Significance of Process Control Strategy in Dynamic Modeling and Optimization of Full-Scale Post Aerobic Digestion (PAD)

Introduction: Implementation of PAD at Boulder WRRF has reduced sidestream nitrogen loading and improved overall nitrogen removal. Process modeling of PAD process is challenging given the high OUR combined with reductions in transfer efficiency, complex process control including air flow changes based on pH set point and time of the day schedule. In addition, the mechanism for nitrogen removal have not been fully understood. The aim of this modeling effort was to simulate the key process performance parameters for PAD using BioWin Controller. This study shows the importance of process control in operation and performance of PAD process as well as providing insight into key parameters and mechanism of reaction. Methodological Approach: A calibrated whole-plant model in BioWin (adopted from past work) was used for this modeling effort. Process Data from 1/1/2020 to 2/3/2020 was used, using daily data for influent, PAD and secondary digester. The average ammonia removal efficiency was 70%. Daily nitrite data showed very low concentrations of nitrite with negligible accumulation of nitrite. Average nitrate concentration was approximately 1 mg/L. For the operation of PAD, DO is not measured continuously. City staff have indicated that they were not able to detect measurable DO in PAD using handheld meters. The key operating condition for PAD is presented in Table 1. Three modeling approaches were considered in this study (1) DOSP-based model (2) Air flow-based model (3) Controller-based model. Model Setup and Calibration: The model calibration focused on the performance of PAD for VSR, solids concentration, and nutrient removal. BioWin's Model Builder element was used to represent the PAD process. To mimic the increased solids destruction, aeration rate, and nutrient removal efficiency, a few modifications were required to default values including allowance for VSS components (i.e. endogenous decay products and influent undegradable particulate organics) not normally solubilized in the base model to undergo partial hydrolysis. Endogenous products decay rate increased from 0 to 0.1 d-1, temperature dependency for biomass decay and particulate substrate hydrolysis of 1.100 (from the default value of 1.029), aeration parameters adjustments: K1:2.56 and K2:0.05, Alpha factor: 0.1. The alpha factor was determined from an extensive meta-analysis of data in 10 studies for MLSS and alpha factor, as presented in Figure 1. The control logic for the BioWin Controller used a Proportional-Integral (PI) controller with aeration rate as an input and DO as the measured parameter. The DOSP and proportional gain was set at 0.03 mg/L and 100 scfm. Results and Discussion: Dynamic process modeling was also conducted using DOSP-based model and air flow-based model (using actual data). None of these strategies resulted in a nutrient profile that matched the actual data. This clearly shows the importance of simulation coupled with automation control to mimic actual data. Table 2 shows the average model predicted and measured PAD effluent parameters, using BioWin controller. Overall, the model predicted the key parameters well. The Controller air flow rate is very close to the actual value. The average soluble P predicted by the model is around 40 mg/L. No soluble P measurement was reported during the study period for PAD; however, soluble P measurements for early 2020 in centrate was between 20-50 mg/L. Ammonia removal efficiency predicted by the model is 66%, which is close to the actual value of 70%. Figure 1 shows the time variations of N species compared with actual data. Overall, the model predicts PAD ammonia, nitrate, and soluble P well. Table 3 summarizes the key model predicted rates related to nitrogen transformations by AOB, NOB, and OHOs using BioWin controller, compared with DO-based modeling. Figure 2 shows the concentration of key microorganisms in PAD based on modeling with the controller. The results of RNA analysis for three PAD samples are also presented in Figure 2. The relative distribution based on RNA analysis is consistent with BioWin predictions with the Controller(NOB suppression). The rates presented in Table 3 and Figure 2 predict that AOB is the primary group responsible for ammonia removal and there is minimal NOB growth. This suggests that nitrite accumulation may happen in this process, although high nitrite measurements are not observed. This also suggests that nitritation and denitritation may be the primary route of nitrogen removal; however, the mechanism for NOB suppression is not clear and could be due to sensitivity of NOB to temperature (Lantz et al, 2021), or high FA concentration. FA concentration in this simulation is 5-13 mg/L and studies have shown NOB suppression at FA concentrations of 0.1- 5 mg/L (Liu et al., 2020). Mass Balance: Tables 4 and 5 show the mass balance for N species in PAD. One route for nitrogen removal in PAD is ammonia stripping (4% of the influent TKN). Another route is the production of N2O. Some hydrolysis of particulate TKN is also occurring within PAD. Ammonia removal remains the main mechanism for removal of N in PAD. Conclusions: This modeling effort shows the importance of control strategy coupled with adjustment of key kinetic and aeration parameters for PAD. Out of the three modeling approaches considered in this study, the controller-based approach was able to predict the performance accurately.