Summary: This paper presents the application of Artificial Neural Network (ANN) for prediction of system maximum demand. The average hourly load on each feeder for four days in a week for one month past historical data is chosen as inputs to the ANN. The system maximum demands are chosen as outputs. The ANN was trained using supervised back propagation algorithm. Due to the high capability of parallel information processing of ANNs, the proposed approach is fast and accurate. The outcomes of this approach help the future expansion of existing distribution substation. The proposed algorithm is applied to a 33/11kv Muss Fort substation for system demand prediction. The difference between actual and estimated system maximum demands are good in terms of accuracy.