JSDEWES: The use of Dynamic Mathematical Models for improving the Designs of upgraded Wastewater Treatment Plants

The existing treatment facilities at Ljubljana WWTP consist of mechanical treatment (screens, grit and grease chamber), biological stage with suspended biomass activated sludge process (three parallel aerobic reactors and four parallel secondary settlers) and sludge treatment (anaerobic digestion and sludge drying). The plant is currently operated at 435,000 PE. It efficiently removes carbon and achieves nitrification, but lacks denitrification and P-removal.

The preliminary design for the upgraded plant as prepared by a consulting company defined suitable technological solutions for the expected increase of future input load, efficient removal of N and P components, as well as for the reception and treatment of sludge from other WWTP. To simulate the upgraded plant performance the model was designed in two stages. First, a model of the existing plant was developed since the upgraded plant will rely on current plant configuration. Simulations of the existing plant operation enable to assess the model quality, especially the appropriateness of input wastewater characterization and replication of operating conditions. Second, the model was supplemented with the proposed technological solutions for the plant upgrading and simulated at increased input load.

The model was designed in CNP library in GPS-X [19], thus considering the removal of organics, N and P components. The design of the model for the current configuration included the following:

• Collection of long-term regular data and additional measurements for the model design;

• Selection of modelled objects;

• Determination of physical parameters and plant operating parameters;

• Influent wastewater characterization and adjustment of model parameters;

• Simulation and model evaluation.

Figure 1 shows configuration of Ljubljana WWTP water line with indicated locations where measurements for the purpose of model design were performed. Measurements included regular daily laboratory measurements, on-line measurements, as well as estimated process variables from sample measurements. In addition, an intensive 5-day measurements were performed specifically for influent wastewater characterization.

• Following the scheme in Figure 1, the designed simulation model included the following objects:

• Influent wastewater from the sewer modelled with “Chemical Oxygen Demand (COD) fractions model” [19];

• Reject water from the sludge treatment line (filtrate from a filter press, centrate from centrifuge, condensate from sludge drying) also modelled with “COD fractions model”;

• Grit chamber modelled with “empiric model” [19];

• Aeration tanks modelled as a plug-flow reactor with four tanks using “ASM2d” model for biological reactions [3];

• Secondary clarifier with a “simple1d” model, i.e., 1-dimensional settler model without biological reactions [19].

For each object the corresponding physical parameters (tank volume, tank surface, water depth, etc.) and operational parameters were determined. Airflow to aerobic reactors was determined by controlling Dissolved Oxygen (DO) concentration in aerobic tanks at measured values. Return sludge flow was set to low or high value, depending on the inflow as in real plant operation. Excess sludge flow was determined by controlling Mixed Liquor Suspended Solids (MLSS) after aeration tanks at measured values (junction 13 in Figure 1).

The performance of the model was improved by more precise influent wastewater characterization and adjustment of model parameters as presented in Table 1. Influent characterization was performed according to STOWA protocol [20] for low loaded WWTP using measurements in intensive 5-day measurement campaign. The estimated fbod and ivt values were higher than the pre-set values in GPS-X, but did not give satisfactory performance - therefore default values were used. Based on parameter sensitivity analysis performed in [6], the three most sensitive model parameters (Y_A, Y_H and μ_max,A) were adjusted in “ASM2d” model to obtain a better fit.

In the upgraded plant, each of the existing plug-flow aeration tanks in the three parallel lines of the biological stage will be transformed into four consecutive tanks, accounting for 26%, 18%, 34% and 22% of the total aerobic reactor volume, respectively. As shown in Figure 3, the plant will operate in two different configurations, depending on the wastewater temperature. BioP-DN configuration operates at high temperatures and enables biological P removal, nitrification and denitrification. DN configuration operates at low temperatures and enables only nitrification and denitrification. Operation in these two configurations requires the change of the first two zones and the change of the internal recirculation flow. In both cases the still remaining P after the biological stage is removed by the chemical precipitation. To achieve the necessary biological activity after upgrading within the existing plant volumes, the biological stage is planned to operate at increased MLSS concentrations. Final cloth filtration will be included to control the effluent Total Suspended Solids (TSS) below the limit value, but was not simulated in our study.

