Subtransmission Grid Control via Online Feedback Optimization

arXiv:2212.07795 · eess.SY, cs.SY · Submitted 2022-12-15 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Robotics Radio. Generated commentary on the latest robotics and control papers.

Rosa: We're looking at the paper: "Subtransmission Grid Control via Online Feedback Optimization".

Dev: Here's the gist of the paper. Online Feedback Optimization (OFO) is a feedback real-time control method that can be employed to enable optimal operation of transmission and subtransmission grids.

Rosa: Such controllers can maximize grid efficiency (e.g., minimizing curtailment) while satisfying grid constraints like voltage and current limits.

Dev: Problem Motivation The increasing electric power consumption and the shift towards renewable energy resources demand for new ways to operate transmission and subtransmission grids.

Rosa: Grid operators have deployed different solutions to tackle this issue.

Title and authors: Rosa: Now we're moving into the title and authors of this paper, "Subtransmission Grid Control via Online Feedback Optimization." The authors are Lukas Ortmanna, Jean Maeght b, Patrick Panciatici b, Florian Dörfler c, and Saverio Bolognani c.

Dev: Those are a good mix of people here. You’ve got the field roboticist perspective from Rosa running trials outside the lab and the controls engineer focus on real-time loops and safety cases.

Taro: I'm curious if any of them have experience with real-world failure modes, like what happens when the world misbehaves, because that's where my testing usually gets complicated.

Rosa: The paper is from a group that clearly has strong ties to power systems research, which is expected given the specific grid benchmark they chose—the real French subtransmission grid.

Dev: That context is important because it means their focus isn't just on abstract theory; they’re grounding this in something tangible, which helps with the safety case side of things.

Taro: I hope this doesn't get too bogged down in synthetic models again, because we want to see how these things actually behave when they hit real-world complexities.

Rosa: They explicitly use a real French subtransmission grid as their benchmark for robustness against model mismatch and constraint satisfaction, which is pretty telling about their practical goals.

Dev: That level of testing against a real system is what separates this from just theoretical papers that don't have to deal with the messy details of physical reality.

The paper's summary: Rosa: So, summarizing what the paper says, they’re presenting the Online Feedback Optimization method as a way to achieve optimal operation of these grids. It’s a feedback control approach designed to maximize grid efficiency by minimizing curtailment while respecting voltage and current limits.

Dev: That boils down to using real-time measurements to drive the controller, updating the inputs based on what you see, which is how they manage the system in real time.

Taro: So it’s about this iterative update process where the inputs are adjusted based on a combination of current state and some estimated sensitivity model.

Rosa: Right. They use an integral form for this update, where the inputs at step k plus one are calculated based on the previous input plus some term that depends on how you're doing now, the disturbance you’ve seen, and their estimate of how much things will change.

Dev: That update involves solving a mixed integer quadratic optimization problem to figure out what the next best move is for the system.

Taro: So it’s not just a simple calculation; it's an iterative decision-making process that incorporates discrete choices, which makes sense for managing those physical components.

Rosa: Right. And they use a tuning matrix G to fine-tune the transient behavior of this closed-loop system to manage how things transition between states.

Dev: That tuning matrix G is crucial because it dictates the rate at which different inputs change and also how much those tap changers are used during those transitions.

The paper's improvements: Taro: So, when we talk about improvements, I see they are really focusing on making this method robust against model mismatch and constraint satisfaction issues.

Rosa: That’s right. They show that this controller is highly precise and not sensitive to model mismatch when compared to the current practice of curtailing injection at a fixed level.

Dev: So, the main improvement here is moving away from fixed curtailment strategies by showing that OFO can extract fifty percent more wind power in the real-world benchmark than what's currently done.

Taro: That fifty percent increase is pretty compelling when you think about how much energy we could save if we could actually achieve those gains.

Rosa: They also showed that the controller can track the time-varying local optimum of the optimization problem, which means it performs practically identically to what you'd get from a perfectly omniscient solver.

Dev: So, even though they acknowledge there are limitations regarding model mismatch, this paper is really pushing for better decision-making accuracy over just relying on fixed rules.

Conclusion: Rosa: Wrapping up the whole thing, the main implication is that this OFO controller can safely operate a subtransmission grid in real time under model mismatch and with limited model information, which is pretty significant.

Dev: It means we don't need perfect information to run a grid optimally; we just need this feedback mechanism to keep it tracking the local optimum as conditions change.

Taro: For my testing perspective, it shows that even when things are unpredictable, the system can recover gracefully and stay in a safe operating point.

Rosa: And they've extended OFO to handle discrete actuators, which is a major step forward for versatile control systems on these grids.

Dev: It’s about giving operators a unified controller for both active and reactive power and those physical switching actions, which simplifies things significantly.

Lukas Ortmanna, Jean Maeght, Patrick Panciatici, Florian Dörfler, Saverio Bolognani

OST – Eastern Switzerland University of Applied Sciences · § Réseau de Transport d’ Electricity (RTE) · Automatic Control Laboratory, ETH Zurich

eess.SY, cs.SY

Submitted: 2022-12-15

Updated: 2025-12-05

Journal ref: Applied Energy, 2026 404, 127142

DOI: 10.1016/j.apenergy.2025.127142

Code: https://github.com/Lukas738/SimulinkMATPOWER

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 87/100

The gist: Online Feedback Optimization (OFO) is a feedback real-time control method that can be employed to enable optimal operation of transmission and subtransmission grids.

Key concepts

Online Feedback Optimization (OFO)
A feedback control method where the controller continuously updates its actions based on real-time grid measurements. It solves a mixed integer quadratic optimization problem at each step to determine the best next move for the grid, ensuring it stays near the ideal operating point.
Mixed Integer Quadratic Optimization Problem (MIQP)
The mathematical problem that OFO solves at every time step. It involves both continuous variables (like power injections) and discrete variables (like whether a tap changer is on or off). Solving this problem helps find the optimal balance between maximizing power output and satisfying all grid constraints.
Model Mismatch Robustness
The ability of the OFO controller to perform well even when its internal model of the grid does not perfectly match reality. The paper shows OFO is highly precise and not overly sensitive to these mismatches, which is crucial for real-world deployment where perfect information is impossible.
Sensitivity ($ abla_u h(u, d)$)
A key piece of model information that describes how much the grid output changes when an input (like power injection) is slightly altered. OFO uses this sensitivity to make informed updates, approximating it with a fixed value for practical, real-time operation.

Terminology

Summary

Online Feedback Optimization (OFO) is a feedback real-time control method that can be employed to enable optimal operation of transmission and subtransmission grids. p. 0

Where the paper says it

Subtransmission Grid Control via Online Feedback Optimization ⋆ Lukas Ortmanna, ∗ , Jean Maeght b , Patrick Panciatici b , Florian Dörfler c , Saverio Bolognani c aOST – Eastern Switzerland University of Applied Sciences, Rapperswil, Switzerland bRéseau de Transport d’Électricité (RTE), Paris, France cAutomatic Control Laboratory, ETH Zurich, Zurich, Switzerland Abstract The increasing electric power consumption and the shift towards renewable energy resources demand for new ways to operate transmission and subtransmission grids. Online Feedback Optimization (OFO) is a feedback real-time control method that can be employed to enable optimal operation of these grids.

Such controllers can maximize grid efficiency (e. p. 0
Where the paper says it

Online Feedback Optimization (OFO) is a feedback real-time control method that can be employed to enable optimal operation of these grids. Such controllers can maximize grid efficiency (e.

g., minimizing curtailment) while satisfying grid constraints like voltage and current limits. p. 0
Where the paper says it

Such controllers can maximize grid efficiency (e. , minimizing curtailment) while satisfying grid constraints like voltage and current limits.

