Significant experimental literature exists on nutrient recovery from manure, focusing on technologies such as gas-permeable membranes (GPM), air stripping (AS), reverse osmosis (RO), electrodialysis (ED), and struvite precipitation (SP). However, none of these processes have been reported to achieve a concentration factor higher than 10. Additionally, most of these methods require costly pH adjustments, high temperatures, or high pressures. ED is an electrified process that has been proven to have better efficiency compared to conventional pressure-driven processes for desalination. Furthermore, ED can be operated with very minimal chemical additions, making it the process with the most potential. These reasons justify a detailed modeling and optimization study to improve the current knowledge of ED capabilities for nutrient recovery.
The modeling and optimization literature for ED primarily focuses on desalination, resulting in a lack of optimization studies that explore the impact of various parameters at high-concentration factors. This talk will present two mathematical models (1-D, 0-D) of ED adapted from the desalination literature to better explain nutrient recovery. Both models encompass phenomena of ionic mass transfer and water transport, current-voltage relationships, and mass conservation equations. The simplified 0-D ED model employs algebraic representation for conservation equations and is computationally tractable for global optimization. This lower fidelity model is used to investigate the effects of various parameters such as membrane area, current, cell-pair voltage, and flow split.
The conventional heuristic design splits the feed stream equally between the concentrate and dilute channels of ED, inherently limiting the maximum achievable concentration. With osmosis counteracting the effect of ionic transfer, it is imperative to introduce less water into the concentrate channel. Therefore, we introduced a variable called flow split, which determines the proportion of feed flow directed to concentrate and dilute channels. Lower values of flow split favor larger feed flow through the dilute channel, while higher values prefer the opposite. For high-purity separations, the optimizer chooses to maintain very low flow split values. However, upon reaching the lower limit of ‘zero,’ the optimizer shifts its strategy to using the membrane area for controlling osmosis. With these new strategies, ED capabilities are extended to produce higher purity products up to 5 wt% N from a relatively dilute feed of 0.13 wt% N (Concentration Factor = 35).
The optimal designs are validated by comparing the 0-D optimization model predictions to higher-fidelity 1-D model simulations of the same designs. The latter simulations are performed using Aspen Plus with a custom ED unit operation developed using the CAPE-Open interface. These simulations help elucidate the shortcomings of the 0-D model. Furthermore, the presentation will discuss how optimal designs vary in response to factors such as variability in feed concentration, target dilute concentration, and the cost of the membranes. The significant contribution of this work lies in its optimization and validation framework, which can be readily extended to evaluate diverse test cases.