Modeling, Instrumentation, Automation, and Optimization of Wastewater Treatment Facilities

Abstract

MODELING Mechanistic ModelsA review by Hu et al. (2003) of available activated sludge models led them to conclude that the most complete model was the Barker and Dold model and that the highly interactive nature of its processes, particularly with respect to the chemical oxygen demand (COD) loss mechanism, resulted in a non-unique identification of certain model parameters using the Wentzel et al. pilot scale data sets. Muller et al. (2003) validated the theoretical reduction of the anoxic yield to 79% of the aerobic yield value of 0.67 mgCOD/mgCOD, and noted that in updating the Activated Sludge Model 1 (ASM1) model parameters to reflect this change, the value of the hydrolysis reduction coefficient _ g should be adjusted to reflect the higher denitrification potential associated with the lower anoxic yield. The applicability to the ASM3 parameters was unclear. Jimenez et al. (2003) fitted pilot test data with a first order particulate COD removal rate, and concluded that flocculation of colloidal COD is slow and comparable to the removal rate of soluble COD and that flocculation of suspended solids is rapid but not instantaneous. Manga et al. (2003) used a modified version of ASM2, including the modeling of phosphate/glycogen accumulating organism (PAO/GAO) competition to model the data obtained at the anaerobic, anoxic, and aerobic compartments of a pilot scale modified University of Cape Town (UCT) process operated at solids residence times (SRTs) of 12, 14, and 16 days. Good agreement of effluent soluble P and nitrate-1437a N was obtained at the SRTs of 16 and 14 days, as opposed to the 12 day SRT. Copp et al. (2003) presented a state variable mapping between ASM1 and anaerobic digestion model 1 (ADM1) based on mass balance for COD and total Kjeldahl nitrogen (TKN) to facilitate the coupling of the two models. Fonseca et al. (2003) described a customized MS Excel interface to a plant's GPS-X TM simulator software that facilitated ease-of-use of the simulator by the plant operators by including a familiar description of the plant's processes, measurements and nomenclature and restricting access for modifying process model parameters. Using computational fluid dynamics (CFD) simulation, Vermande et al. (2003) obtained good agreement of the horizontal velocity profile along the length of an oxidation ditch bioreactor, in particular for velocity measurements near the ditch's bottom and near the water surface. Littleton et al. (2003) summarized the results of an investigation of simultaneous biological nutrient removal in Orbal TM processes and concluded that both the bioreactor macroenvironment and floc microenvironment are important factors in establishing a balance among the required nutrient removal microorganism populations. A full-scale extended aeration multi-loop bioreactor was modeled by Insel et al. (2003b) using ASM2d and 20 days of 15-minute measurements of flow rate and nutrient concentrations. They calibrated the inert particulate COD fraction of influent wastewater and 8 of the ASM2d parameters and...

Shall we use hardware sensor measurements or soft-sensor estimates? Case study in a full-scale WWTP

“…This leads to considerably time-delayed responses unsuitable for the monitoring and control of the dynamic treatment processes. Fortunately, considerable developments in the on-line instrumentation in municipal wastewater treatment plants (WWTPs) have occurred during the past decades, providing more real-time information on the evolving process conditions (Vanrolleghem and Lee, 2003;Vanrolleghem et al, 2006;Madsen et al, 2011;Sweeney and Kabouris, 2013;Olsson et al, 2014). In spite of these developments, the instruments still tend to get fouled in the biological treatment processes, giving rise to a poor performance of sensors (Olsson, 2012).…”