A STEP CLOSER TO REALITY: NEW DEVELOPMENTS IN PREDICTING DISTRIBUTED PAO STATES FOR ENHANCED BIOLOGICAL PHOSPHORUS REMOVAL

Computer simulation of activated sludge system performance is a critical tool for design, operation, and troubleshooting, and simulation of nutrient removal systems has proven to be particularly challenging. EBPR systems cycle bacteria through anaerobic and aerobic reactors, which selects for polyphosphate accumulating organisms (PAOs) containing several microbial storage products (polyphosphate, glycogen, and polyhydroxyalkanoates). Conventional simulation programs utilize a “lumped” approach where process rates are calculated using bulk concentrations of biomass and microbial storage products as inputs to sets of biokinetic equations. However, a newly developed activated sludge simulation program (DisSimulator) demonstrated that a range of PAO states (internal microbial storage product contents) is likely EBPR system, due to the variety of hydraulic residence times experienced by individual PAOs (Schuler, A.J., in press). The total calculated process rates calculated using the state distributions were considerably lower than those predicted using the lumped approach. The result was that lumped simulations consistently overestimated EBPR performance - this would tend to produce less conservative EBPR system designs than the distributed approach, and this could lead to undersized systems.In the current research, the effects of increasing the number of anaerobic and aerobic reactors in series on PAO states and EBPR performance were evaluated. It is known that as the number of reactors in series increases, the distribution of hydraulic residence times (HRTs) decreases and plug flow conditions are approached (uniform hydraulic residence time for all hydraulic elements), as is assumed in lumped simulations. Because HRT variation is the primary factor effecting PAO state distributions, there is a need to determine how changing hydraulic configurations affects EBPR performance with respect to state distributions. It was demonstrated that diversity decreased as the number of completely mixed reactors in series increased, and this was because hydraulic residence time distributions decreased with increasing numbers of reactors in series (plug flow was approached). Although increasing the number of reactors in series brought lumped and distributed predictions closer together, there were still large differences in these predictions, and so accounting for distributed states in full-scale systems is still likely to be important even in systems with several reactors in series. Based on these results, it appears that continued development of the distributed approach to activated sludge simulation has the potential to improve design and operation of biological nutrient removal systems.