To provide a water-demand prediction system enabling a prediction correctly grasping seasonal variation even with a reduced load on the system.

A water-demand prediction system includes: an analysis part 21 for breaking down a distributed water volume X(t) into a clock-time component s(t) repeating in week cycles and a tendency component m(t) representing seasonal variation; and a distributed water actual value accumulation part 1 for accumulating, as time-series data, a distributed water actual value X(t) in an analysis target period. Further, the analysis part 21 performs processing for acquiring the tendency component by inputting the distributed water actual volume X(t) and removing the clock-time component, and processing for acquiring the regression coefficient of the tendency component by expressing the tendency component using a seasonal variation-approximating expression with time t as a variable, and the system further comprises a distributing water calculation part 41 for obtaining, from the analysis part 21, a prediction model expression including a clock-time component and a tendency component to calculate a distributing water prediction value X^(t). Furthermore, the distributing water calculation part repeatedly calculates distributing water prediction values using the same seasonal variation-approximating expression as a tendency component during a model-updating period designated in excess of the cycle T of the clock-time component.