Recursive identification and adaptive prediction of wastewater flows

Abstract

The forecasting of wastewater flowrates can help to reduce overflows and the operational costs of wastewater pumping stations and treatment plants. Such flows normally exhibit a diurnal flow pattern, but with large variations dependent on rainfall and groundwater infiltration. These factors have to be taken into account in the forecasting model. In this paper, a direct k-step adaptive predictor is used to forecast the wastewater flowrate. The time-varying dynamics of dry and wet weather sewer flow are modelled using the multi-input/single-output ARMAX model. The parameters are recursively estimated at each time step by the method of extended least squares; a forecast of the wastewater flow k-steps ahead is then made on the basis of the updated model. The model uses the measured sewer flow, the prefiltered area-averaged rainfall intensities and/or dimensionless flow pattern. By using this new method to represent the rainfall disturbances, reliable predictions up to 2 hours ahead for wet weather sewer flow can be made. This method is illustrated for a sewered catchment in Melbourne, Australia, which has a sewer system separate from the stormwater system.