Ozonation Digital Twin For Predicting Micropollutants Removal At The Wervershoof (NL) WWTP, Pilot and Demonstration-Scale Testing

INTRODUCTION
Indirect reuse of wastewater treatment plant (WWTP) effluent can play an important role in building a resilient water future (Bullard et al., 2019). When reuse is considered, the presence of micropollutants (MPs) is an issue that needs to be addressed. Wervershoof WWTP (the Netherlands), operated by water authority Hoogheemraadschap Hollands Noorderkwartier (HHNK), illustrates of how the different parts of the water cycle are interlinked (Hoekstra et al., 2021). Ijssel Lake is the receiving water body, a main drinking water source of the North-Holland province. HHNK, PWN Water Supply Company North-Holland (PWN), and PWNT are working closely together on an advanced treatment project to tackle MPs problem. In this collaborative framework, knowledge in drinking water production and wastewater treatment with today's power of digital twins for advanced testing and monitoring, were combined to target MPs removal. A 700 m3/h ozonation facility that allows the testing of three different ozone dosing systems will be used. Reuse will also be explored using an additional pilot-scale facility of 5 m³/h combining ozone, coagulation, and ceramic membrane filtration. For each of those systems, AM-Team is currently building a tailored digital twin based on the mechanistic ozonation model AMOZONE to: 1) predict in real-time MP removal, bromate formation, and other key variables, 2) assess the impact of upstream WWTP dynamics on ozonation, 3) run in-parallel virtual piloting test and 4) maximise process efficiency and performance. This paper outlines the digital twin concept for advanced treatment and provides the first model calibration results. This is the first step in obtaining digital twins for aiding the removal of MPs at the Wervershoof WWTP.
MATERIALS AND METHODS
Digital twin AMOZONE is a kinetic process engineering model for ozonation and advanced oxidation processes that mechanistically predicts ozone (O3) and hydroxyl radical (HO*) concentrations in real water matrices (Audenaert et al., 2019). The removal of MPs is predicted based on literature reaction rate constants (O3 and HO*) and the model is fed with real water matrix data and configured with actual process settings (Figure 1a). The model contains equations for bromate formation and chemical reactions of other species (carbonates, nitrite, ammonia, …). The model allows the assessment of different reactor configurations and O3 injection strategies (Figure 1b). For model calibration, experiments from bench scale bubble column were used and the real wastewater characteristics (e.g. DOC, pH, Alk, etc.) and process settings (volume, gas flow rate and concentration, …) were implemented as model inputs. Experimental results of MP removal and bromate formation were used to assess the model performance. Experimental data The Wervershoof WWTP is a conventional activated sludge plant with a capacity around 270.000 person equivalent. The results from testing with a 6L bench-scale bubble column (Figure 2) using the real WWTP effluent were used to calibrate and validate the AMOZONE model. The column was used to study MP degradation and bromate formation at different O3/DOC dosing ratios (0.2 to 1.5 g/g). A fixed gas flow was injected for a given period to achieve the required O3/DOC dosage. Prior to the experiments, the bubble column was filled with 6L of the intended water matrix. The fluid flow direction was from the top of the reactor to the bottom. Another pilot scale ozone installation treating 10 m3/h and using a hybrid (Roturi) injection system (manufacturer Up2e!), was used to perform the same tests on the same WWTP effluent. Results from this installation were used for further validation of the AMOZONE model. All experiments were run with the real WWTP effluent. In parallel with calibration of the kinetic model based on bench-scale data, all four ozonation systems will be modelled in 3D using advanced computational fluid dynamics (CFD) simulations with integrated ozonation kinetics. Thus, the final digital twins will not be based on ideal hydraulic assumptions (e.g. 'completely mixed') and will include real non-ideal flow behaviour.
RESULTS AND DISCUSSION
The comparison between predicted and measured MP removal during model calibration is shown in Figure 2. The experimental results had decreasing uncertainty (on the ozone dose) with increasing ozone dose particularly for the bench-scale bubble column. Interestingly, the model showed increasing agreement with the experimental results at increasing ozone doses. Important to note is that literature reaction rate constants for the individual MPs were used in the model and remained unchanged also during the validation of the model. For the experiment with the 0.7 O3/DOC ratio, HO* radical formation seemed to be overestimated by the model, leading to higher predicted removal of ozone resistant MPs. The calibrated model also replicated the bromate experimental data, however, further verifications of the experimental results are ongoing. Despite the same water origin, the samples used for calibration and validation were significantly different from each other in composition (DOC, N-species, pH, …), these differences were considered by the model (see Figure 1a). The agreement between the model results and the experimental data increased with the increasing ozone dose, confirmed the reduced experimental uncertainty on the actual dose applied (Figure 2 and Figure 3). The overestimation of HO* levels became more visible at lower O3/DOC ratios, e.g. Diclofenac.
An additional validation step was performed with the Roturi pilot results maintaining unchanged the calibrated and validated model parameters derived from the bubble column experiments. The average removal of the 11 Dutch indicator MPs is reported in Figure 4 along with the comparison with the experimental data. As visible from the overestimation of the MPs average removal at the lower O3/DOC ratio, the modelled and experimental dose might not be aligned. This could be due to the different transfer efficiency of the injection system which was difficult to assess at O3/DOC ratio lower than 0.3 due to generator limitations. Currently, the experimental results of bromate need further verification before the comparison with the predicted bromate formation.
CONCLUSIONS
In the framework of developing a full-scale digital twin, the initial calibration and validation results of the AMOZONE digital twin model look promising. The model was calibrated and validated using experimental results from ozonation tests performed on Wervershoof WWTP effluent. The model was also additionally validated against the Roturi, a completely different experimental setup in terms of size, O3 injection system, but treating the same WWTP effluent. The model captured the main trends in MP removal at different doses. Different ozone doses, secondary effluent characteristics and the MP specific rate constants were considered. This was the first step in obtaining digital twins for the removal of MPs at the Wervershoof WWTP. The digital twins will help in building a sustainable water future in North-Holland.