Problem Motivation

The increasing electric power consumption and the shift towards renewable energy resources demand for new ways to operate transmission and subtransmission grids. Grid operators have deployed different solutions to tackle this issue. p. 1

Where the paper says it

In some areas, the grid’s capacity is partly reached, and the uncontrolled generation of renewable power can cause overloaded lines and overvoltages. Grid operators have deployed different solutions to tackle this issue.

The task of real-time efficient operation of the transmission grid can be stated as an online optimization problem. p. 1
Where the paper says it

This opportunity comes from the higher degree of controllability of the grid, given by the flexibility of a fine network of renewable generators connected to almost every bus. The task of real-time efficient operation of the transmission grid can be stated as an online optimization problem.

From a control perspective, one wants to design a feedback control policy (that is, a strategy driven by real-time grid measurements) so that the state of the controlled power grid tracks the optimal state defined as the solution of a nonlinear OPF. p. 1
Where the paper says it

In fact, at every instant in time, the optimal operation of the grid can be defined via a Optimal Power Flow (OPF) problem, whose parameters, however, are time-varying as they depend on the instantaneous loads, the availability of generation, etc. From a control perspective, one wants to design a feedback control policy (that is, a strategy driven by real-time grid measurements) so that the state of the controlled power grid tracks the optimal state defined as the solution of a nonlinear OPF.

Benchmark and Modeling

A novel benchmark is presented and published that corresponds to the real French subtransmission grid on which the proposed controller is analyzed in terms of robustness against model mismatch, constraint satisfaction, and tracking performance. The Unicorn 7019 benchmark represents the real French subtransmission grid. p. 2

Where the paper says it

On the provided benchmark, it is demonstrated how it virtually reinforces the grid (reducing curtailment of renewable energy), and the dynamic tracking performance, the resulting constraint satisfaction, and the robustness to model mismatch. The Unicorn 7019 Benchmark In this section, the Unicorn 7019 benchmark is presented, which represents the real French subtransmission grid.

This grid model consists of 7019 buses, 9657 branches, 1465 generators, and 907 tap changers. p. 2
Where the paper says it

It is implemented in MATPOWER [25] and Matlab Simulink through the toolbox SimulinkMATPOWER [26]. The grid model consists of 7019 buses, 9657 branches, 1465 generators, and 907 tap changers.

The task in the benchmark is to minimize the losses and active power curtailment in the Blocaux area using the active and reactive power injections of the wind farms and the on-load tap changers, while satisfying grid constraints. p. 2
Where the paper says it

The wind generation exceeds the power transfer capacity of the grid, and during the summer of 2021, the wind farms were curtailed at a fixed level to prevent overloaded lines. The task in the benchmark is to minimize the losses and active power curtailment in the Blocaux area using the active and reactive power injections of the wind farms and the on-load tap changers, while satisfying 3

Controller Design Methodology

The OFO controller has an integral form where inputs are iteratively updated as u(k + 1) = u(k) + σ(u(k), d(k), ym(k)). The core of the update involves solving a mixed integer quadratic optimization problem (MIQP). p. 6

Where the paper says it

For example, if the set U is described by the fully general inequality g(u) ≤ 0, then the Taylor expansion g(u + w) ≈ g(u) + ∇g(u) ⊤w immediately yields the desired linearization to be employed in (3): ∇g(u) ⊤ | {z } A (u + w) ≤ −∇g(u) ⊤u − g(u) | {z } b . Problem (3) is a mixed integer quadratic optimization problem (MIQP) that needs to be solved at every time step.

The tuning matrix G is used to adjust the transient behavior of the closed-loop system. p. 6
Where the paper says it

The numerical simulations will show that this seems to be a favorable setting for convergence, even in a dynamic setting. Controller Tuning Besides defining the optimization problem (1), the only parameter in the OFO controller that is to be tuned consists of the tuning matrix G, which is used to adjust the transient behavior of the closed-loop system.

Performance and Results

The OFO controller can help utilize the grid to its full extent, virtually reinforce it, and operate it optimally and in real-time by using the flexibility offered by renewable generators connected to distribution grids. The results show that the OFO controller is tracking the locally optimal solution of the optimization problem defined in Section 4. p. 8

Where the paper says it

Offline Optimization Online Feedback Optimization Without model mismatch Locally optimal solution Same locally optimal solution With model mismatch Constraint violations or suboptimal Constraints are satisfied, converges to the best achievable solution given the model mismatch Computation Solving a computationally intense non-convex problem Calculating an easier update step that is computationally lighter Tracking behavior No tracking (only adjusts the inputs when a new estimate of d becomes available) Tracks the time-varying optimal solution using measurements Communication infrastructure Sending setpoints Sending setpoints & receiving measurements Control strategy Feedforward Feedback Decision basis Model-based Measurement-based Necessary information Model h(u, d), disturbance d, and sensitivity ∇uh(u, d) Sensitivity ∇uh(u, d) Table 1: Comparison of Offline Optimization and Online Feedback Optimization in the first place, and its solution used as set-points for the grid actuators. This potential approach will be referred to as Offline Optimization, as no online information (measurements) is incorporated in the decision.

The simulations show that the OFO controller can optimally solve this non-trivial task of deciding which wind park has to be curtailed by which value at what time while making sure all voltages and lines are within their limits.

Key Findings

The OFO controller is shown to be highly precise and not sensitive to model mismatch when compared to the current practice of curtailing injection at a fixed level. p. 0

Where the paper says it

A novel benchmark is presented and published that corresponds to the real French subtransmission grid on which the proposed controller is analyzed in terms of robustness against model mismatch, constraint satisfaction, and tracking performance. It is shown that OFO controllers can help utilize the grid to its full extent, virtually reinforce it, and operate it optimally and in real-time by using the flexibility offered by renewable generators connected to distribution grids.

The OFO controller can extract 50% more wind power in the real-world benchmark than the current industrial practice. p. 1
Where the paper says it

The current practice for voltage control depends on the country and ranges from manual to automatic control with large sampling times. The fast-changing power injections of renewable energy sources like wind and solar require higher control rates to enforce voltage and current limits in the grid.

Furthermore, simulations on the real benchmark show that OFO can be a powerful tool to safely operate the subtransmission grid in real-time, under model mismatch, and with limited model information. p. 0
Where the paper says it

A novel benchmark is presented and published that corresponds to the real French subtransmission grid on which the proposed controller is analyzed in terms of robustness against model mismatch, constraint satisfaction, and tracking performance. It is shown that OFO controllers can help utilize the grid to its full extent, virtually reinforce it, and operate it optimally and in real-time by using the flexibility offered by renewable generators connected to distribution grids.

The choice of G affects two important features of the grid transient behavior: the rate of change of different inputs and tap changer usage. p. 6
Where the paper says it

Overall, the choice of G affects two important features of the grid transient behavior of the grid. Rate of change of different inputs For simplicity, let G be a diagonal matrix G = diag(g1, g2, .

p. 6
Where the paper says it

Overall, the choice of G affects two important features of the grid transient behavior of the grid. Rate of change of different inputs For simplicity, let G be a diagonal matrix G = diag(g1, g2, .

Necessary Information

OFO is a mostly model-free approach aside from one key piece of model information: the sensitivity ∇uh(u, d) that describes the effect of a change in the input u on the output y. This sensitivity can be computed once initially for a grid state with high power generation and is approximated with this fixed sensitivity. p. 9

Where the paper says it

Namely, these are the sensitivities of the outputs v and ℓ with respect to the inputs q, p, and tap. The sensitivity is calculated once initially for a grid state with high power generation and the true time-varying sensitivity is approximated with this fixed sensitivity.

OFO controllers guarantee constraint satisfaction at steady-state thanks to including repeated measurements of the constrained variables as feedback from the system. p. 8
Where the paper says it

When using Offline Optimization, this suboptimality can yield constraint violations that may not be tolerable by the grid. OFO controllers, instead, guarantee constraint satisfaction at steady-state thanks to including repeated measurements of the constrained variables as feedback from the system.

