Performance comparison of different optimization methods for the wastewater treatment plant model based on the Activated Sludge Model No. 3 and design of the plant control

Abstract

Abstract Calibration of the wastewater treatment plant (WWTP) models is a challenging task, because reliable collection of measured data and good process knowledge are required in each particular case. This research presents the case study of a Romanian municipal WWTP with the Anaerobic-Anoxic-Oxic (A 2 O) configuration. Activated Sludge Model No. 3 (ASM3) is the core of the developed WWTP model. The proposed WWTP model calibration methodology assumes the selection and computation of a set of influent variables, associated to a group of process and settler parameters, such as they fit the model predictions to the plant measurements. The objective function used for calibration consists in the absolute difference between the measured and model predicted effluent data and its constrained minimization finds the decision variables as calibration targets. Three optimization approaches were selected and investigated: genetic, genetic-hybrid and multi objective Pareto algorithms. The calibration performance of the different optimization methods was evaluated and compared. The genetic-hybrid optimization method showed the best performance as the model predicted data approximates well the effluent measured data both in steady and dynamic state. The properly calibrated WWTP model was complemented with nitrification and denitrification control loops, aimed to improve the municipal WWTP operation by enhancing the effluent quality and reducing the aeration and pumping energy.