| 摘要 |
Assume a decision strategy that is optimal for a primary objective function(s). Without unacceptably affecting the value of the primary objective function(s), Robustness Enhancing Optimizer (REO) modifies the optimal strategy's decision variable values to maximize the likelihood that implementing the new strategy in the real world will be satisfactory (will satisfy optimization problem constraints in the real world). Modification is via linked system parameter sensitivity analysis and multi-realization optimization. REO creates new realizations by changing system parameter(s) in a selected manner (for example, multiplying by a selected factor smaller or larger than 1 proportionally reduces or increases the parameter). REO is especially valuable where probability density functions or statistically-based realizations are unavailable, or where one wants to increase the robustness of satisfying constraints for multiple realities in a particular setting.
|
