Using regression trees for raw effluents quality prediction

Nowadays is quite relevant to any Waste Water Treatment Plant (WWTP) manager to know how and in what conditions its facilities and respective treatment units work, especially the ones related to the application of biological treatments. Among the different treatment stages that occurs usually in a WWTP to deal with domestic wastewater, the stage where organic load is removed (secondary treat- ment) assume an indispensable function ensuring a minimum quality for the treated wastewater. It is the most sensitive unit in a WWTP, reacting easily to load vari- ations, flow rate or residual concentrations of harmful elements. Factors like these ones may lead to the death of microorganisms responsible for the treatment or to significant changes in the kinetics of organic matter degradation, which affects the quality of the final effluent. All this assumes a major role in the removal of the pollutant load of wastewaters. In this work we present a study carry out on a spe- cific WWTP, located in the Northern of Portugal, in order to monitor and control its secondary treatment units, in particular, predicting the impact of changes in its raw effluents to provide operational elements that may contribute to ensure an ade- quate quality for final effluents. We used several analysis parameters – pH, Chemical Oxygen Demand (COD), Biochemical Oxygen Demand after 5 days (BOD5), and others – to support the development of a data mining model, using regression trees, especially oriented to make such prediction. This predictive application will give us the possibility to determine such changes and analyse the overall operation of the WWTP for different periods of time, analysing their impact in the operation of the WWTP