Parameter subset selection for the dynamic calibration of activated sludge models (ASMs): experience versus systems analysis.

In this work we address the issue of parameter subset selection within the scope of activated sludge model calibration. To this end, we evaluate two approaches: (i) systems analysis and (ii) experience-based approach. The evaluation has been carried out using a dynamic model (ASM2d) calibrated to describe nitrogen and phosphorus removal in the Haaren WWTP (The Netherlands). The parameter significance ranking shows that the temperature correction coefficients are among the most influential parameters on the model output. This outcome confronts the previous identifiability studies and the experience based approaches which excluded them from their analysis. Systems analysis reveals that parameter significance ranking and size of the identifiable parameter subset depend on the information content of data available for calibration. However, it suffers from heavy computational demand. In contrast, although the experience-based approach is computationally affordable, it is unable to take into account the information content issue and therefore can be either too optimistic (giving poorly identifiable sets) or pessimistic (small size of sets while much more can be estimated from the data). An appropriate combinations of both approaches is proposed which offers a realistic (doable) and sound approach for parameter subset selection in activated sludge modelling.