Inverse Modeling of Nitrification-Denitrification Processes: A Case Study on the Blue Plains Wastewater Treatment Plant in Washington, DC.

Evaluating the uncertainty associated with the predictions of Activated Sludge Models (ASMs) is essential in designing and optimization of biological wastewater treatment systems. The sources of ASM model prediction uncertainties can be classified into influent and environmental factor uncertainties, parameter uncertainty and epistemic uncertainty due to model abstraction. In this study, observed data obtained from nitrification-denitrification processes at the Blue Plains advanced wastewater treatment plant is used to quantifying the uncertainty associated with ASM parameter estimation. The posterior distributions obtained for parameters are then used in a Monte Carlo simulation to assess the uncertainties of predicted effluent constituents expressed as the 95% credible intervals of effluent concentrations. Using the outcome of the parameter estimation step, two scenarios of methanol loading scheme including one based on flow-rate (baseline) and the second one based on total nitrogen in the influent were evaluated. The treatment performance of the system was quantified based on the probability distributions of effluent violation with respect to certain water quality standards. The results show that methanol loading based on influent flow-rate is less effective compared to determining methanol loading based on total nitrogen loading in the influent.