The SBR reactors were modelled with “Advanced SBR” blocks with Mantis model [19] for biological reactions and double exponential function for settling velocity of suspended solids. The model parameters values were kept at pre-set values in GPS-X. In this case, a more detailed model calibration could not be performed since the biological stage will be changed from conventional activated sludge process (currently in operation) to SBR, so additional plant measurements for fine SBR model tuning are not at disposal yet.

To model time delays in the operating schedule among the reactors, as well as time delay of methanol addition, the “Timer Based Splitter” blocks were used. For combining the inflow and methanol addition, and for combining the flow from several reactors, the “Combiner” blocks were used. The simulation scheme of SBR plant in GPS-X is shown in Figure 5.

In modelling SBR reactors the selectors and the recycle sludge from the second to the first part of the reactor were not considered since such a model was not available. It was also assumed that the selectors and the recycle sludge mainly affect the quality of the sludge and prevent sludge washout, while they do not affect the effluent quality significantly.

Filling and decanting of reactors was carried out in such a way that simultaneously two reactors are filled with half of the influent flow, while a single reactor is decanting with a constant flow equal to the prescribed maximum influent flow. The lowest water height thus achieved in the reactor equals the prescribed minimum water level. In simulations this is achieved by a control function that prevents lowering the water level below the prescribed height.

The aeration control of reactors was performed with PI controllers controlling DO concentrations in the reactors at a constant reference value. Reactors are aerated only in the filling and aeration phase.

Waste sludge control in the settling phase was performed with PI controllers implemented with “Modelling Toolbox” blocks. The controller controls daily average TSS concentration at the reactor surface at a reference value by adjusting the waste sludge flow. Removal of waste sludge from the reactor is performed only during the last 20 minutes of the settling phase.

During the first half of the filling and aeration phase methanol could be added in the reactor.

Simulations of the upgraded plant operation were performed by increasing the inflow for 30%, while wastewater characterisation and influent concentrations were the same as in the existing plant simulations. The average inflow thus obtained in one-year plant operation was 110,519 m³/d, while input load was 33.6 t BOD/d, 62.6 t COD/d, 5.5 t N/d, 1.07 t P/d and 33.8 t TSS/d for the BOD5, COD, TN, TP and TSS, respectively. Comparison with predicted input load in the preliminary design (Table 3) shows that for TN and TP the load in simulations is higher than in preliminary, since it takes into account also the load from reject water. It was also assumed that because of the inclusion of primary clarifier and the increased influent PE, the amount of sludge digested will increase, imposing also higher amounts of reject water from sludge line to water line. Highly concentrated reject water is returned to water line in the periods of sludge drying and centrifuge operation. These periods currently amount for approximately 64% of total time. Increased amount of sludge will require almost constant sludge drying and centrifuge operation, and therefore also constant return of highly concentrated reject water to water line.

The upgraded plant was simulated at increased MLSS concentration of 4,500 mg/l in the biological stage. The external recycle flow was simulated as 1.6 times influent flow, while internal recycle flow was 4 times influent flow.

With the simulation model of the upgraded plant three different operation scenarios were considered:

• Plant operated in BioP-DN configuration;

• Plant operated in DN configuration;

• Plant operated in both configurations, depending on the temperature.

Simulations have shown that in all three cases the effluent COD, BOD5 and TSS are almost the same and below the required limit values as given in Table 4. In simulations, the increase of effluent TSS because of higher TSS values in aerobic reactors (4,500 mg/l) was not noticed, which was most probably due to a limited capability of the settler model prediction.

Effluent TN and TP concentrations depend on the simulated configuration as shown in Figure 7. DN configuration gives high TP effluent concentrations, while BioP-DN configuration gives high TN and NH4-N concentrations at low temperatures. A combination of both (third simulation) gives satisfactory plant performance. NH4-N, TN and TP are 99.5%, 90.5% and 65.5% of time below the limit values, respectively. The surplus TP concentrations are eliminated by chemical precipitation.

