Process optimization of a technical scale phosphorus recovery system through struvite crystallization at the City of Penticton Advanced Wastewater Treatment Plant

• Main Author: Forrest, Alexander Lebaron
• Format: Others
• Language: English
• Published: 2009
• Online Access: http://hdl.handle.net/2429/15946

This project was divided into three main studies; on-site magnesium detection in wastewater samples, the accurate estimation of the solubility constant of struvite and the optimization o f technical scale reactors for phosphorus recovery in the form of struvite. Two reactors were installed at the City of Penticton AWWTP where they were run for 8 consecutive months; concurrent work was underway on Mg⁺² detection at the laboratory facilities on-site. Finally, the work on solubility was completed at the laboratory facilities at UBC. Ion selective electrodes (ISE) were tested on wastewater samples to determine Mg⁺² concentrations. It was found that both ISE tested produced unreliable results, as they both proved to be non-specific to Mg⁺². A modification using polyaluminum chloride (PAC) was developed to remove the interference of phosphates from the colorimetric technique. It was found to produce reliable results within 10 percent deviation of those results predicted by atomic absorption. The resulting technique averaged about 10 minutes per sample. Three different solubility (pK[sub sp]) prediction models were then examine which provided the best estimate of pK[sub sp]. It was found that the pK[sub sp] had a linear relationship with pH below an inflection point at a pH value of 7.0, becoming relatively constant in the working pH range of 7.0 to 9.0 (typical pH values used for struvite recovery). The model that provided the best estimate was then used to generate values for 3 different temperatures tested; 9, 15, and 19 °C . From the associated plot, the enthalpy of formation of struvite (ΔH⁰) was estimated and found to be in relative agreement with those values taken from literature. The values reported in this work are then 31.62 kJ/mol for ΔH⁰ and 13.28 for pK[sub sp]. The reactors produced about 0.75-1.5 kg of struvite per day, depending on the levels o f the control parameters, including: the upflow velocity (v); the recycle ratio (RR) ; the Mg:P ratio; and the supersaturation ratio (SSR). The generated product was significantly smaller and less dense than the product generated at different study sites operating under similar conditions. This was believed to be the result of some non-quantified inhibiting factor being present (initially suggested as being PAC sludge present from the WTP). The previously studied concept of the crystal retention time (CRT) was abandoned, as it appeared to have little or no influence on the final product. Conversely, the crystal mean size (CMS) of the product proved to be a much better indicator than previous studies had indicated. Multivariate analysis on this variable revealed 2nd order interactions between SSR and Mg:P levels and SSR and RR levels. Two novel devices were designed and built to take in-situ samples of the reactor. This was undertaken in order to construct a vertical profile of the SSR and to estimate the fluidized bed porosity during operation. It was found that the zone of maximum primary growth was in the lowermost section while the zone of greatest secondary growth potential existed midway through the reactor. These were quantified on residence time, chemical driving potential and bed porosity. ICP/MS analysis, zeta potential, polyelectrolyte surface charge and SEM imagery were all employed as techniques in a comparison study of the crystals grown in this study with those at other sites. Although the samples proved to be statistically different, there was no evidence in these tests to provide positive indications as to what those inhibitors may be. Finally, three different models for prediction of the effluent condition were examined. It was found that the two deterministic models, which had been previously studied, provided the weakest estimates (0.25 and 0.62 for values of R²). A new artificial neural network (ANN) model was developed that provided an R² value of 0.86. It was concluded that this model should see further application. === Applied Science, Faculty of === Civil Engineering, Department of === Graduate