Abstract

Optimization and control of waste watertreatment plants (WWTP) is an ongoing effort to make theprocess more efficient and cost-effective. As found in literature,data mining models such as neural networks have been appliedto simulate and model various aspects of the plant such asperformance, quality parameters and process parameters. Inthis paper, we introduce bagging model, an ensemble datamining model, to predict the performance of the WWTP.Ensemble models have been shown to stabilize the baseclassifier used and avoid overfitting the data. Bagging was usedto predict the performance of individual units (primary settlerand secondary settler) and the global plant performance. Thepredicted performance of individual units was also used asinputs to predict the global performance thereby enabling goodprocess control via predictive data models. Upon application tothe WWTP dataset, it was found that bagging models performat par or even better than ANN or SVM for the prediction andhence are suitable models that can be implemented for processcontrol of the water treatment plants.