Dynamic mathematical modelling of full-scale wastewater treatment plants requires an optimal level of detail to be accurate, effective, but still manageable. Single process models tend to oversimplify the complexities of a large-scale plant and result in erroneous calibrated parameter values or limited predictive power for the model – on the other hand, modelling of every process and process unit is usually impractical, since it cannot be supported by reliable plant data, and requires prohibitive sampling costs and effort.
Level of aggregation, settler dimensionality and reactivity, and plug-flow hydraulics were investigated with the help of a sophisticated dynamic modelling package. The results show that there is no general rule or global ‘optimal level’ of modelling – the required modelling detail is a function of influent flow and loading levels, and processes to be simulated (BOD removal, nitrification-denitrification, biological phosphorus removal, settling). In the case of real-life plants, supportable modelling level is often constrained by data availability and reliability.