The suboptimality under model mismatch does not arise from the method but is fundamental to the problem of making decisions based on inaccurate information. p. 8
Where the paper says it

Offline Optimization Online Feedback Optimization Without model mismatch Locally optimal solution Same locally optimal solution With model mismatch Constraint violations or suboptimal Constraints are satisfied, converges to the best achievable solution given the model mismatch Computation Solving a computationally intense non-convex problem Calculating an easier update step that is computationally lighter Tracking behavior No tracking (only adjusts the inputs when a new estimate of d becomes available) Tracks the time-varying optimal solution using measurements Communication infrastructure Sending setpoints Sending setpoints & receiving measurements Control strategy Feedforward Feedback Decision basis Model-based Measurement-based Necessary information Model h(u, d), disturbance d, and sensitivity ∇uh(u, d) Sensitivity ∇uh(u, d) Table 1: Comparison of Offline Optimization and Online Feedback Optimization in the first place, and its solution used as set-points for the grid actuators. This potential approach will be referred to as Offline Optimization, as no online information (measurements) is incorporated in the decision.

Conclusion

By extending OFO controllers with the capability to handle discrete actuators, it is possible to design a real-time controller for a real subtransmission grid benchmark that uses a diverse and complex range of actuators. This controller operates the grid safely at the locally optimal operation point and tracks the local optimum when it changes over time. p. 5

Where the paper says it

y = h(u, d) u ∈ U, y ∈ Y ui ∈ Z ∀ discrete input i (1) At a local optimum (u ⋆ , y⋆ ), the system satisfies the constraints on u and y, which means that it is a feasible operating point. Furthermore, the cost function is locally minimized, which corresponds to minimizing the active power curtailment that is necessary to satisfy line limits.

Most importantly, the simulations on the real benchmark show that OFO can be a powerful tool to safely operate the subtransmission grid in real-time, under model mismatch, and with limited model information. p. 0
Where the paper says it

A novel benchmark is presented and published that corresponds to the real French subtransmission grid on which the proposed controller is analyzed in terms of robustness against model mismatch, constraint satisfaction, and tracking performance. It is shown that OFO controllers can help utilize the grid to its full extent, virtually reinforce it, and operate it optimally and in real-time by using the flexibility offered by renewable generators connected to distribution grids.

A remaining open question is how several independently controlled subtransmission grids would interact with each other. p. 2
Where the paper says it

However, these models lack essential properties of real power grids e. , locally controlled tap changers or independently operating voltagecontrolled generators.

References

[1] “World energy outlook 2021,” IEA, Tech. Rep., 2021.

[3] “Schéma décennal de développement du réseau (sddr) 2019 - rapport complet,” RTE Réseau de transport d’électricité, Tech. p. 0

Where the paper says it

Subtransmission Grid Control via Online Feedback Optimization ⋆ Lukas Ortmanna, ∗ , Jean Maeght b , Patrick Panciatici b , Florian Dörfler c , Saverio Bolognani c aOST – Eastern Switzerland University of Applied Sciences, Rapperswil, Switzerland bRéseau de Transport d’Électricité (RTE), Paris, France cAutomatic Control Laboratory, ETH Zurich, Zurich, Switzerland Abstract The increasing electric power consumption and the shift towards renewable energy resources demand for new ways to operate transmission and subtransmission grids. Online Feedback Optimization (OFO) is a feedback real-time control method that can be employed to enable optimal operation of these grids.

Rep., 2019.

[4] A. Hauswirth, Z. He, S. Bolognani, G. p. 0

Where the paper says it

Subtransmission Grid Control via Online Feedback Optimization ⋆ Lukas Ortmanna, ∗ , Jean Maeght b , Patrick Panciatici b , Florian Dörfler c , Saverio Bolognani c aOST – Eastern Switzerland University of Applied Sciences, Rapperswil, Switzerland bRéseau de Transport d’Électricité (RTE), Paris, France cAutomatic Control Laboratory, ETH Zurich, Zurich, Switzerland Abstract The increasing electric power consumption and the shift towards renewable energy resources demand for new ways to operate transmission and subtransmission grids. Online Feedback Optimization (OFO) is a feedback real-time control method that can be employed to enable optimal operation of these grids.

p. 0
Where the paper says it

Subtransmission Grid Control via Online Feedback Optimization ⋆ Lukas Ortmanna, ∗ , Jean Maeght b , Patrick Panciatici b , Florian Dörfler c , Saverio Bolognani c aOST – Eastern Switzerland University of Applied Sciences, Rapperswil, Switzerland bRéseau de Transport d’Électricité (RTE), Paris, France cAutomatic Control Laboratory, ETH Zurich, Zurich, Switzerland Abstract The increasing electric power consumption and the shift towards renewable energy resources demand for new ways to operate transmission and subtransmission grids. Online Feedback Optimization (OFO) is a feedback real-time control method that can be employed to enable optimal operation of these grids.

Hug, and F. Dörfler, “Optimization algorithms as robust feedback controllers,” Annual Reviews in Control, vol. p. 0
Where the paper says it

Subtransmission Grid Control via Online Feedback Optimization ⋆ Lukas Ortmanna, ∗ , Jean Maeght b , Patrick Panciatici b , Florian Dörfler c , Saverio Bolognani c aOST – Eastern Switzerland University of Applied Sciences, Rapperswil, Switzerland bRéseau de Transport d’Électricité (RTE), Paris, France cAutomatic Control Laboratory, ETH Zurich, Zurich, Switzerland Abstract The increasing electric power consumption and the shift towards renewable energy resources demand for new ways to operate transmission and subtransmission grids. Online Feedback Optimization (OFO) is a feedback real-time control method that can be employed to enable optimal operation of these grids.

57, p. 100941, 2024.

[5] A. Bernstein, E. Dall’Anese, and A. Simonetto, “Online primal-dual methods with measurement feedback for time-varying convex optimization,” IEEE Trans. p. 8

Where the paper says it

Offline Optimization Online Feedback Optimization Without model mismatch Locally optimal solution Same locally optimal solution With model mismatch Constraint violations or suboptimal Constraints are satisfied, converges to the best achievable solution given the model mismatch Computation Solving a computationally intense non-convex problem Calculating an easier update step that is computationally lighter Tracking behavior No tracking (only adjusts the inputs when a new estimate of d becomes available) Tracks the time-varying optimal solution using measurements Communication infrastructure Sending setpoints Sending setpoints & receiving measurements Control strategy Feedforward Feedback Decision basis Model-based Measurement-based Necessary information Model h(u, d), disturbance d, and sensitivity ∇uh(u, d) Sensitivity ∇uh(u, d) Table 1: Comparison of Offline Optimization and Online Feedback Optimization in the first place, and its solution used as set-points for the grid actuators. This potential approach will be referred to as Offline Optimization, as no online information (measurements) is incorporated in the decision.

Signal Process., vol. 67, no. 8, pp. 1978–1991, 2019.

[6] L. S. Lawrence, J. W. Simpson-Porco, and E. Mallada, “Linear-convex optimal steady-state control,” IEEE Transactions on Automatic Control, vol. p. 8

Where the paper says it

Offline Optimization Online Feedback Optimization Without model mismatch Locally optimal solution Same locally optimal solution With model mismatch Constraint violations or suboptimal Constraints are satisfied, converges to the best achievable solution given the model mismatch Computation Solving a computationally intense non-convex problem Calculating an easier update step that is computationally lighter Tracking behavior No tracking (only adjusts the inputs when a new estimate of d becomes available) Tracks the time-varying optimal solution using measurements Communication infrastructure Sending setpoints Sending setpoints & receiving measurements Control strategy Feedforward Feedback Decision basis Model-based Measurement-based Necessary information Model h(u, d), disturbance d, and sensitivity ∇uh(u, d) Sensitivity ∇uh(u, d) Table 1: Comparison of Offline Optimization and Online Feedback Optimization in the first place, and its solution used as set-points for the grid actuators. This potential approach will be referred to as Offline Optimization, as no online information (measurements) is incorporated in the decision.

