Time series models of the activated sludge process are very useful in design and real time operation of wastewater treatment systems which deal with variable influent flows and pollution loads. In contrast to common deterministic dynamic mathematical models which require knowledge of a large number of coefficients, the time series models can be developed from input and output monitoring data series. In order to avoid “black box” approaches, time series models can be made compatible and identical in principle, with their dynamic mass balance model equivalents. In fact, these two types of models may differ only in nomenclature. ARMA-Transfer Function models can be used for systems which are linear or can be linearized such as typical BOD or suspended solids influent-effluent relationships for which the type of model is known. For systems which are highly nonlinear, and/or the input-output model is unknown, neural network models can be used. Both ARMA-TF models and neural network models can be made self-learning, that is, the performance of the model can be periodically improved manually or in an automated mode as new information is collected by monitoring. Application examples are included.

This content is only available as a PDF.
You do not currently have access to this content.