(188f) Machine Learning Modeling of Electrochemical Recovery of Phosphorus from Municipal Wastewater in a Batch Reactor

Phosphorus is an irreplaceable macronutrient necessary for plant growth and other industrial processes, whereas when in excess in water bodies can lead to eutrophication. Recovery of phosphorus from wastewater became important as demand for fertilizer and supply of non-renewable rock phosphate increase and decrease respectively¹. Electrochemical recovery of phosphorus from wastewater has been established to be efficient, modular and robust with little or no chemical dosage^2,3. This approach can also be used to achieve a process that enables easy automation with little or no human intervention. However, the efficiency of electrochemical driving force depends partly on the water chemistry¹ which varies based on the type of wastewater and treatment method. To achieve an efficient automated process, process parameters need to concurrently change based on the composition of the wastewater and the intended percentage recovery⁴. Therefore, it is important to develop a model that takes both process and stream parameters as input variables to determine the percentage recovery of phosphorus as an output variable.

This research study aims to develop an experimental data-driven model to predict the recovery of phosphorus using various machine learning algorithms. One major advantage of machine learning is that it doesn’t require understanding of the science behind the process⁵ nor does it make any presumption about the data⁶. Hao et al. developed a model for removal of phosphorus in an electrochemical fixed bed reactor packed with magnetite⁴, however, process variables such as temperature and turbulence that could affect the reaction kinetics and mass transfer respectively were not considered in the model development. Recently, the use of sacrificial magnesium anode as opposed to an inert anode with magnesium salt^2,7 has been found to facilitate phosphorus recovery, especially when struvite (a slow release fertilizer) is the desired precipitate⁸. Therefore, different anode types (sacrificial versus non-sacrificial) will also be used as an input variable for the model development.

Preliminary data has shown recovery of phosphorus to be directly and inversely related to the initial concentration of ammonia and phosphorus respectively in the wastewater being treated as indicated in the attached figure. This initial data, which varies the concentration of major ions in the wastewater and process parameters including cathodic potential, size of the electrode, temperature, and turbulence, will be used for model development. Machine learning algorithms such as multiple linear regression, random forest and support vector machine will be used to train, test, and validate different models. These various models and feature importance tests will be assessed for precision and ranking of parameters with the most significant impact on the recovery efficiency. These results will be presented at the meeting.

References:

(1) Kékedy-Nagy, L.; English, L.; Anari, Z.; Abolhassani, M.; Pollet, B. G.; Popp, J.; Greenlee, L. F. Electrochemical Nutrient Removal from Natural Wastewater Sources and Its Impact on Water Quality. Water Res. 2022, 210, 118001. https://doi.org/10.1016/j.watres.2021.118001.

(2) Belarbi, Z.; Daramola, D. A.; Trembly, J. P. Bench-Scale Demonstration and Thermodynamic Simulations of Electrochemical Nutrient Reduction in Wastewater via Recovery as Struvite. J. Electrochem. Soc. 2020, 167 (15), 155524. https://doi.org/10.1149/1945-7111/abc58f.

(3) Lei, Y.; Zhan, Z.; Saakes, M.; van der Weijden, R. D.; Buisman, C. J. N. Electrochemical Recovery of Phosphorus from Acidic Cheese Wastewater: Feasibility, Quality of Products, and Comparison with Chemical Precipitation. ACS EST Water 2021, 1 (4), 1002–1013. https://doi.org/10.1021/acsestwater.0c00263.

(4) Hao, J.; Zhang, J.; Li, X.; Qiao, M.; Zhao, X. Machine Learning Facilitates the Application of Electrochemically Induced Precipitation for the Removal of Phosphorous. ACS EST Water 2023, 3 (2), 616–625. https://doi.org/10.1021/acsestwater.2c00631.

(5) Cao, B.; Adutwum, L. A.; Oliynyk, A. O.; Luber, E. J.; Olsen, B. C.; Mar, A.; Buriak, J. M. How To Optimize Materials and Devices via Design of Experiments and Machine Learning: Demonstration Using Organic Photovoltaics. ACS Nano 2018, 12 (8), 7434–7444. https://doi.org/10.1021/acsnano.8b04726.

(6) Nageshwari, K.; Senthamizhan, V.; Balasubramanian, P. Sustaining Struvite Production from Wastewater through Machine Learning Based Modelling and Process Validation. Sustain. Energy Technol. Assess. 2022, 53, 102608. https://doi.org/10.1016/j.seta.2022.102608.

(7) Belarbi, Z.; Trembly, J. P. Electrochemical Processing to Capture Phosphorus from Simulated Concentrated Animal Feeding Operations Waste. J. Electrochem. Soc. 2018, 165 (13), E685. https://doi.org/10.1149/2.0891813jes.

(8) Tan, X.; Yu, R.; Yang, G.; Wei, F.; Long, L.; Shen, F.; Wu, J.; Zhang, Y. Phosphate Recovery and Simultaneous Nitrogen Removal from Urine by Electrochemically Induced Struvite Precipitation. Environ. Sci. Pollut. Res. 2021, 28 (5), 5625–5636. https://doi.org/10.1007/s11356-020-10924-8.