Short-Term Forecasting for Urban Water Consumption

Abstract

An approach is presented for short-term (i.e., daily and monthly) forecasting of municipal water use that utilizes a deterministic smoothing algorithm to predict monthly water use. The smoothing algorithm considers level, trend, and seasonality components of the time series. Daily deviations from the monthly average are then forecasted for up to six days using autocorrelation and weather dependence. While providing accurate operational forecasts, the approach required about six years of daily data to develop and validate the models. The approach is applied and evaluated for a number of municipalities near Tampa, Fla. Results show that the approach provides accurate daily forecasts as measured using a validation period of about three years.