66, no. 11, pp. 5377–5384, 2020.

[7] M. Colombino, E. Dall’Anese, and A. Bernstein, “Online optimization as a feedback controller: Stability and tracking,” IEEE Transactions on Control of Network Systems, vol. p. 8

Where the paper says it

Offline Optimization Online Feedback Optimization Without model mismatch Locally optimal solution Same locally optimal solution With model mismatch Constraint violations or suboptimal Constraints are satisfied, converges to the best achievable solution given the model mismatch Computation Solving a computationally intense non-convex problem Calculating an easier update step that is computationally lighter Tracking behavior No tracking (only adjusts the inputs when a new estimate of d becomes available) Tracks the time-varying optimal solution using measurements Communication infrastructure Sending setpoints Sending setpoints & receiving measurements Control strategy Feedforward Feedback Decision basis Model-based Measurement-based Necessary information Model h(u, d), disturbance d, and sensitivity ∇uh(u, d) Sensitivity ∇uh(u, d) Table 1: Comparison of Offline Optimization and Online Feedback Optimization in the first place, and its solution used as set-points for the grid actuators. This potential approach will be referred to as Offline Optimization, as no online information (measurements) is incorporated in the decision.

7, no. 1, pp. 422–432, 2019.

[8] G. Bianchin, J. Cortes, J. I. Poveda, and E. Dall’Anese, “Time-varying optimization of lti systems via projected primal-dual gradient flows,” IEEE Transactions on Control of Network Systems, 2021. p. 1

Where the paper says it

This method was specifically developed to drive a dynamical system to the solution of an optimization problem while guaranteeing constraint satisfaction [4, 5, 6, 7, 8, 9]. It needs minimal model information, it responds to time-varying disturbances on the systems, and it was experimentally validated in different small-scale power systems [10, 11, 12].

[9] A. Simonetto, E. Dall’Anese, S. Paternain, G. Leus, and G. B. Giannakis, “Time-varying convex optimization: Time-structured algorithms and applications,” Proceedings of the IEEE, vol. p. 8

Where the paper says it

Offline Optimization Online Feedback Optimization Without model mismatch Locally optimal solution Same locally optimal solution With model mismatch Constraint violations or suboptimal Constraints are satisfied, converges to the best achievable solution given the model mismatch Computation Solving a computationally intense non-convex problem Calculating an easier update step that is computationally lighter Tracking behavior No tracking (only adjusts the inputs when a new estimate of d becomes available) Tracks the time-varying optimal solution using measurements Communication infrastructure Sending setpoints Sending setpoints & receiving measurements Control strategy Feedforward Feedback Decision basis Model-based Measurement-based Necessary information Model h(u, d), disturbance d, and sensitivity ∇uh(u, d) Sensitivity ∇uh(u, d) Table 1: Comparison of Offline Optimization and Online Feedback Optimization in the first place, and its solution used as set-points for the grid actuators. This potential approach will be referred to as Offline Optimization, as no online information (measurements) is incorporated in the decision.

108, no. 11, pp. 2032–2048, 2020.

[10] L. Ortmann, A. p. 0

Where the paper says it

∗Corresponding author Email addresses: lukas. ortmann@ost.

p. 0
Where the paper says it

∗Corresponding author Email addresses: lukas. ortmann@ost.

Hauswirth, I. Caduff, F. Dörfler, and S. p. 0
Where the paper says it

Subtransmission Grid Control via Online Feedback Optimization ⋆ Lukas Ortmanna, ∗ , Jean Maeght b , Patrick Panciatici b , Florian Dörfler c , Saverio Bolognani c aOST – Eastern Switzerland University of Applied Sciences, Rapperswil, Switzerland bRéseau de Transport d’Électricité (RTE), Paris, France cAutomatic Control Laboratory, ETH Zurich, Zurich, Switzerland Abstract The increasing electric power consumption and the shift towards renewable energy resources demand for new ways to operate transmission and subtransmission grids. Online Feedback Optimization (OFO) is a feedback real-time control method that can be employed to enable optimal operation of these grids.

Bolognani, “Experimental validation of feedback optimization in power distribution grids,” Electric Power Systems Research, vol. p. 0
Where the paper says it

Subtransmission Grid Control via Online Feedback Optimization ⋆ Lukas Ortmanna, ∗ , Jean Maeght b , Patrick Panciatici b , Florian Dörfler c , Saverio Bolognani c aOST – Eastern Switzerland University of Applied Sciences, Rapperswil, Switzerland bRéseau de Transport d’Électricité (RTE), Paris, France cAutomatic Control Laboratory, ETH Zurich, Zurich, Switzerland Abstract The increasing electric power consumption and the shift towards renewable energy resources demand for new ways to operate transmission and subtransmission grids. Online Feedback Optimization (OFO) is a feedback real-time control method that can be employed to enable optimal operation of these grids.

189, p. 106782, 2020.

[11] L. Ortmann, A. Prostejovsky, K. Heussen, and S. Bolognani, “Fully distributed peer-to-peer optimal voltage control with minimal model requirements,” Electric Power Systems Research, vol. p. 1

Where the paper says it

It needs minimal model information, it responds to time-varying disturbances on the systems, and it was experimentally validated in different small-scale power systems [10, 11, 12]. Related earlier works also employed feedback controllers to drive a power system to an optimal operation point, but in limited specialized contexts (see [13] 2

189, p. 106717, 2020.

[12] B. Kroposki, A. Bernstein, J. King, and F. Ding, “Good grids make good neighbors,” IEEE Spectrum, vol. p. 0

Where the paper says it

Subtransmission Grid Control via Online Feedback Optimization ⋆ Lukas Ortmanna, ∗ , Jean Maeght b , Patrick Panciatici b , Florian Dörfler c , Saverio Bolognani c aOST – Eastern Switzerland University of Applied Sciences, Rapperswil, Switzerland bRéseau de Transport d’Électricité (RTE), Paris, France cAutomatic Control Laboratory, ETH Zurich, Zurich, Switzerland Abstract The increasing electric power consumption and the shift towards renewable energy resources demand for new ways to operate transmission and subtransmission grids. Online Feedback Optimization (OFO) is a feedback real-time control method that can be employed to enable optimal operation of these grids.

57, no. NREL/JA-5D00-78521, 2020.

[13] D. K. Molzahn, F. Dörfler, H. p. 0

Where the paper says it

Subtransmission Grid Control via Online Feedback Optimization ⋆ Lukas Ortmanna, ∗ , Jean Maeght b , Patrick Panciatici b , Florian Dörfler c , Saverio Bolognani c aOST – Eastern Switzerland University of Applied Sciences, Rapperswil, Switzerland bRéseau de Transport d’Électricité (RTE), Paris, France cAutomatic Control Laboratory, ETH Zurich, Zurich, Switzerland Abstract The increasing electric power consumption and the shift towards renewable energy resources demand for new ways to operate transmission and subtransmission grids. Online Feedback Optimization (OFO) is a feedback real-time control method that can be employed to enable optimal operation of these grids.

Sandberg, S. H. Low, S. Chakrabarti, R. Baldick, and J. Lavaei, “A survey of distributed optimization and control algorithms for electric power systems,” IEEE Transactions on Smart Grid, vol. p. 0
Where the paper says it

Subtransmission Grid Control via Online Feedback Optimization ⋆ Lukas Ortmanna, ∗ , Jean Maeght b , Patrick Panciatici b , Florian Dörfler c , Saverio Bolognani c aOST – Eastern Switzerland University of Applied Sciences, Rapperswil, Switzerland bRéseau de Transport d’Électricité (RTE), Paris, France cAutomatic Control Laboratory, ETH Zurich, Zurich, Switzerland Abstract The increasing electric power consumption and the shift towards renewable energy resources demand for new ways to operate transmission and subtransmission grids. Online Feedback Optimization (OFO) is a feedback real-time control method that can be employed to enable optimal operation of these grids.

