Keywords
multi-objective optimization; life cycle assessment; sensitivity analysis; ex- pensive optimization; computational sustainability.
Location
Session D10: The Role of Modelling in Sustainable Development
Start Date
11-7-2016 4:50 PM
End Date
11-7-2016 5:10 PM
Abstract
In the frameworks of environmental management and computational sustainability, this paper aims at improving, in a cost-effective manner, the sustainability of existing drinking water production plants (DWPPs) via multi-objective constrained optimization (MOO). Specifically, the paper explores the ability of global sensitivity analysis (SA) methods to support MOO of DWPPs by means of a quick identification of a subset of the most sensitive decision variables from a large set, which leads to better quality solutions under limited time budget. To this end, the paper conducts a comparative analysis for this MOO problem of two major SA methods, namely Morris and Sobol’, both provided by the free software SimLab. The numerical results show that, coupled with the meta-heuristic algorithm SPEA2, both SA methods can properly filter out insensitive decision variables, reducing the full set of decision variables by a factor of 3, and improve the Pareto front approximation quality.
Included in
Civil Engineering Commons, Data Storage Systems Commons, Environmental Engineering Commons, Hydraulic Engineering Commons, Other Civil and Environmental Engineering Commons
Sensitivity Analysis in Support to Drinking Water Production Plants Optimization
Session D10: The Role of Modelling in Sustainable Development
In the frameworks of environmental management and computational sustainability, this paper aims at improving, in a cost-effective manner, the sustainability of existing drinking water production plants (DWPPs) via multi-objective constrained optimization (MOO). Specifically, the paper explores the ability of global sensitivity analysis (SA) methods to support MOO of DWPPs by means of a quick identification of a subset of the most sensitive decision variables from a large set, which leads to better quality solutions under limited time budget. To this end, the paper conducts a comparative analysis for this MOO problem of two major SA methods, namely Morris and Sobol’, both provided by the free software SimLab. The numerical results show that, coupled with the meta-heuristic algorithm SPEA2, both SA methods can properly filter out insensitive decision variables, reducing the full set of decision variables by a factor of 3, and improve the Pareto front approximation quality.
