Modeling of simultaneous carbon and nitrogen removal (SCNR) in an anaerobic/anoxic reactor treating salmon fishery wastewater

Abstract

Mathematical modeling can help to clarify some aspects of the process that promote the simultaneous removal of organic matter and nitrogen (SCNR) from the fish processing industry wastewater that still require elucidation, such as the competition between methanogenic and denitrifying microorganisms for the substrate, for example. Therefore, the aim of this paper is to present the application of the Anaerobic Digestion Model No. 1 (ADM1) for modeling the SCNR process in a single stage anaerobic/anoxic reactor fed with a mix of a synthetic substrate and a saline protein-rich salmon processing industry. The ADM1 was not specifically developed for the treatment of this wastewater, therefore, initially, a parametric sensitivity analysis was performed to define the most sensitive parameters. The sensitivities were quantified in terms of the variation of measurable process variables under the perturbation of model parameters in their neighborhood domain. After this analysis, the model was calibrated and validated using previously published experimental data. Based on a sensitivity analysis, five parameters of the original ADM1 and of the denitrification processes were estimated using iterative methods, which optimized the parameters with experimental results. Namely: proteins hydrolysis constant (khyd,pr = 0.05 h−1); Monod maximum specific uptake rate for amino acids utilizers (km,aa = 1.87 h−1); half saturation value for amino acids utilizers (KS,aa = 0.23 kg COD m−3); Monod maximum specific uptake rate for acetate utilization rates by the denitrifiers (kno3,ac = 0.09 h−1) and half saturation value for the utilization of the acetate by the denitrifiers (KS,no3,ac = 0.38 kg COD m−3). Simulation with calibrated parameters indicated good agreement with experimental data, with the mean absolute error below 40 %. Considering these optimal values, the complete reduction of the oxidized forms of nitrogen occurred up to 11 h of reaction time. Cross-validation results indicated that the calibrated model was able to predict the experimental results of effluent COD, acetate, propionate, butyrate, nitrate and nitrite flows with good accuracy showing a mean absolute error below 40 % (19 %–36 %).