Bottom two diagrams show simulated wastewater temperature (for all cases), and the switching between DN and BioP-DN for the third simulation case, respectively

Simulations confirm satisfactory plant performance after upgrading, but also pointed out some potential bottlenecks:

• Significant contribution of reject water to input load, which will even increase after upgrading because of larger sludge production. This adds an additional input load to the one predicted from the increase of influent PE;

• Slightly higher temperature for switching between the configurations (around 18 °C) compared to expected 15 °C in the preliminary design. This implies a shorter annual period when the plant could operate with biological P removal;

• Limited denitrification capacity of BioP-DN configuration. As seen in the first diagram in Figure 7, the average effluent TN concentration in BioP-DN configuration (days 25-175) is higher than in DN configuration (days 175 to 390), and is very close to the limit value. In this period, the 85^th percentile of data exceeds the limit value of 10 mg/l. The reason for this is the course of the internal recycle leading from the penultimate nitrification tank, and thus not returning into denitrification zone the NO3-N generated in the last nitrification tank. The proposed measures to improve this performance are:

• Improved denitrification by tighter NO3-N control;

• The change of the internal recycle from 4^th to 2^nd tank in BioP-DN configuration;

• Or additional reject water treatment, e.g., by deammonification.

Operation of Domžale-Kamnik upgraded plant was first simulated for the total SBR volume of 19,000 m³ as planned in the preliminary design. The height of the reactors was 5.5 m, while the lowest water level in the reactors was predicted as 4.13 m. In simulations the influent daily average values were used as measured at the Domžale-Kamnik WWTP after the mechanical stage. Only the days with no major failure of sensors were included. In total 316 days of operation were collected with the following measurements: wastewater temperature, flow, COD, TKN and NH4-N. The measured average flow at the plant (around 19,200 m³/d) was lower than that used in preliminary design (25,000 m³/d). Therefore the inflow was increased by about 30%. The input load thus obtained was still slightly lower than the planned one in Table 5.

Because of the increase of influent flow, the minimum water level in the reactor was lowered from 4.13 m to 3.7 m. In this way the plant is able to treat larger amount of wastewater without overflow. SBR plant operation was simulated for three different waste flows, i.e. 500 m³/d, 900 m³/d and 1,000 m³/d, to potentially obtain better plant performance at higher biomass concentrations. In the filling and aeration phase the set point for dissolved oxygen concentration of PI controllers was set to 2 mg/l. During the filling phase, the methanol was added in the reactor with the flow of 30 l/h. Methanol was added as an external carbon source to complete the denitrification and thus not limit total nitrogen removal because of low denitrification rate. The obtained simulation results for the effluent are shown in Figure 8.

From the figure it can be seen that effluent TN and NH4-N concentrations exceed the limit values of 10 mg/l and 5 mg/l for tertiary treatment, respectively, as also given in Table 4. At high waste sludge flow, the amount of sludge in the reactors is lower, resulting in high effluent NH4-N concentrations. On the other hand, at low waste sludge flow, the effluent TN and TSS concentrations are too high. Hence it was concluded that for the efficient SBR operation it was necessary to use larger reactor volumes.

The operation of SBR was simulated also at increased total reactor volume of 26,400 m³ as proposed in the revised preliminary design. For the inflow new measurements from the real plant operation were collected. Again, average daily measurements were taken after the mechanical stage for 365 days of plant operation. For the missing data, the annual average measured values were used. The influent flow was also in this case, lower than that used in the preliminary design. Therefore the inflow was increased for 27% to reach the design daily average flow of 25,000 m³/d. Also in this case, the input load was still slightly lower than the one planned in Table 5.

The set point for the DO concentration of PI controllers was set to 2 mg/l. The set point for the average daily TSS concentration at the reactor surface was set to 2,750 mg/l to adjust the waste sludge flow by PI controller. Methanol was not added in this case.

Simulated effluent SBR concentrations are shown in Figure 9. For the effluent parameters with the legislation limit values also the percentages of time below the limit values were calculated. They are given in Table 6.

Simulation results show that with the increased reactor volume of 26,400 m³, the average daily COD, TSS in NH4-N concentrations meet the legislation limits in more than 80% of time during one year of plant operation. BOD5 and TN are below the limits approximately 80% of time. Effluent BOD5 exceeds the limit value especially because of the high soluble biodegradable matter in the effluent, while TN exceeds the limit value because of high values of NO3-N and soluble organic nitrogen. The average concentration of soluble organic nitrogen in the effluent is around 4 mg/l.