8, no., pp., 2941–2962.

[14] S. Bolognani, R. p. 0

Where the paper says it

Subtransmission Grid Control via Online Feedback Optimization ⋆ Lukas Ortmanna, ∗ , Jean Maeght b , Patrick Panciatici b , Florian Dörfler c , Saverio Bolognani c aOST – Eastern Switzerland University of Applied Sciences, Rapperswil, Switzerland bRéseau de Transport d’Électricité (RTE), Paris, France cAutomatic Control Laboratory, ETH Zurich, Zurich, Switzerland Abstract The increasing electric power consumption and the shift towards renewable energy resources demand for new ways to operate transmission and subtransmission grids. Online Feedback Optimization (OFO) is a feedback real-time control method that can be employed to enable optimal operation of these grids.

Carli, G. Cavraro, and S. Zampieri, “Distributed reactive power feedback control for voltage regulation and loss minimization,” IEEE Transactions on Automatic Control, vol. p. 3
Where the paper says it

Measurement devices take voltage magnitude measurements v and apparent power flow measurements ℓ. These are Wind Farms Blocaux Area p(t + 1) Maximum Available Active Power Active Power Set-Points pmax(t) Reactive Power Set-Points q(t + 1) Tap Changers Tap Changer Position Set-Points tap(t + 1) Lines Buses Voltage Magnitude Measurements v(t) `(t) Power Flow Measurements y(t) Online Feedback u(t + 1) Optimization Controller Changing Wind y(t) Figure 2: Block diagram of the control setup with the controller in blue and the grid in gray.

60, no., pp., 966–981, 2015.

[15] S. Li, W. Wu, and Y. Lin, “Robust data-driven and fully distributed volt/var control for active distribution networks with multiple virtual power plants,” IEEE Transactions on Smart Grid, vol. p. 2

Where the paper says it

Most papers analyze the Volt/VAR problem, e. [14, 15, 16, 17] and until now nearly all solutions were applied to distribution grids, e.

13, no., pp., 2627–2638, 2022.

[16] G. Qu and N. Li, “Optimal distributed feedback voltage control under limited reactive power,” IEEE Transactions on Power Systems, vol. p. 1

Where the paper says it

It needs minimal model information, it responds to time-varying disturbances on the systems, and it was experimentally validated in different small-scale power systems [10, 11, 12]. Related earlier works also employed feedback controllers to drive a power system to an optimal operation point, but in limited specialized contexts (see [13] 2

35, no., pp., 315–331, 2019.

[17] H. J. Liu, W. Shi, and H. Zhu, “Distributed voltage control in distribution networks: Online and robust implementations,” IEEE Transactions on Smart Grid, vol. p. 0

Where the paper says it

It is shown that OFO controllers can help utilize the grid to its full extent, virtually reinforce it, and operate it optimally and in real-time by using the flexibility offered by renewable generators connected to distribution grids. Keywords: grid congestion control, online feedback optimization, voltage control, optimal curtailment, renewable energy integration ⋆The research leading to this research has been funded by the Swiss Federal Office of Energy through the project “UNICORN” (SI/501708).

9, no., pp., 6106–6117, 2017.

[18] Y. Guo, X. Zhou, C. Zhao, L. Chen, G. Hug, and T. H. Summers, “An online joint optimization–estimation architecture for distribution networks,” IEEE Transactions on Control Systems Technology, 2023. p. 0

Where the paper says it

It is shown that OFO controllers can help utilize the grid to its full extent, virtually reinforce it, and operate it optimally and in real-time by using the flexibility offered by renewable generators connected to distribution grids. Keywords: grid congestion control, online feedback optimization, voltage control, optimal curtailment, renewable energy integration ⋆The research leading to this research has been funded by the Swiss Federal Office of Energy through the project “UNICORN” (SI/501708).

[19] J. C. Olives-Camps, Á. R. del Nozal, J. M. Mauricio, and J. M. MazaOrtega, “A holistic model-less approach for the optimal real-time control of power electronics-dominated AC microgrids,” Applied Energy, vol. p. 0

Where the paper says it

Subtransmission Grid Control via Online Feedback Optimization ⋆ Lukas Ortmanna, ∗ , Jean Maeght b , Patrick Panciatici b , Florian Dörfler c , Saverio Bolognani c aOST – Eastern Switzerland University of Applied Sciences, Rapperswil, Switzerland bRéseau de Transport d’Électricité (RTE), Paris, France cAutomatic Control Laboratory, ETH Zurich, Zurich, Switzerland Abstract The increasing electric power consumption and the shift towards renewable energy resources demand for new ways to operate transmission and subtransmission grids. Online Feedback Optimization (OFO) is a feedback real-time control method that can be employed to enable optimal operation of these grids.

, 335, p., 120761, 2023.

[20] A. D. Domínguez-García, M. Zholbaryssov, T. Amuda, and O. Ajala, “An online feedback optimization approach to voltage regulation in inverterbased power distribution networks,” in 2023 American Control Conference (ACC), 2023, pp. p. 0

Where the paper says it

It is shown that OFO controllers can help utilize the grid to its full extent, virtually reinforce it, and operate it optimally and in real-time by using the flexibility offered by renewable generators connected to distribution grids. Keywords: grid congestion control, online feedback optimization, voltage control, optimal curtailment, renewable energy integration ⋆The research leading to this research has been funded by the Swiss Federal Office of Energy through the project “UNICORN” (SI/501708).

, 1868–1873.

[21] S. Nowak, Y. C. Chen, and L. Wang, “Measurement-based optimal DER dispatch with a recursively estimated sensitivity model,” IEEE Transactions on Power Systems, vol. p. 8

Where the paper says it

Offline Optimization Online Feedback Optimization Without model mismatch Locally optimal solution Same locally optimal solution With model mismatch Constraint violations or suboptimal Constraints are satisfied, converges to the best achievable solution given the model mismatch Computation Solving a computationally intense non-convex problem Calculating an easier update step that is computationally lighter Tracking behavior No tracking (only adjusts the inputs when a new estimate of d becomes available) Tracks the time-varying optimal solution using measurements Communication infrastructure Sending setpoints Sending setpoints & receiving measurements Control strategy Feedforward Feedback Decision basis Model-based Measurement-based Necessary information Model h(u, d), disturbance d, and sensitivity ∇uh(u, d) Sensitivity ∇uh(u, d) Table 1: Comparison of Offline Optimization and Online Feedback Optimization in the first place, and its solution used as set-points for the grid actuators. This potential approach will be referred to as Offline Optimization, as no online information (measurements) is incorporated in the decision.

35, no., pp., 4792–4802, 2020.

[22] M. Picallo, L. Ortmann, S. p. 0

Where the paper says it

∗Corresponding author Email addresses: lukas. ortmann@ost.

p. 0
Where the paper says it

∗Corresponding author Email addresses: lukas. ortmann@ost.

