This study aims to minimize the environmental impacts of thermal seawater desalination by optimizing the required fossil fuel mixture. Life cycle assessment (LCA) is applied to simulate the environmental impacts for each fuel mixture. To prevent mixture designs inherited collinearly from correlating LCA results, fuel mixtures are first sampled prior to conducting LCA and then later optimized using a regression-based methodology to reduce entailed environmental impacts. Setting the functional unit to 1 m3 of desalinated water induces different reference flows of energy requirements depending on the fuels used. Increasing the level of any fuel type within the fuel mixture scenario will cause a decrease in the level of the other fuel type(s) included. An augmented simplex lattice mixture (ASLM) design is applied to indicate correct experimental conditions and to prevent the correlation due to collinearity inherited from the nature of mixture problems. Regression models are formulated to represent life cycle impact assessment (LCIA) results in a closed form suitable for response surface methodology (RSM) optimization. An overall composite sustainability index (CSI) is a single index calculated by aggregating and normalizing corresponding LCIA responses of different units, ranges, and scales using the geometric mean-based method. The results indicate that marine sediment ecotoxicity (MSE) is the category most adversely affected by multistage flash distillation (MSF). On a nationwide scale, the LCA optimized results scored a 17% reduction in associated environmental impacts, which corresponds to a 4.2% reduction in the county’s carbon footprint and a 62% reduction in MSE while incurring a minor retrofitting cost to desalination facilities. High MSE results due to excessive fossil fuel consumption/burning in MSF should gain as much attention as paid toward global warming potential. High MSE entails the risk of having heavy metals entering the food chain. On the other hand, the geometric mean approach is found to be an effective model to aggregate the LCIA results into a single index while avoiding the subjectivity of the value judgment used in LCIA weighting. This approach serves as a unit-free rescaling method that is robust to outliers or large values examined across different LCIA impacts.
Read full abstract