved in two steps: the ltering step and the forecasting step in an open loop con guration. Because under normal operating conditions the power system behaves in a quasi-static manner, a simpli ed model of the dynamic behavior of the power system states is considered. Two diOE erent ANN models have been used for these two steps of power system state forecasting problem. For the ltering step, a functional link network (FLN), and for the forecasting step, a time delay neural network (TDNN) have been used to simulate the dynamic behavior of the power system states. The proposed method has been tested on two IEEE test systems, and a practical Indian system and results have been compared with an extended Kalman lter (EKF) based technique [Leite da Silva et al., 1983].