Register now After registration you will be able to apply for this opportunity online.
Robustify Feedback Optimization through Regularization
Optimal steady-state operations are crucial for engineering systems. A promising paradigm called feedback optimization (FO) features autonomous optimality seeking with a minimal requirement on model information, i.e., the input-output sensitivity. In applications, however, uncertainties (e.g., random failures and parameter shifts) may cause a model mismatch, thus resulting in closed-loop sub-optimality. To address this critical issue, we will explore robustifying FO against structured model mismatch through regularization. To this end, we will formulate a min-max closed-loop optimization problem and solve the reformulated regularized problem in an online fashion. We will characterize the optimality and stability of the closed-loop behavior. Furthermore, we will numerically validate the effectiveness of the proposed algorithm.
Keywords: Feedback optimization, robust optimization,power system
Please see the attached file.
Please see the attached file.
The goals of this project are as follows.
- Learn about feedback optimization and robust optimization;
- Design a feedback optimization algorithm that is robust against model mismatch and random disturbances;
- Characterize the theoretical performance (e.g., optimality and stability) of the closed-loop interconnection of the algorithm and the plant;
- Conduct numerical experiments to illustrate the performance of the algorithm.
**Publications**: If the final results are promising, we will aim for publication in top-tier control conferences or journals.
The goals of this project are as follows.
- Learn about feedback optimization and robust optimization;
- Design a feedback optimization algorithm that is robust against model mismatch and random disturbances; - Characterize the theoretical performance (e.g., optimality and stability) of the closed-loop interconnection of the algorithm and the plant; - Conduct numerical experiments to illustrate the performance of the algorithm.
**Publications**: If the final results are promising, we will aim for publication in top-tier control conferences or journals.
Please send your resume/CV (including lists of relevant publications/projects) and transcript of records in PDF format via email to Zhiyu He (zhiyhe@ethz.ch), Dr. Keith Moffat (kmoffat@ethz.ch), and Dr. Saverio Bolognani (bsaverio@ethz.ch). We look forward to your application.
Please send your resume/CV (including lists of relevant publications/projects) and transcript of records in PDF format via email to Zhiyu He (zhiyhe@ethz.ch), Dr. Keith Moffat (kmoffat@ethz.ch), and Dr. Saverio Bolognani (bsaverio@ethz.ch). We look forward to your application.