Combining numerical simulation with response surface modelling for optimization of reject water partial...

Optimization of a single-stage, partial nitritation/anammox (PN/A) process for a reject water treatment in a continuous-flow, moving bed biofilm reactor (MBBR) was presented. Response surface method (RSM) was combined with simulation experiments conducted with the validated mathematical model of PN/A in MBBR. The total inorganic nitrogen (TIN) removal efficiency was the response parameter. Eight independent variables were taken into consideration: reject water flow rate (Q), inflow concentrations of the total ammonium nitrogen (TAN), chemical oxygen demand (COD), alkalinity (ALK), pH, temperature (T), dissolved oxygen concentration in the bulk liquid (DO) and aeration time within 60 min intermittent aeration cycle (AERON). Eleven interactions between independent variables were found as significant (p < 0.05). The interaction of AERON*DO had the highest impact on the PN/A process. Optimal values of the controlled variables were found for two cases of MBBR operation. Verification of the optimization was done by the simulation and comparison with the data from the empirical experiments. Under the conditions of the fixed hydraulic retention time of about 38 h, volumetric nitrogen loading rate of 0.48 kgN/m³d, T of 22.5°C, TAN of 750 gN/m³ and optimized values of DO = 3.0 gO₂/m³, AERON = 0.54 h, pH = 7.5, ALK = 80 molHCO₃/m³, COD = 775 gO₂/m³, the predicted TINrem was 78% which is consistent with PN/A performance observed in the technical-scale MBBR systems.