(589d) Using Data-Driven Modeling and Systems Optimization for Advancing Sustainable Nutrient Recovery Technologies from Concentrated Wastewater Sources

The discharge of nitrogen (N), phosphorus (P), and potassium (K) from wastewater treatment plants contributes significantly to nutrient pollution, leading to harmful algal blooms and eutrophication in freshwater ecosystems [1,2]. On the other hand, as per recent statistics [3], nearly 13% of the global nutrient demand can be supplemented via the recovery of these key fertilizer nutrients from wastewater. Nutrient recovery from wastewater thus offers a win-win solution of food and environment. Toward this direction, a novel membrane separation process has been experimentally developed in the team’s previously work [4], demonstrating superior N/P/K separation efficiency and energy savings. In this presentation, we aim to quantitatively model and analyze this process to identify the optimal process design conditions.

Machine learning-aided modeling is first leveraged to unravel the structure-property-process relationships governing membrane performance based on experimental data encompassing various membrane properties, process conditions, and performance metrics. To address the challenge of limited data from experiment, the Synthetic Minority Oversampling Technique (SMOTE) is utilized to generate synthetic data based on the original experimental data [5]. A neural network (NN) model is then trained to extract underlying nonlinear input-output relationships. This SMOTE-integrated NN modeling approach has been demonstrated in a previous study for microwave-assisted ammonia production, and we aim to extend its application to membrane separations [6]. The resulting data-driven structure-property-process relationships are further integrated with first-principles membrane models, from which sizing and costing can be evaluated. On this basis, process design optimization is conducted to assess the economic viability of our proposed nutrient recovery process at an industrial scale. Inverse design is also performed to identify the corresponding structural parameters for membrane synthesis. Our analyses strive to provide insights into process bottlenecks or opportunities to enhance economic competitivity and production scalability of such sustainable nutrient recovery processes.

Reference:

[1] Chislock, M. F., Doster, E., Zitomer, R. A., & Wilson, A. E. (2013). Eutrophication: causes, consequences, and controls in aquatic ecosystems.Nature Education Knowledge, 4(4), 10.

[2] Preisner, M., Neverova-Dziopak, E., & Kowalewski, Z. (2021). Mitigation of eutrophication caused by wastewater discharge: A simulation-based approach.Ambio, 50(2), 413-424.

[3] Facts about wastewater and nutrient management. United Nations Environmental Programme.

[4] Piash, K. S., Anwar, R., Shingleton, C., Erwin, R., Lin, L. S., & Sanyal, O. (2022). Integrating chemical precipitation and membrane separation for phosphorus and ammonia recovery from anaerobic digestate. AIChE Journal, 68(12), e17869.

[5] Chawla, N. V., Bowyer, K. W., Hall, L. O., & Kegelmeyer, W. P. (2002). SMOTE: synthetic minority over-sampling technique. Journal of artificial intelligence research, 16, 321-357

[6] Masud, M. A. A.; Araia, A.; Wang, Y.; Hu, J.; Tian, Y., Machine Learning-Aided Process Design for Microwave-Assisted Ammonia Production. Computer-Aided Chemical Engineering. Under review