Uncertainty Analysis and Multi-Objective Control Design of Methanol-Feed Nitrification-Denitrification Process in Municipal Wastewater Treatment Plant

In this study, a full-scale methanol-fed nitrification-denitrification process with methanol as external carbon source was studied. This process has been widely used in the municipal wastewater treatment plants to lessen the total nitrogen amount of treated wastewater prior to discharge to the receiving water or into the water recycling system. When methanol is available as an electron donor over a long term, a new type of specialist heterotrophic bacteria is enriched in the activated sludge culture; these bacteria are collectively referred to as methylotrophs and they have a unique reactions and stoichiometric and kinetic properties. The common activated sludge models (ASMs) do not include the methylotrophic reactions into their reaction network table. Therefore, ASM1 was modified by adding three new reactions and two new constituents to take into consideration the methylotrophic microorganism as well as generalist heterotrophic and autotrophic bacteria. To calibrate the modified ASM, approximately 7,000 samples were collected over a period of 150 days from effluent as well as in the reactor tanks from Blue Plains advanced wastewater treatment plant in Washington DC. To earn a prior knowledge about the endogenous respiration rate and the observed biomass yield of the denitrifying methylotrophic biomass, a series of secondary batch tests were conducted to measure denitrification rates and observed biomass yields. The regression analysis on the declining denitrification rate data showed 95% confidence intervals of 0.130 +/- 0.017 day--1 for the endogenous respiration rate under aerobic conditions at 20 °C, 0.102 +/- 0.013 day--1 under anoxic conditions at 20 °C, and 0.214 +/- 0.044 day --1 under aerobic conditions at 25 °C. The observed biomass yield value showed an increasing trend from approximately 0.2 to 0.6 when the starvation time increased from 0 to 10 days. The Bayesian parameter estimation framework was used to evaluate the ability of the measured data from the lab-scale and full-scale nitrification-denitrification bioreactor to reduce the uncertainty associated with the bio-kinetic and stoichiometric parameters of the modified ASM. In this framework, the batch test results plus the reported values in the literatures and other studies were used as prior knowledge for the unknown parameter values, and the collected measured data were used as evidence points to obtain the posterior distribution for the unknown parameter of the modified ASM. A hybrid genetic algorithm and the Bayesian inference were used to perform deterministic and probabilistic parameter estimations, respectively. Sensitivity analysis was also performed to explain the ability of the data to provide information about each of the parameters. The results showed that the uncertainty on the estimates of the most sensitive parameters (including growth rate, decay rate, and yield coefficients) decreased with respect to the prior information. Both the deterministic and stochastic calibrated ASM were used for model uncertainty analysis, alternative methanol dose criteria analysis to minimize the methanol consumption, prediction of system performance under different operating conditions, and control of the processes in the wastewater treatment plant. In the application point of view, the calibrated modified ASM was used in a feed-forward multi-objective model predictive control (MPC) scheme to minimize the overall operating cost of the methanol-fed nitrification-denitrification system while meeting the regulatory thresholds for effluent quality.