Real-time Curative Actions for Power Systems via Online Feedback Optimization
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: "Real-time Curative Actions for Power Systems via Online Feedback Optimization".
Dev: Here's the gist of the paper. Real-time Curative Actions for Power Systems via Online Feedback Optimization proposes calculating complex curative actions in real-time after a contingency using Online Feedback Optimization (OFO) to enhance N-1 security Need for Real-Time Control Curative or remedial actions are the set of immediate actions intended to bring the power grid to a safe operating point after a contingency Due to the shift from steady to volatile energy resources, the grid state will frequently change and the curative actions would need to be pre-planned increasingly often Furthermore, with the shift from large bulk production to many small decentralized energy sources more devices need to be actuated simultaneously to achieve the same outcome Overall, operators will need to determine curative actions more often, and those actions will be more complicated.
Rosa: The proposed control strategy is based on OFO, a methodology that allows converting iterative optimization algorithms into real-time robust feedback controllers These controllers can then be used to drive a system to the optimum of a constrained optimization problem which defines the safe operating region of the grid.
Dev: The control update is given by pOF O(t + 1) vOF O(t + 1) v(t) (t)ym(t).
Rosa: The controller iteratively gathers power system measurements and updates setpoints for the controllable power plants in the grid.
Title and authors: Dev: Let's talk about who wrote this, Rosa and Dev have been digging into "Real-time Curative Actions for Power Systems via Online Feedback Optimization." The authors are Ortmann, Hotz, Bolognani, and Dorfler.
Rosa: I'm curious what the title says about the main problem they are trying to solve here. It's about getting those immediate actions right after a failure.
Dev: They explain that because the energy world is shifting from steady sources to lots of small, decentralized ones, pre-planning these actions is becoming way too hard and frequent.
Taro: So it's not just about fixing one thing; it’s about handling a situation where you need many devices acting at the same time for one goal.
Rosa: That's right. The paper points out that the more decentralized sources we have, the more simultaneous actions are needed to get back to a safe operating point.
Dev: And they are proposing OFO as a way to deal with this complexity by calculating those necessary actions in real time instead of relying on old plans.
Taro: What I find important is that it's not just about the calculation; it's about making sure these real-time calculations result in safe actions, which brings up some tough questions for me regarding failure modes.
The paper's summary: Rosa: So the summary of this paper explains that they are using OFO because it’s a method that lets you turn iterative optimization algorithms into controllers that work in real time and handle constraints.
Dev: They show how this controller can be used to push the system toward what they define as the safe operating region, which is basically the safe zone for the grid.
Taro: I see them using an example from a two thousand three Swiss-Italian blackout to illustrate how this works in practice, specifically showing how it could reduce the voltage difference across a breaker so operators can safely reclose it <ref:2212.07765#pg2>.
Rosa: That's a key part of their demonstration; they set up a grid model where opening a breaker causes high angle differences, and the OFO controller is what reduces that.
Dev: The paper details the structure of this controller, showing blocks like the power system stabilizer, the automated voltage regulator, and then this OFO controller itself driving things like active power and voltage set points.
Taro: So they are using a specific example—the IEEE thirty-nine bus model—to show how this mathematical approach translates into controlling those physical components of the grid <ref:2212.07765#pg2,the IEEE 39 bus model>.
Rosa: It’s about showing that you can use this OFO structure to iteratively gather measurements and update the set points for all the controllable power plants in response to what's happening.
The paper's improvements: Dev: Now, let's talk about the suggested improvements they bring to this approach. They focus on making sure this system is tough when things don't match up perfectly with their simulation model.
Rosa: They address the issue of model mismatch by using a sensitivity term, which is something called gradient of u sub h of u, y sub m, which helps the controller know how much to adjust even if the real world differs from their mathematical picture.
Taro: That sensitivity analysis is interesting because it’s their way of checking robustness; they show that even when there's severe model mismatch, the controller can still reduce the absolute voltage difference and doesn't become unstable.