Bolognani, and F. p. 0
Where the paper says it

Subtransmission Grid Control via Online Feedback Optimization ⋆ Lukas Ortmanna, ∗ , Jean Maeght b , Patrick Panciatici b , Florian Dörfler c , Saverio Bolognani c aOST – Eastern Switzerland University of Applied Sciences, Rapperswil, Switzerland bRéseau de Transport d’Électricité (RTE), Paris, France cAutomatic Control Laboratory, ETH Zurich, Zurich, Switzerland Abstract The increasing electric power consumption and the shift towards renewable energy resources demand for new ways to operate transmission and subtransmission grids. Online Feedback Optimization (OFO) is a feedback real-time control method that can be employed to enable optimal operation of these grids.

p. 0
Where the paper says it

Subtransmission Grid Control via Online Feedback Optimization ⋆ Lukas Ortmanna, ∗ , Jean Maeght b , Patrick Panciatici b , Florian Dörfler c , Saverio Bolognani c aOST – Eastern Switzerland University of Applied Sciences, Rapperswil, Switzerland bRéseau de Transport d’Électricité (RTE), Paris, France cAutomatic Control Laboratory, ETH Zurich, Zurich, Switzerland Abstract The increasing electric power consumption and the shift towards renewable energy resources demand for new ways to operate transmission and subtransmission grids. Online Feedback Optimization (OFO) is a feedback real-time control method that can be employed to enable optimal operation of these grids.

Dörfler, “Adaptive real-time grid operation via online feedback optimization with sensitivity estimation,” Electric Power Systems Research, vol. p. 0
Where the paper says it

Subtransmission Grid Control via Online Feedback Optimization ⋆ Lukas Ortmanna, ∗ , Jean Maeght b , Patrick Panciatici b , Florian Dörfler c , Saverio Bolognani c aOST – Eastern Switzerland University of Applied Sciences, Rapperswil, Switzerland bRéseau de Transport d’Électricité (RTE), Paris, France cAutomatic Control Laboratory, ETH Zurich, Zurich, Switzerland Abstract The increasing electric power consumption and the shift towards renewable energy resources demand for new ways to operate transmission and subtransmission grids. Online Feedback Optimization (OFO) is a feedback real-time control method that can be employed to enable optimal operation of these grids.

, 212, p., 108405, 2022.

[23] Z. Tang, E. Ekomwenrenren, J. W. Simpson-Porco, E. Farantatos, M. Patel, and H. Hooshyar, “Measurement-based fast coordinated voltage control for transmission grids,” IEEE Transactions on Power Systems, vol. p. 4

Where the paper says it

It is assumed that the wind farms also communicate the currently available wind power to the controller, for example based on a wind speed measurement. The controller in the Blocaux area (and in any other subtransmission area) is not tasked to perform frequency control because this is done at the transmission grid level.

, 36, no., pp., 3416–3429, 2020.

[24] Z. Tang, D. J. Hill, and T. Liu, “Distributed coordinated reactive power control for voltage regulation in distribution networks,” IEEE Transactions on Smart Grid, vol. p. 2

Where the paper says it

For a voltage control problem in a distribution grid, the authors of [24] include tap changers in their control decisions. The application of online optimization solutions to power grids is, therefore, mostly limited to voltage control in distribution grids and, with one exception, limited to continuous inputs.

, 12, no., pp., 312–323, 2020.

[25] R. D. Zimmerman, C. E. Murillo-Sánchez, and R. J. Thomas, “Matpower: Steady-state operations, planning, and analysis tools for power systems research and education,” IEEE Transactions on power systems, vol. p. 1

Where the paper says it

This method was specifically developed to drive a dynamical system to the solution of an optimization problem while guaranteeing constraint satisfaction [4, 5, 6, 7, 8, 9]. It needs minimal model information, it responds to time-varying disturbances on the systems, and it was experimentally validated in different small-scale power systems [10, 11, 12].

, 26, no., pp., 12–19, 2010.

[26] L. Ortmann, “Github repository,” https://github. p. 0

Where the paper says it

∗Corresponding author Email addresses: lukas. ortmann@ost.

com/Lukas738/SimulinkMATPOWER, 2023. p. 2
Where the paper says it

The Unicorn 7019 Benchmark In this section, the Unicorn 7019 benchmark is presented, which represents the real French subtransmission grid. It is implemented in MATPOWER [25] and Matlab Simulink through the toolbox SimulinkMATPOWER [26].

[27] ——, “Gitlab repository,” https://gitlab. p. 9

Where the paper says it

This is done in line 4 of the file "OFO_with_integer. m" which can be found on gitlab [27].

ost.ch/lukas.ortmann/subtransmission-grid-control-via-online-feedback-optimization, 2025. p. 0
Where the paper says it

Subtransmission Grid Control via Online Feedback Optimization ⋆ Lukas Ortmanna, ∗ , Jean Maeght b , Patrick Panciatici b , Florian Dörfler c , Saverio Bolognani c aOST – Eastern Switzerland University of Applied Sciences, Rapperswil, Switzerland bRéseau de Transport d’Électricité (RTE), Paris, France cAutomatic Control Laboratory, ETH Zurich, Zurich, Switzerland Abstract The increasing electric power consumption and the shift towards renewable energy resources demand for new ways to operate transmission and subtransmission grids. Online Feedback Optimization (OFO) is a feedback real-time control method that can be employed to enable optimal operation of these grids.

[28] S. Bolognani and F. p. 0

Where the paper says it

Subtransmission Grid Control via Online Feedback Optimization ⋆ Lukas Ortmanna, ∗ , Jean Maeght b , Patrick Panciatici b , Florian Dörfler c , Saverio Bolognani c aOST – Eastern Switzerland University of Applied Sciences, Rapperswil, Switzerland bRéseau de Transport d’Électricité (RTE), Paris, France cAutomatic Control Laboratory, ETH Zurich, Zurich, Switzerland Abstract The increasing electric power consumption and the shift towards renewable energy resources demand for new ways to operate transmission and subtransmission grids. Online Feedback Optimization (OFO) is a feedback real-time control method that can be employed to enable optimal operation of these grids.

Dörfler, “Fast power system analysis via implicit linearization of the power flow manifold,” in 53rd Annual Allerton Conference on Communication, Control, and Computing (Allerton), 2015, pp. p. 0
Where the paper says it

Subtransmission Grid Control via Online Feedback Optimization ⋆ Lukas Ortmanna, ∗ , Jean Maeght b , Patrick Panciatici b , Florian Dörfler c , Saverio Bolognani c aOST – Eastern Switzerland University of Applied Sciences, Rapperswil, Switzerland bRéseau de Transport d’Électricité (RTE), Paris, France cAutomatic Control Laboratory, ETH Zurich, Zurich, Switzerland Abstract The increasing electric power consumption and the shift towards renewable energy resources demand for new ways to operate transmission and subtransmission grids. Online Feedback Optimization (OFO) is a feedback real-time control method that can be employed to enable optimal operation of these grids.

, 402–409.

[29] V. Häberle, A. Hauswirth, L. Ortmann, S. Bolognani, and F. Dörfler, “Nonconvex feedback optimization with input and output constraints,” IEEE Control Systems Letters, vol. p. 4

Where the paper says it

Online Feedback Optimization with Integer Constraints This section, provides a mathematical formulation of the control design problem, derives the OFO controller, and comments on the necessary model information. Input, Output, and Optimization Problem The input u = [q ⊤, p⊤, tap⊤] ⊤ consists of the reactive power setpoints q ∈ R 42 and active power setpoints p ∈ R 42 for the 42 wind parks, and the tap positions tap ∈ Z 10 for the 10 tap changers within the Blocaux area, for a total of 94 signals.

, 5, no., pp., 343–348, 2020.

[30] J. Nocedal and S. J. Wright, Numerical optimization, ser. p. 6

Where the paper says it

The numerical simulations will show that this seems to be a favorable setting for convergence, even in a dynamic setting. Controller Tuning Besides defining the optimization problem (1), the only parameter in the OFO controller that is to be tuned consists of the tuning matrix G, which is used to adjust the transient behavior of the closed-loop system.

Springer series in operations research. p. 0
Where the paper says it

It is shown that OFO controllers can help utilize the grid to its full extent, virtually reinforce it, and operate it optimally and in real-time by using the flexibility offered by renewable generators connected to distribution grids. Keywords: grid congestion control, online feedback optimization, voltage control, optimal curtailment, renewable energy integration ⋆The research leading to this research has been funded by the Swiss Federal Office of Energy through the project “UNICORN” (SI/501708).

