(612d) Nitrate and Salt Transport Coefficients for Spiral-Wound RO Membranes Derived from Field Data of Treatment of Impaired Brackish Groundwater

Reverse osmosis (RO) membrane treatment of impaired brackish groundwater often must handle the removal of multiple contaminants. In the present work, salinity reduction and nitrate removal, which is of particular concern in agricultural regions, was analyzed to extract the fundamental membrane element solute transport coefficients and water permeability, and assess their dependence on system operating conditions. The above membrane transport parameters were extracted based on high-resolution field data from two different spiral-wound RO systems (4,000 gallons/day capacity) treating/desalting impaired groundwater of ~50-65 mg/L nitrate as nitrogen, and salinity in the range of ~1,200 – 1,500 mg/L total dissolved solids. The RO treatment system operation was monitored with seventeen process and water quality sensors (e.g., pressure, flow rate, conductivity, nitrate concentration, temperature). The operation of the field systems was over a period of about twelve months, covering a wide range of operating conditions (e.g., permeate recovery, feed nitrate concentration, and temperature). System performance was initially described via machine learning models to identify the most significant attributes that affect membrane/process performance. The relative significance of each model attribute was quantified by its Gini importance, and the incremental contribution to the R-squared (R2) of the derived model. The solute transport coefficients were significantly affected by feed temperature and membrane water permeability. The ratios of salt to nitrate passage and salt/nitrate transport coefficients were also evaluated based on models that account for the impact operating conditions, demonstrating prediction performance of up to R2~0.960. The above results suggest the potential use of the correlation of the above transport parameters for process optimization/control and for assessing nitrate sensor fault/drift and for data imputation.