Abstract:This work uniquely combines an affine linear decision rule known from adjustable robustness with min-max-regret robustness. By doing so, the advantages of both concepts can be obtained with an adjustable solution that is not over-conservative. This combination results in a bilevel optimization problem. For solving this problem, a three-stage algorithm which uses adaptive discretization of the uncertainty set via two criteria is presented and its convergence is proven. The algorithm is applicable for an example of optimizing a robust pump operation plan for a drinking water supply system facing uncertain demand. The algorithm shows a notable ability to scale, presenting an opportunity to solve larger instances that might challenge existing optimization approaches.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is how to combine adjustable robustness and min - max - regret robustness to deal with uncertainty optimization problems. Specifically, the paper proposes a new optimization framework, which combines affine linear decision rules with the min - max - regret method to obtain more flexible and not overly conservative solutions.
### Problem Background
In many practical applications, optimization problems are usually accompanied by uncertainty. Although traditional robust optimization methods can handle optimization problems under uncertain conditions, they are often too conservative, resulting in a decline in the quality of solutions. For this reason, researchers have proposed two main robust optimization methods: the min - max - regret method and the adjustable robustness method.
- **Min - Max - Regret Method**: Aims to minimize the regret value in the worst - case scenario, that is, the gap between the decision result and the ex - post optimal solution.
- **Adjustable Robustness Method**: Allows some decisions to be adjusted after the uncertainty part is revealed, usually achieved through affine linear decision rules.
### Core Contributions of the Paper
The core contribution of the paper lies in combining these two methods to form a new optimization problem and proposing an adaptive three - stage algorithm to solve this problem. Specifically:
1. **Problem Description**:
- Replace the decision variable with the affine linear decision rule \(x=\pi_0 + \Pi u\), where \(\pi_0\) and \(\Pi\) are parameters to be optimized.
- Define an optimization problem containing uncertainty, with the goal of minimizing the maximum regret value.
2. **Algorithm Design**:
- Propose an adaptive three - stage algorithm, which approximates the global optimal solution by gradually increasing the discretized scenario set.
- In the first stage of the algorithm, solve the discretized min - max - regret problem, in the second stage, check the feasibility of the solution, and in the third stage, evaluate the optimality of the solution.
3. **Convergence Proof**:
- Prove that the algorithm can converge to the global optimal solution under certain conditions.
- Give the error estimate between the approximate solution and the true solution.
### Application Example
The paper takes optimizing the pump operation plan in the drinking water supply system as an example to show the effectiveness and scalability of this algorithm. Through this example, the author proves that the algorithm can provide a robust optimization scheme in the face of uncertain demand.
### Summary
This paper proposes a new robust optimization framework by combining affine linear decision rules and the min - max - regret method, and develops an effective adaptive three - stage algorithm to solve such problems. This method not only improves the flexibility of solutions but also reduces conservatism, thus performing well in dealing with uncertainty optimization problems.