Black & Veatch’s Frank Rogalla looks at optimising plant with modelling software
Factors affecting overall process efficiency include mixed liquor recycle (MLR) rates, dissolved oxygen (DO) concentrations, solids retention time (SRT), temperature and wastewater characteristics including the carbon to nitrogen (C:N) ratio. Engineers and operators examine these factors closely when planning operations, however, there are some situations that are unavoidable. For example, a WwTW may experience high influent ammonia for a short period of time, but operators may be unsure if the system will be able to handle the extra load or if additional basins need to be brought online. Process modelling tools allow operators to examine such situations without changing operational conditions. Model output can provide guidance in selecting target SRTs for a plant, MLR rates, solids handling schedules, etc. Models can also provide insight into the effects of uncontrollable changes within the plant.
Sensitivity analyses were performed in which operational parameters were varied for different plants. Effluent characteristics were compared across a range of the parameter of interest. By plotting the data, optimal operational conditions were determined for the specific plant. Once such ‘optimum’ values were determined from the sensitivity analyses, the information was used to fine-tune the operational condition at the plant. This technique provides a good estimate of optimum conditions, is simple to follow and can be an extremely powerful device in testing a wide range of operating parameters.
For most of the sensitivity analyses, the BioWin (Envirosim) package was used. A ‘baseline’ model was established by building the plant unit-by-unit with the respective process components. The baseline model was setup and calibrated using actual plant data. Figure 1 shows the configuration of a typical BioWin model.
Once the configuration was built, details of each unit process were entered according to the specifics of the particular plant. Influent wastewater characterisation required using a separate stoichiometry spreadsheet to estimate the fractionation of COD, or the fractions of COD that were biodegradable/soluble, biodegradable/particulate, unbiodegradable/soluble, and so on. In all 17 wastewater influent parameters are input in addition to flowrate, COD, total kjeldahl nitrogen (TKN), total phosphorus, nitrate, alkalinity, inert suspended solids (ISS), magnesium and dissolved oxygen. If some of the plant data is not available the model is set with default values. Influent characteristics can also be set up as dynamic waste flows if diurnal plant data is available. Parameters such as temperature, autotrophic and heterotrophic kinetics, settling parameters and biomass stoichiometry can be varied from default values if data is available.
Influent characteristics are by far the most time consuming and potentially difficult part of setting up the model. Clarifier input includes volume (or area) and depth, underflow rate, percent solids removal, and sludge blanket fraction of settler height. Activated sludge basin input includes volume (or area) and depth, and airflow specifications (if aerated) such as DO setpoint. MLR can be set up according to a specified flowrate or a percentage of the influent. Wastage rates can be specified as a specific flowrate, or paced as a percentage of the influent, or determined by BioWin by specifying an SRT.
Once BioWin configurations were established and parameters set, a ‘check’ feature located any incompletely set-up items. Finally, models were run at steady state to solve effluent characteristics, solids loading rates on the clarifiers, sludge production quantities, etc. The model can be run dynamically.
When performing the sensitivity analyses, one factor of interest was changed in the previously calibrated baseline models to observe changes in effluent quality. The models were run over a range of values for each process parameter. For the analyses, all factors except for the factor of interest were kept constant. Once the models had been run over a range of values, the model output showed effluent quality versus the adjusted variable, describing how sensitive the WwTW was to changes in a particular parameter, and pointing to optimum operating conditions.
The first sensitivity analysis was performed for a WwTW in Florida that is being expanded to achieve an average effluent total nitrogen concentration of 7.0mg/l or less. The plant is converting from a two-stage BNR process to a four-stage Bardenpho process. As part of the upgrade, sensitivity analyses were performed to optimise the design. In the sensitivity analyses the MLR was varied. At rates below 400% of influent flow, higher quantities of nitrates left the first aerobic reactor rather than returning to the first anoxic, or pre-denitrification reactor. The second anoxic reactor was then unable to denitrify these higher levels of nitrates, and effluent concentrations rose.
MLR rates of 500 and 600% did not provide any significant benefit with regards to effluent characteristics when compared to 400%. This becomes important when power costs associated with pumping become an issue. A suitable operating MLR for this plant was determined to be in the range of 400-500%. Furthermore, detrimental effects were noted as effluent nitrate and total N increased above 600% recycle. This was likely due to the return of high DO concentrations to the first anoxic zone.
Using the sensitivity analysis feature in GPS-X (Hydromantis) in an activated sludge plant in Connecticut the program repeatedly ran the model and increased the SRT by one day between each run to automatically generate the sensitivity analysis output. The purpose of this analysis was to see at what SRT the model predicted nitrification would occur.
The first step of nitrification, converting ammonia to nitrite, started to occur between an SRT of 7-8 days. At this point nitrite was observed in the effluent but no nitrate was seen. At some point between 10-11 days SRT, nitrate was observed in the effluent and there was a decrease in nitrite concentration. This indicated the second stage of nitrification, oxidation of nitrite to nitrate, was established. For this system, to ensure nitrification is complete, the SRT should be maintained above 12 days SRT, including the anoxic reactor volume.
The minimum effluent nitrogen concentration occurred at an SRT of 9-10 days, just prior to the onset of the second stage of nitrification. Some novel activated processes claim to achieve high degrees of denitrification, and hence low effluent total-nitrogen, by encouraging the reduction of nitrite to nitrogen gas rather than reducing the fully oxidised nitrate. This sensitivity analysis supports the theory behind the idea, however it is unlikely a plant could be operated to control the SRT at a point where nitrite formation is occurring but nitrate is not yet produced. The SRT at which nitrification starts to occur is very temperature dependent and it would be impractical to vary the SRT control point to match this.
The effect which increasing the influent flow would have on the effluent total-nitrogen was simulated at a facility in Florida. It has a permit for total-nitrogen and uses a process that includes partially aerobic (low DO) zones to make use of simultaneous nitrification and denitrification to produce very low effluent total nitrogen. The model was run as a dynamic simulation in Biowin.
Sensitivity analysis showed the WwTW should be able to achieve the lowest effluent total nitrogen as the plant approaches its design capacity. At lower flows there was insufficient load to drive down DO concentrations to produce the conditions required to remove nitrates. At higher flows the plant had difficulty in achieving full nitrification.
To upgrade a facility in Michigan to provide both biological N and P removal, the model showed removal to improve as the influent COD concentration increases. The influent COD needs to be above 500mg/l to ensure good P-removal. As the influent COD increased further, the improvement in P removal was much less pronounced than the improvement in denitrification, indicated by a near-linear decrease in effluent nitrate until about 700 mg/l COD l