The paper discusses the implementation of a fuzzy-logic approach to provide a structural framework for the representation, manipulation and utilization of data and information concerning the prediction of power commitments. A neural network would then be implemented to accommodate and manipulate the large amount of sensor data involved. A training facility could allow the system to replace the requirement for skilled dispatchers in scheduling the generators. An algorithm has been implemented and trained to predict the total power demand on an hourly basis. The parameters taken into consideration cover environmental and weather-related conditions. Prediction of the power demand at each geographical load point, and hence the country-wide demand, has been tested in Jordan. Results concerning the daily prediction have been obtained. It is found to be very promising, especially in that the prediction is evaluated in a fuzzy environment.