Designing Experiments to Determine the Coefficients of Activated Sludge Models by Identification Algorithms

Abstract

A calibration technique comprised of an experimental procedure and a curve fitting mathematical method has been proposed in order to estimate the coefficients of the IAWPRC activated sludge model restricted to oxic reactions. A continuous activated sludge plant having two in-series biodegradation reactors is suggested for the experiment. This plant is operated by reducing, in steps, the first reactor volume. It was found that this experimental procedure enables scanning of a significant part of the kinetics contained in the model. The curve fitting mathematical method was an identification algorithm based on recursive non-linear Kaiman filter. In order to assess the validity of the calibration technique, simulated results of measurable variables, generated from presumed real coefficients, have been employed as real results by the identification algorithm. The effect of several erroneous initial estimates of the coefficients, which are required to start the identification, has been investigated. It was found that the proposed calibration technique gave rise to acceptable identifications of the real coefficients independent of the initial estimate used. The convergence rate for every coefficient, however, depends on the initial estimate. Those rates were properly related to the differences observed between the real results and the predictions obtained with the initial estimates for the measurable variables.