Wastewater treatment plant performance prediction with support vector machines

Abstract

Wastewater treatment plants are essential infrastructures to maintain the environmental balance of the regions where they were installed. The dynamic and complex wastewater treatment procedure must be handled efficiently to ensure good quality effluents. This paper presents a research and development work implemented to predict the performance of a wastewater treatment plant located in the northern Portugal, serving a population of about 45,000 inhabitants. The data we used were recorded based on the daily averaged values of the measured parameters during the period of one year. The predictive models were developed supported by two implementations of Support Vector Machines methods for regression, due to the presence of two lines of treatment in the selected case of study, using two of the most relevant output parameters of a wastewater treatment plant: the biochemical oxygen demand and the total suspended solids. We describe here the wastewater treatment plant we studied as well the data sets used in the mining processes, analyzing and comparing the regression models for both predictive parameters that were selected.