Full-scale Ozonation Plant for Maximum Micropollutant Removal and Minimal Bromate Formation, Phase 2: Digital Twin Design for Real-time Monitoring

INTRODUCTION The removal of organic micropollutants (OMPs) prior to discharge into the receiving water bodies is gaining interest globally. When it comes to ozonation, The Netherlands is currently investigating with water authorities the feasibility to remove at least 70% of 7 out of 11 guide OMPs (Table 1), with minimal bromate (BrO3-) formation (<1µg/L yearly average). In this framework, the water authority De Dommel (WSDD), in close collaboration with the bouwteam (WSDD, AM-team and CLCWater (combination of ADS, Witteveen+Bos, Moekotte, Nijhuis en Pannekoek)), is constructing a full-scale ozone installation at the waste water treatment plant (WWTP) of Hapert. The integration of Computational Fluid Dynamics (CFD) and the kinetic AMOZONE model was used in the initial phase of this project and reported at WEFTEC2023 (Muoio et al. 2023). In this work, the simulations performed at the design stage were used to design the digital twin (DT). Considering the water matrix, the reactor's geometry and its behaviour at different flows, the ozone generator, and gas diffusers specifications, the DT was design to support the operator's decision making process to achieve both maximum OMPs removal and minimal BrO3- formation. While the full-scale installation is under construction, the DT is currently used for operators' training and preparation to explain the relevant details of ozonation and the impacts of operational changes with the actual water matrix. This crucial step is essential for the staff preparation in understanding the details involved in the correct operation of a complex installation such as an ozone reactor and to finally connect the DT real-time with the installation. METHODS The AMOZONE model (Audenaert et al., 2019) is a detailed mechanistic kinetic model for the prediction of OMPs removal and BrO3- formation including a large number of chemical reactions derived from experience and literature (inter alia: Bourgin et al., 2018; Buffle et al., 2004; von Gunten, 2003). In the CFD-AMOZONE results, two OMPs (benzotriazole and venlafaxine) to evaluate the average removal. In fact, benzotriazole has one of the lowest sensitivities, while venlafaxine can be used as good indicator of the average removal of the 11 compounds (Figure 1). The compartmentalization of the 3D CFD results was performed considering the hydraulics behavior at different flows and the resulting chemical concentrations considering the water matrix fluctuations (Rehman, 2016). This allowed to design a flow-sheet model containing all the spatial information of the detailed CFD-AMOZONE results, but able to run long-term dynamics for parallel computing with the real installation. RESULTS AND DISCUSSION The several simulations performed in the first part of this project with the CFD- AMOZONE tool for defining the best design of the ozone installation at different operational conditions were used to design the DT. The volume compartmentalization allowed to produce a fast model that can run on a regular PC but which includes key information from the hydraulics and chemical points of view that allow very realistic simulations. With the DT operational alternatives can be run using real data from the effluent of the biological installation. In Figure 2, results of dynamic simulations at different ozone dosages report the bromate formation through 15 days with 15min data frequency. At the highest O3/DOC dosage the bromate threshold of 1 ug/L is exceeded at several moments, while reducing the dosage below a ratio of 0.4 shows a much conservative scenario. The OMPs removal is also a dynamic outcome of the ozone dosing strategy, and resulted to be maintained above 80% in average by operating at a dosing ratio of 0.4 O3/DOC. The addition of H2O2 was also tested (data not shown) leading to further BrO3- reduction and to an increase in the removal of those OMPs with higher sensitivity to hydroxyl radicals. Currently simulations are in progress to assure a minimized used of H2O2. For the safety of the microbiome at the natural pond at the effluent of the treatment plant. Finally, the DT confirmed that working at reduced ozone gas concentration can create a double advantage by lowering both bromate formation and energy need while maintaining the required OMPs removal. CONCLUSIONS In the first part of this project the CFD-AMOZONE integration supported the design of the ozonation reactor at Hapert WWTP. During the construction of the ozone installation a DT was created as a result of the 3D results generated in the first phase. The DT is currently used for operators' training and preparation. In this way, the relevant details of ozonation can be illustrated to those who will operate the installation with an intuitive tool, considering practical implications of operational changes and water matrix dynamics on the actual installation which they'll soon operate. Several scenario analysis for the Hapert ozonation reactor were tested with the DT using the AMOZONE model in flow-sheet configuration to maximize the removal of micropollutants and minimize the formation of bromate. The DT allowed to test different ozone dosages in long/term dynamic simulations and define the minimum required dose to contain bromate formation below 1 ug/L and micropollutants removal above the utility's target.