Understanding the limits of struvite based phosphorus recovery through mechanistic modelling and data-driven performance evaluations

Aguiar, Samuel Enrique

Permalink

https://hdl.handle.net/2142/127384 Copy

Description

Title

Understanding the limits of struvite based phosphorus recovery through mechanistic modelling and data-driven performance evaluations

Author(s)

Aguiar, Samuel Enrique

Issue Date

2024-12-05

Director of Research (if dissertation) or Advisor (if thesis)

Cusick, Roland D

Doctoral Committee Chair(s)

Cusick, Roland D

Committee Member(s)

• Guest, Jeremy
• Espinosa-Marzal, Rosa M
• Jun, Young-Shin

Department of Study

Civil & Environmental Eng

Discipline

Environ Engr in Civil Engr

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

Ph.D.

Degree Level

Dissertation

Keyword(s)

• Wastewater Treatment
• Nutrient Recovery
• Phosphorus
• Struvite
• Crystallization
• Kinetic Modeling
• Machine Learning
• Particle Size
• Fines Loss

Abstract

Managing phosphorus (P) within the Food-Energy-Water (FEW) nexus is key to reducing agricultural reliance on mined rock phosphate fertilizers1,2 and eutrophication of sensitive natural waterways.3–6 One way water resource recovery facilities (WRRF) have implemented P recovery is through struvite crystallization in P rich sidestreams. These systems typically operate with high P removal ranging from 78 – 93%, but can suffer from fine particle washout resulting in highly variable total P recovery from 7 – 91%.7–16 Though some experiential knowledge has been developed to improve reactor performance, a generally shallow understanding of struvite crystallization particle dynamics has led to overall poor performance in terms of P recovery, uncontrolled production of fine struvite crystals, and weak optimization strategies that give operators little discretion in adjusting reactor conditions. This dissertation analyzes struvite precipitation with an emphasis on the importance of particle size and how the loss of fine particles can influence the performance of crystallization reactors. The results discussed range from purely experimental findings based on lab scale crystallization studies, to modeling studies providing some context for those results, and data-driven evaluations of reactor performance. The primary objectives were as follows: (1) Assess the plantwide implications of fine loss to determine if low yield operation can cause major disruptions in effluent composition or other unintended consequences at the WRRF, (2) Evaluate historical performance of a full-scale struvite crystallizer using data-driven methods to uncover correlations between the limited operational parameters available at WRRFs and reactor performance, (3) Develop a comprehensive data set that investigates the influence of supersaturation, organic additives, and calcium co-precipitation on the kinetics of P removal and particle dynamics, and (4) Evaluate various kinetic model structures with increasing complexity and predictive capability to show the importance of surface area-based modeling. The first study in this dissertation showed loss of fines (low yield) 1) is a significant contributor to intraplant P recycling and 2) shift the mass flow of struvite directly to treated effluent water resulting in increases in effluent P. This chapter confirmed the validity of concerns related to the washout of fines and provided motivation for improving understanding of the importance of particle size and developing model frameworks which could address fines loss. The next chapter focused on developing a data-driven approach to evaluating reactor performance through the limited set of operational parameters available at WRRFs. A correlation between conversion and calcium co-precipitants supersaturation was found, but yield seemed to be controlled through an independent set of operational parameters that were not included in this analysis. Data-driven models were shown to predict P removal with high confidence, but were only moderately successful in predicting conversion, and were very weak indicators of yield. While some success can be found through historical data analysis from full-scale nutrient recovery systems, there is a significant gap between the data that is routinely measured at WRRFs and the types of data necessary to fully implement data-driven models. The final chapter returned the focus of the work to connecting crystallization phenomena observed in kinetics experiments with model frameworks capable of accounting for particle characteristics like mass and surface area. Supersaturation, citrate and aspartate, and calcium were investigated primarily for their effects on P removal rate and particle size changes in struvite precipitation experiments. Surface area-based kinetics models are shown to be an acceptable compromise that includes physical particle characteristics and could realistically be implemented at WRRFs. Overall, this dissertation shows the importance of considering particle size and fines loss when evaluating struvite crystallization reactors.

Graduation Semester

2024-12

Type of Resource

Thesis

Handle URL

https://hdl.handle.net/2142/127384

Copyright and License Information

Copyright 2024 Samuel Enrique Aguiar

Owning Collections