Management of watewater treatment plants: a statistical approach

Abstract

The progressive deterioration of water resources and the large amount of polluted water generated in modern societies give Wastewater Treatment (WWT) processes a fundamental importance in water prevention and control. Inside a biological Wastewater Treatment Plant (WWTP), the activated sludge process is the most commonly used technology to remove organic pollutants from wastewater. This most cost-effective technology is very flexible and can be adapted to different kinds of wastewater. Therefore, it is very important to understand and to model the management processes involved that can lead to benefits for the overall WWTP, in particular in cost-effectiveness. In this work the discussion focuses on the dynamic monitoring procedure based on the statistical modeling approach, in order to quantify and to characterize significant statistical patterns of interaction between wastewater flows, hydro-meteorological variables (such as rainfall), and physicochemical variables. Activated sludge processes (ASPs) within wastewater treatment plants are commonly operated conservatively, aiming to maintain the healthy operation of the plant in the presence of varying plant loads. While the primary aim of wastewater treatment systems is to provide plant effluent of a suitable quality, this must be achieved in the presence of both physicochemical and financial restrictions on plant operation. Operating costs, in particular energy costs, associated with, for example, process aeration, are driven up by the inclusion of ‘safety margins’. A statistical exploratory analysis, calibration models and linear models were performed in order to obtain an accurate prediction and forecast of the relevant predictors (wastewater effluent variables) in the flows’ behavior and which have the greatest impact on cost reduction. The statistical modeling procedure was applied to a set of nine Wastewater Treatment Plants located in the Northwest region of Portugal (five in rural regions and four in urban ones), and the dataset consists of monthly measurements during a period of two years, from January 2015 to December 2016. By accommodating the well-known seasonal regimes of dry and wet seasons, the statistical results will provide a better representation of the plants’ real situation in order to design an efficient management process.