New York: Springer, 1999. p. 0
Where the paper says it

Subtransmission Grid Control via Online Feedback Optimization ⋆ Lukas Ortmanna, ∗ , Jean Maeght b , Patrick Panciatici b , Florian Dörfler c , Saverio Bolognani c aOST – Eastern Switzerland University of Applied Sciences, Rapperswil, Switzerland bRéseau de Transport d’Électricité (RTE), Paris, France cAutomatic Control Laboratory, ETH Zurich, Zurich, Switzerland Abstract The increasing electric power consumption and the shift towards renewable energy resources demand for new ways to operate transmission and subtransmission grids. Online Feedback Optimization (OFO) is a feedback real-time control method that can be employed to enable optimal operation of these grids.

[31] M. Colombino, J. W. Simpson-Porco, and A. Bernstein, “Towards robustness guarantees for feedback-based optimization,” in 58th Conference on Decision and Control (CDC), 2019, pp. p. 8

Where the paper says it

Offline Optimization Online Feedback Optimization Without model mismatch Locally optimal solution Same locally optimal solution With model mismatch Constraint violations or suboptimal Constraints are satisfied, converges to the best achievable solution given the model mismatch Computation Solving a computationally intense non-convex problem Calculating an easier update step that is computationally lighter Tracking behavior No tracking (only adjusts the inputs when a new estimate of d becomes available) Tracks the time-varying optimal solution using measurements Communication infrastructure Sending setpoints Sending setpoints & receiving measurements Control strategy Feedforward Feedback Decision basis Model-based Measurement-based Necessary information Model h(u, d), disturbance d, and sensitivity ∇uh(u, d) Sensitivity ∇uh(u, d) Table 1: Comparison of Offline Optimization and Online Feedback Optimization in the first place, and its solution used as set-points for the grid actuators. This potential approach will be referred to as Offline Optimization, as no online information (measurements) is incorporated in the decision.

, 6207–6214.

[32] W. Chan, Z. He, K. Moffat, S. Bolognani, M. p. 0

Where the paper says it

Subtransmission Grid Control via Online Feedback Optimization ⋆ Lukas Ortmanna, ∗ , Jean Maeght b , Patrick Panciatici b , Florian Dörfler c , Saverio Bolognani c aOST – Eastern Switzerland University of Applied Sciences, Rapperswil, Switzerland bRéseau de Transport d’Électricité (RTE), Paris, France cAutomatic Control Laboratory, ETH Zurich, Zurich, Switzerland Abstract The increasing electric power consumption and the shift towards renewable energy resources demand for new ways to operate transmission and subtransmission grids. Online Feedback Optimization (OFO) is a feedback real-time control method that can be employed to enable optimal operation of these grids.

Muehlebach, and F. Dörfler, “Robust feedback optimization with model uncertainty: A regularization approach,” arXiv:2503. p. 8
Where the paper says it

Offline Optimization Online Feedback Optimization Without model mismatch Locally optimal solution Same locally optimal solution With model mismatch Constraint violations or suboptimal Constraints are satisfied, converges to the best achievable solution given the model mismatch Computation Solving a computationally intense non-convex problem Calculating an easier update step that is computationally lighter Tracking behavior No tracking (only adjusts the inputs when a new estimate of d becomes available) Tracks the time-varying optimal solution using measurements Communication infrastructure Sending setpoints Sending setpoints & receiving measurements Control strategy Feedforward Feedback Decision basis Model-based Measurement-based Necessary information Model h(u, d), disturbance d, and sensitivity ∇uh(u, d) Sensitivity ∇uh(u, d) Table 1: Comparison of Offline Optimization and Online Feedback Optimization in the first place, and its solution used as set-points for the grid actuators. This potential approach will be referred to as Offline Optimization, as no online information (measurements) is incorporated in the decision.

24151 [math.OC], 2025.

[33] Z. He, S. Bolognani, J. p. 0

Where the paper says it

Subtransmission Grid Control via Online Feedback Optimization ⋆ Lukas Ortmanna, ∗ , Jean Maeght b , Patrick Panciatici b , Florian Dörfler c , Saverio Bolognani c aOST – Eastern Switzerland University of Applied Sciences, Rapperswil, Switzerland bRéseau de Transport d’Électricité (RTE), Paris, France cAutomatic Control Laboratory, ETH Zurich, Zurich, Switzerland Abstract The increasing electric power consumption and the shift towards renewable energy resources demand for new ways to operate transmission and subtransmission grids. Online Feedback Optimization (OFO) is a feedback real-time control method that can be employed to enable optimal operation of these grids.

He, F. Dörfler, and X. p. 0
Where the paper says it

Subtransmission Grid Control via Online Feedback Optimization ⋆ Lukas Ortmanna, ∗ , Jean Maeght b , Patrick Panciatici b , Florian Dörfler c , Saverio Bolognani c aOST – Eastern Switzerland University of Applied Sciences, Rapperswil, Switzerland bRéseau de Transport d’Électricité (RTE), Paris, France cAutomatic Control Laboratory, ETH Zurich, Zurich, Switzerland Abstract The increasing electric power consumption and the shift towards renewable energy resources demand for new ways to operate transmission and subtransmission grids. Online Feedback Optimization (OFO) is a feedback real-time control method that can be employed to enable optimal operation of these grids.

Guan, “Model-free nonlinear feedback optimization,” IEEE Transactions on Automatic Control, vol. p. 6
Where the paper says it

It is important to notice that (3) does not contain the nonlinear power flow equations that model the grid, and it is therefore significantly easier to solve than an OPF like (1). The key idea of the OFO controller is to repeatedly solve a simpler optimization program that incorporates the real-time feedback measurements from the system, knowing that the repeated closed-loop application of these updates will drive the system to the solution of (1) (see [4] for a review of the formal convergence guarantees).

, 69, no., pp., 4554–4569, 2024.

[34] J. Lofberg, “Yalmip: A toolbox for modeling and optimization in matlab,” in IEEE International Conference on Robotics and Automation, 2004, pp. p. 2

Where the paper says it

The Unicorn 7019 Benchmark In this section, the Unicorn 7019 benchmark is presented, which represents the real French subtransmission grid. It is implemented in MATPOWER [25] and Matlab Simulink through the toolbox SimulinkMATPOWER [26].

, 284–289.

[35] G. Belgioioso, S. Bolognani, G. Pejrani, and F. Dörfler, “Tutorial on congestion control in multi-area transmission grids via online feedback equilibrium seeking,” in 62nd IEEE Conference on Decision and Control (CDC), 2023. p. 0

Where the paper says it

Subtransmission Grid Control via Online Feedback Optimization ⋆ Lukas Ortmanna, ∗ , Jean Maeght b , Patrick Panciatici b , Florian Dörfler c , Saverio Bolognani c aOST – Eastern Switzerland University of Applied Sciences, Rapperswil, Switzerland bRéseau de Transport d’Électricité (RTE), Paris, France cAutomatic Control Laboratory, ETH Zurich, Zurich, Switzerland Abstract The increasing electric power consumption and the shift towards renewable energy resources demand for new ways to operate transmission and subtransmission grids. Online Feedback Optimization (OFO) is a feedback real-time control method that can be employed to enable optimal operation of these grids.

How the training signal is built

The cost function encodes the control goal, i.e., to minimize curtailment. p. 2

Where the paper says it

The wind generation exceeds the power transfer capacity of the grid, and during the summer of 2021, the wind farms were curtailed at a fixed level to prevent overloaded lines. The task in the benchmark is to minimize the losses and active power curtailment in the Blocaux area using the active and reactive power injections of the wind farms and the on-load tap changers, while satisfying 3

To design an appropriate regularization of the problem at hand, two additional terms besides the cost of curtailment are included. p. 5
Where the paper says it

Furthermore, the cost function is locally minimized, which corresponds to minimizing the active power curtailment that is necessary to satisfy line limits. Implicitly, (1) defines a timevarying reference that the grid needs to track, as opposed as standard control design tasks where the reference to be tracked is set a-priori, or fixed.

