Influent load prediction using low order adaptive modeling

Paper in proceeding, 2005

This paper presents the use of models for predicting influent flow to the Rya wastewater treatment plant. Momentaneous influent flow is calculated using physical modelling and is then used to identify prediction models using ARX filters. Dry weather and stormwater runoff flows are modelled separately with the dry weather model predicting the flow as the average of past dry weather days and the stormwater model also using past flow data but with rainfall predictions as an extra input signal. The results indicate that accurate 24 hour predictions can be made during dry weather. With the addition of a stormwater runoff model with a shorter prediction horizon the combined models is believed to be useful for improving pump control performance for all weather conditions.