The economical operation of the Grande Dixence S.A. pumping plants involves an optimal utilization of the electrical energy for pumping the discharges of five glacierized basins. For each basin, a short-range forecasting model (3 d) has been implemented in order to determine, at any time, the future evolution of the hourly discharge curve. The forecasting model is based on multiple regressions and involves following variables: discharge, air temperature, global radiation, precipitation and the altitude of the snow-cover limit. The selection of the variables not only depends on the physical laws of the melting of glaciers but also on the possibilities of measuring them accurately. The melt-water run-off is not a stationary process, but this difficulty can be by-passed by introducing the altitude of the snow-cover limit and by subdividing the melt season into intervals. The model must automatically choose the equations corresponding to the actual interval. The quality of the forecasts essentially depends on the weather forecast; some difficulties, such as the precipitation in the forecasted period, have not yet been resolved. The utilization of discharge forecasts in an optimizing model for the operation of a pumping plant with a compensation basin is compulsory and, further, enables the maximum usage of low-price electrical energy, limits water losses, and guarantees operational security.