The transport of solutes across spiral-wound reverse osmosis (RO) membranes is impacted by membrane material and operating conditions (e.g., flow rate, permeate flux, feed quality and temperature). Solute transport coefficients are also affected by membrane surface fouling that may develop rapidly or slowly over time. Moreover, the rate and severity of fouling are affected by temperature and feed water quality which vary temporally. Here we present a systematic data-driven approach to quantify solute transport coefficients and spiral-wound membrane performance and its degradation under field conditions. The study, based on a three-year operation of three RO wellhead water treatment/desalination systems in Salinas Valley, California having permeate production capacities of 2,500–4,500 gallons/day, treating brackish water of salinity up to 1,500 mg/L total dissolved solid and alkalinity. The three system provided ~4.5-11 million data samples per system each consisting of information from 33 process sensors and operational tags. Data-driven machine learning system performance models, accounting for concentration polarization, were developed to quantify the impacts of spiral-wound membrane compaction, temperature, and fouling on permeate flux, specific energy consumption, and membrane solute transport coefficients. Temperature influenced both the progression of fouling and the hydraulic permeability of the active layer, having a greater impact on permeate flux than estimated from standard ASTM based normalization of permeate flux and solute rejection. Specific energy consumption increased with decreasing temperature (during fall and winter). Based on the data-driven performance models a decision support approach was developed to identify the periods of membrane compaction and fouling progression. In addition, optimal scheduling was assessed for membrane replacement versus clean-in-place (CIP) considering their costs, permeate production, and excess energy cost.