Dev: They state that this robustness comes from the feedback nature of the approach, specifically how their control law acts as an integrator driven by a gradient step.
Rosa: So they are claiming that even with a lot of mismatch between their model and reality, the system stays stable because of how it's built to react.
Taro: But I have to ask what this means for actual deployment; testing robustness against real-world topologies that are different from the simulation model is going to be pretty crucial.
Conclusion: Rosa: So, wrapping up the paper "Real-time Curative Actions for Power Systems via Online Feedback Optimization," the authors show that OFO controllers have the potential to derive curative actions in real time after a contingency.
Dev: They suggest this method could automate some of those curative actions during emergency power system operations, specifically by iteratively adjusting active power and voltage set points across all generators to minimize the absolute voltage difference over a breaker.
Taro: I think what this means for us is that determining these actions in real time aligns with what European and North American grid codes are already doing, which definitely reduces the workload in the control room.
Rosa: It seems like a significant step toward making emergency power system operations more automated and faster than waiting for pre-planned responses.
Dev: We still have some reservations though; they point out that further research is needed into the stability of connecting this controller with actual power system dynamics because timescale separation issues might make current conservative results too cautious for this kind of time-critical application.
Taro: And they also flag that because we expect the system to be far from normal operating points during a contingency, they need more extensive certification of robustness against model mismatch in real-world conditions.
Rosa: So the big picture is that while it’s promising for real-time action, they need to do more work on proving its stability under extreme real-world conditions before we see it fully deployed.
Lukas Ortmann, Gianni Hotz, Saverio Bolognani, Florian Dorfler
ETH Zurich
eess.SY, cs.SY
Submitted: 2022-12-15
Updated: 2023-04-18
Journal ref: 2023 IEEE Belgrade PowerTech
DOI: 10.1109/PowerTech55446.2023.10202825
License: http://creativecommons.org/licenses/by/4.0/
Importance score: 94/100
The gist: Real-time Curative Actions for Power Systems via Online Feedback Optimization proposes calculating complex curative actions in real-time after a contingency using Online Feedback Optimization (OFO)…
Key concepts
- Curative Actions
- These are immediate measures taken after a grid contingency (like a line failure) to bring the power system back to a safe operating state. They involve quickly adjusting equipment settings to prevent instability and maintain grid security.
- Online Feedback Optimization (OFO)
- OFO is a control strategy that uses iterative optimization algorithms to create real-time feedback controllers. These controllers continuously gather system measurements and update setpoints for power plants, driving the system towards an optimal, safe operating region defined by constraints.
- Model Mismatch Robustness
- This refers to the controller's ability to perform correctly even when the mathematical model used for simulation does not perfectly match the real physical grid. The authors use sensitivity analysis to prove that their control law remains stable and effective despite significant differences between the model and reality.
Terminology
Summary
Real-time Curative Actions for Power Systems via Online Feedback Optimization proposes calculating complex curative actions in real-time after a contingency using Online Feedback Optimization (OFO) to enhance N-1 security
Need for Real-Time Control
Curative or remedial actions are the set of immediate actions intended to bring the power grid to a safe operating point after a contingency Due to the shift from steady to volatile energy resources, the grid state will frequently change and the curative actions would need to be pre-planned increasingly often Furthermore, with the shift from large bulk production to many small decentralized energy sources more devices need to be actuated simultaneously to achieve the same outcome Overall, operators will need to determine curative actions more often, and those actions will be more complicated.
