TY - JOUR
T1 - Approach to robust multi-objective optimization and probabilistic analysis
T2 - The ROPAR algorithm
AU - Marquez-Calvo, Oscar O.
AU - Solomatine, Dimitri P.
PY - 2019
Y1 - 2019
N2 - This paper considers the problem of robust optimization, and presents the technique called Robust Optimization and Probabilistic Analysis of Robustness (ROPAR). It has been developed for finding robust optimum solutions of a particular class in model-based multi-objective optimization (MOO) problems (i.e. when the objective function is not known analytically), where some of the parameters or inputs to this model are assumed to be uncertain. A Monte Carlo simulation framework is used. It can be straightforwardly implemented in a distributed computing environment which allows the results to be obtained relatively fast. The technique is exemplified in the two case studies: (a) a benchmark problem commonly used to test MOO algorithms (a version of the ZDT1 function); and (b) a design problem of a simple storm drainage system, where the uncertainty is associated with design rainfall events. It is shown that the design found by ROPAR can adequately cope with these uncertainties. The approach can be useful for assisting in a wide range of risk-based decisions.
AB - This paper considers the problem of robust optimization, and presents the technique called Robust Optimization and Probabilistic Analysis of Robustness (ROPAR). It has been developed for finding robust optimum solutions of a particular class in model-based multi-objective optimization (MOO) problems (i.e. when the objective function is not known analytically), where some of the parameters or inputs to this model are assumed to be uncertain. A Monte Carlo simulation framework is used. It can be straightforwardly implemented in a distributed computing environment which allows the results to be obtained relatively fast. The technique is exemplified in the two case studies: (a) a benchmark problem commonly used to test MOO algorithms (a version of the ZDT1 function); and (b) a design problem of a simple storm drainage system, where the uncertainty is associated with design rainfall events. It is shown that the design found by ROPAR can adequately cope with these uncertainties. The approach can be useful for assisting in a wide range of risk-based decisions.
KW - Multi-objective optimization
KW - Robust optimization
KW - Uncertainty
KW - drainage system
UR - http://www.scopus.com/inward/record.url?scp=85064548851&partnerID=8YFLogxK
U2 - 10.2166/hydro.2019.095
DO - 10.2166/hydro.2019.095
M3 - Article
AN - SCOPUS:85064548851
SN - 1464-7141
VL - 21
SP - 427
EP - 440
JO - Journal of Hydroinformatics
JF - Journal of Hydroinformatics
IS - 3
ER -