Two approaches to wastewater treatment plant modelling

Authors: Boris KOMPARE, Meta LEVSTEK, Nataša ATANASOVA
Citation: Acta hydrotechnica, vol. 24, no. 40, pp. 45-64, 2006.
Abstract: The main goal of any model is to serve the purpose for which it was built. Models of wastewater treatment plants (WWTP) are built to better explain how processes work in the system, to reveal some new knowledge, or, simply, to predict their future behaviour. Therefore it should be strongly encouraged to use a variety of approaches and methods to find a model that will best satisfy our needs. In this paper we are using two basic approaches to model a pilot WWTP within the central WWTP Domžale-Kamnik (Slovenia). The first includes conceptual modelling, where models are constructed based on theoretical knowledge about the domain. For this approach we apply the well-known conceptual model ASM1 to simulate the processes in the WWTP. The second approach is concerned with learning from data. In this case we use a machine learning method for inducing regression trees. The on-line measurements of nitrogen compounds enabled the ML tool to build a model for prediction of ammonia outflow concentration. The models obtained are discussed in the light of the field expert's expectations and needs.
Keywords: wastewater treatment plant, activated sludge, modelling, conceptual models, machine learning, regression trees
Full text: a40bk.pdf