Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Online optimization in closed loop
Feedback-based optimization is a powerful new paradigm that combines ideas from control and optimization and is highly relevant for the real-time optimization of power grids. In our projects, we offer students the opportunity to contribute to our state-of-the-art research.
Keywords: optimization, control theory, power systems,
Feedback-based optimization is a powerful new paradigm that combines idea from control and optimization and is highly relevant for the real-time optimization of power grids.
In simple terms, we take established optimization algorithms such as gradient descent and implement them as nonlinear feedback controllers. The resulting closed-loop system not only converges an optimizer. But, through the feedback loop, it can track the optimizer and is more robust than the traditional "feedforward" paradigm.
The focus of our student projects varies continuously depending on the state of our research and the student's interest and background.
Current topics involve:
- singular perturbation analysis of dynamic feedback optimization schemes
- guaranteed tracking performance in time-varying online optimization
- robustness of feedback optimization towards model mismatch
Feedback-based optimization is a powerful new paradigm that combines idea from control and optimization and is highly relevant for the real-time optimization of power grids.
In simple terms, we take established optimization algorithms such as gradient descent and implement them as nonlinear feedback controllers. The resulting closed-loop system not only converges an optimizer. But, through the feedback loop, it can track the optimizer and is more robust than the traditional "feedforward" paradigm.
The focus of our student projects varies continuously depending on the state of our research and the student's interest and background.
Current topics involve:
- singular perturbation analysis of dynamic feedback optimization schemes
- guaranteed tracking performance in time-varying online optimization
- robustness of feedback optimization towards model mismatch
In our student projects, we try to explore meaningful applications and/or reinforce the theoretical foundations of our methods.
The skills that students acquire in the process include knowledge about (non-convex) optimization, (nonlinear) control,
dynamical systems, and differential geometry.
Students are generally expected to a solid background in optimization and control and display mathematical maturity.
In our student projects, we try to explore meaningful applications and/or reinforce the theoretical foundations of our methods.
The skills that students acquire in the process include knowledge about (non-convex) optimization, (nonlinear) control, dynamical systems, and differential geometry.
Students are generally expected to a solid background in optimization and control and display mathematical maturity.
Adrian Hauswirth (hadrian@ethz.ch)
Saverio Bolognani (bsaverio@ethz.ch
Adrian Hauswirth (hadrian@ethz.ch) Saverio Bolognani (bsaverio@ethz.ch