A model for short-term load prediction (24 hours) has been developed. It is implemented as a part of an interactive program system for load prediction within the different areas of the Norwegian Power Pool. The model consists of two parts, describing the slow, seasonal variations (normal conditions), and short-term deviations from normal conditions, respectively. Kalman filtering techniques are used for updating the states of the models, and the parameters are estimated with a maximum likelihood method. The model has been tested with load data from various areas in Norway, and the system has been in on-line use at the Norwegian Power Pool since October 1979. Better control of the power system has been obtained through improved prediction and production planning.