The term-sum(p), when minimized, promoted maximal power production (minimal curtailment). p. 5
Where the paper says it

y = h(u, d) u ∈ U, y ∈ Y ui ∈ Z ∀ discrete input i (1) At a local optimum (u ⋆ , y⋆ ), the system satisfies the constraints on u and y, which means that it is a feasible operating point. Furthermore, the cost function is locally minimized, which corresponds to minimizing the active power curtailment that is necessary to satisfy line limits.

The term wq∥q∥2 can be interpreted as an approximation of the additional losses due to reactive power flows on the medium voltage lines connecting the wind farms to the high voltage grid. p. 3
Where the paper says it

Measurement devices take voltage magnitude measurements v and apparent power flow measurements ℓ. These are Wind Farms Blocaux Area p(t + 1) Maximum Available Active Power Active Power Set-Points pmax(t) Reactive Power Set-Points q(t + 1) Tap Changers Tap Changer Position Set-Points tap(t + 1) Lines Buses Voltage Magnitude Measurements v(t) `(t) Power Flow Measurements y(t) Online Feedback u(t + 1) Optimization Controller Changing Wind y(t) Figure 2: Block diagram of the control setup with the controller in blue and the grid in gray.

The term w⊤taptap is a linear approximation of the effect of the tap position on the grid losses (which are reduced for higher tap positions). p. 9
Where the paper says it

This matrix C and vector c are used in the controller in (3). The tap changers can take 33 discrete positions, and the change of a tap position has to be an integer.

The gradient of our cost function is ∇f(u, y) = ∇uf(u, y) ∇yf(u, y) ≈ [2wqq-142 wtap 089]", where 142 is a vector of ones and 089 is a vector of zeros of dimension 42 and 89, respectively.

Where the gains come from

The OFO controller can optimally solve this non-trivial task of deciding which wind park has to be curtailed by which value at what time while making sure all voltages and lines are within their limits. The results show that the OFO controller can extract 50% more wind power in the real-world benchmark than the current industrial practice. p. 1

Where the paper says it

The current practice for voltage control depends on the country and ranges from manual to automatic control with large sampling times. The fast-changing power injections of renewable energy sources like wind and solar require higher control rates to enforce voltage and current limits in the grid.

The OFO controller is shown to be highly precise and not sensitive to model mismatch when compared to the current practice of curtailing injection at a fixed level. The OFO controller guarantees constraint satisfaction at steady-state thanks to including repeated measurements of the constrained variables as feedback from the system. p. 8
Where the paper says it

When using Offline Optimization, this suboptimality can yield constraint violations that may not be tolerable by the grid. OFO controllers, instead, guarantee constraint satisfaction at steady-state thanks to including repeated measurements of the constrained variables as feedback from the system.

The tracking performance is shown to be practically identical to the theoretical optimum that was computed with an omniscient and instantaneous OPF solver. p. 0
Where the paper says it

A novel benchmark is presented and published that corresponds to the real French subtransmission grid on which the proposed controller is analyzed in terms of robustness against model mismatch, constraint satisfaction, and tracking performance. It is shown that OFO controllers can help utilize the grid to its full extent, virtually reinforce it, and operate it optimally and in real-time by using the flexibility offered by renewable generators connected to distribution grids.

The OFO controller can operate the grid safely at the locally optimal operation point and tracks the local optimum when it changes over time. p. 0
Where the paper says it

Subtransmission Grid Control via Online Feedback Optimization ⋆ Lukas Ortmanna, ∗ , Jean Maeght b , Patrick Panciatici b , Florian Dörfler c , Saverio Bolognani c aOST – Eastern Switzerland University of Applied Sciences, Rapperswil, Switzerland bRéseau de Transport d’Électricité (RTE), Paris, France cAutomatic Control Laboratory, ETH Zurich, Zurich, Switzerland Abstract The increasing electric power consumption and the shift towards renewable energy resources demand for new ways to operate transmission and subtransmission grids. Online Feedback Optimization (OFO) is a feedback real-time control method that can be employed to enable optimal operation of these grids.

Tuning Influence

The tuning matrix G influences two important features of the grid transient behavior: the rate of change of different inputs and tap changer usage. The bigger an entry gi of G, the less aggressively the corresponding input ui will be changed at each time step during the transient. p. 6

Where the paper says it

Rate of change of different inputs For simplicity, let G be a diagonal matrix G = diag(g1, g2, . The bigger an entry gi of G, the less aggressively the corresponding input ui will be changed at each time step during the transient.

The higher values in G corresponding to a discrete actuator like tap changers, the less they are used. p. 7
Where the paper says it

While G does not affect the constraint enforcement it does affect how often discrete inputs are used. The higher the values in G corresponding to a discrete actuator like tap changers, the less they are used.

With a value of 5000, they are only used once the grid reaches its capacity limit. p. 1
Where the paper says it

A common approach to prevent overloaded lines is to curtail the renewable generation to a fixed maximum value, depending on the seasonal thermal ratings of the lines, without taking consumption and other generation into account. For example, these actions currently need to be implemented in the Blocaux area in France which will be used as a benchmark in this paper.

The choice of G affects two important features of the grid transient behavior: the rate of change of different inputs and tap changer usage. The results for different values can be seen in Figure 9. p. 7
Where the paper says it

The higher the values in G corresponding to a discrete actuator like tap changers, the less they are used. Numerical results will be shown in Section 5.

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Improve real-time grid operation by enabling controllers to maximize grid efficiency (e.g., minimizing curtailment) while satisfying grid constraints like voltage and current limits using Online Feedback Optimization (OFO).

  2. Enable virtual reinforcement of the transmission grid by using automatic control to mitigate the need to physically extend the grid, potentially saving billions of Euros over ten years through real-time control of active power flows instead of building new lines.

  3. Allow controllers to handle discrete actuators, such as tap changers, by extending OFO with capabilities for discrete inputs, enabling a unified controller that manages both active and reactive power injections and tap changer operations.

  4. Improve robustness against model mismatch by demonstrating that OFO controllers can maintain robustness to model mismatch even on a real grid benchmark, which is essential for deployment where perfect model information is unavailable.

  5. Enhance decision-making accuracy by allowing the controller to track the time-varying local optimum of the optimization problem, achieving performance practically identical to a theoretical optimum computed with an omniscient solver.

  6. Increase renewable energy utilization by showing that OFO controllers can extract 50% more wind power in the real-world benchmark than the current industrial practice.

  7. Provide versatile control objectives by allowing specifications in an optimization problem to easily include grid ancillary services like voltage control, demonstrated through Task 2 where voltage support is incorporated.

Abstract

The increasing electric power consumption and the shift towards renewable energy resources demand for new ways to operate transmission and subtransmission grids. Online Feedback Optimization (OFO) is a feedback real-time control method that can be employed to enable optimal operation of these grids. Such controllers can maximize grid efficiency (e.g., minimizing curtailment) while satisfying grid constraints like voltage and current limits. The OFO control method is tailored and extended to handle discrete inputs and it is explained how to design an OFO controller for the subtransmission grid. A novel benchmark is presented and published that corresponds to the real French subtransmission grid on which the proposed controller is analyzed in terms of robustness against model mismatch, constraint satisfaction, and tracking performance. It is shown that OFO controllers can help utilize the grid to its full extent, virtually reinforce it, and operate it optimally and in real-time by using the flexibility offered by renewable generators connected to distribution grids.

Sources

Related papers