Optimal scenario reduction framework based on distance of uncertainty distribution and output performance: II. Sequential reduction

• Novel method for reduction of a huge number of scenarios generated from the factorial combination.
• Sequential reduction framework greatly reduces the computational complexity.
• Scenario and robust optimization based criteria for quantifying the quality of reduction.

In this paper, a novel sequential scenario reduction framework for general optimization problem is proposed. The proposed method extends the previous work (Li and Floudas, 2014) and aims to tackle optimization problems with a large number of uncertain parameters and a huge number of scenarios generated from the factorial combination. The proposed method first ranks the uncertain parameters based on their effects on the optimal objective using global sensitivity analysis. Then, the parameters are sequentially considered in generating uncertainty scenarios. This method can essentially reduce the computational efforts needed for evaluating the objective values of all scenarios, which is often impractical for a huge number of scenarios. Criteria for quantifying the quality of scenario reduction are also proposed based on robust optimization and scenario optimization. Case studies are presented to illustrate the sequential scenario reduction framework and the results verify the efficiency of the proposed approach.