Performance Assessment of Wastewater Treatment Plant with Machine Learning Tools

Abstract

Wastewater treatment plants (WWTPs) with activated sludge are complex and dynamic systems whose management can be improved by using different modelling and prediction approaches to their work. A machine learning tool for the development of model trees was used in this paper in order to develop a model for chemical oxygen demand (COD) in the wastewater effluent from the WWTP with activated sludge. The data measured in both the influent and the effluent of WWTP were used for modelling. For the COD model the machine learning tool Weka and the algorithm for the development of model trees M5P were used. The obtained model has a high descriptive power and correlation coefficient and thus can be used for modelling purposes. Also, the purpose of this paper is to show the benefits of using machine learning tools for developing WWTP models.