Methodology of Online Feedback Optimization
The proposed control strategy is based on OFO, a methodology that allows converting iterative optimization algorithms into real-time robust feedback controllers These controllers can then be used to drive a system to the optimum of a constrained optimization problem which defines the safe operating region of the grid. The control update is given by pOF O(t + 1) vOF O(t + 1) v(t) (t)ym(t). p. 3 An OFO controller derived from such an
algorithm was developed in [22]. The control update is
pOF O(t + 1) vOF O(t + 1) v(t)
`(t)
ym(t)
Online Feedback u(t + 1) Optimization Controller Power Flow Voltage Magnitude
Voltage
Set-Points
Active Power
Set-Points
Power
Plants
(G)
SubStations Angle Difference ∆θ23−24(t)
Fig. Schematic representation of the proposed OFO-based controller. The
controller iteratively gathers power system measurements and updates setpoints for the controllable power plants in the grid.Where the paper says it
Where the paper says it
Application to Emergency Scenarios
The authors use an OFO controller inspired by the 2003 Swiss-Italian blackout, where it reduces the voltage difference over a breaker to enable the operators to reclose it. The optimization problem minimizes the voltage difference subject to actuator limits and grid constraints. p. 3 The block diagram of
our controller can be seen in Figure 7. We encode the goal of reclosing the breaker in the following
optimization problem
min
u
(v23(u) − v24(u))2 + (θ23(u) − θ24(u))2
subject to pOF O,min < pOF O < pOF O,max ∀ generators
vOF O,min < vOF O < vOF O,max ∀ generators
vmin < v(u) < vmax ∀ buses
`min < `(u) < `max ∀ lines
(1)
that minimizes the voltage difference subject to actuator limits
and grid constraints. Overall, the proposed controller quickly reduces the voltage
difference. The resulting curative actions include iterative
adjustments of the active power and voltage set-points of all
generators, showing how complex coordinated interventions
may be needed in order to effectively tackle a contingency.Where the paper says it
Where the paper says it
Robustness Against Model Mismatch
The controller is designed to be robust with respect to model mismatch by using the sensitivity ∇uh(u, ym). The results of our robustness analysis can be seen in Figure 9 and show that even with severe model mismatch, the controller is able to reduce the absolute voltage difference and does not become unstable. p. 4 Nevertheless, we
choose this source of model mismatch because we consider it
to be the most extreme. The results of our robustness analysis
can be seen in Figure 9 and show that even with severe
model mismatch, the controller is able to reduce the absolute
voltage difference and does not become unstable. However,
some levels of model mismatch cause very slow performance,
and future work should analyze how well the sensitivity needs
to be known to guarantee good performance. Generally, the
robustness against model mismatch is due to the feedback
nature of the approach and the fact that our control law (2) is
an integrator driven by a gradient step, and integral controllers
are known to be robust.Where the paper says it
Where the paper says it
Conclusion and Future Work
Preliminary numerical results show that OFO controllers have the potential to derive curative actions in real-time after the occurrence of a contingency and to automate some curative actions in emergency power system operations. Further research is needed to investigate the stability of the interconnection of the controller with the power system dynamics because timescale separation results like those in [25] turn out to be too conservative for this time-critical application. p. 4 In our opinion, determining curative actions in real-time is in
agreement with the European and North American grid codes,
and it definitely reduces the workload in the control room. Further research is needed to investigate the stability of
the interconnection of the controller with the power system
dynamics because timescale separation results like those in
[25] (which assume that grid dynamics are sufficiently faster
compared to the rate at which set-points are updated by the
controller) turn out to be too conservative for this time-critical
application. Further research is needed to investigate the stability of
the interconnection of the controller with the power system
dynamics because timescale separation results like those in
[25] (which assume that grid dynamics are sufficiently faster
compared to the rate at which set-points are updated by the
controller) turn out to be too conservative for this time-critical
application. Furthermore, because we expect the system to
work far from nominal operating points during contingencies,
robustness to model mismatch needs to be certified for this
application more extensively (possibly building on numerical
tests like those in [24]). Such
controllers could either be implemented as a decision support
tool for the operator or directly as a closed-loop controller. In our opinion, determining curative actions in real-time is in
agreement with the European and North American grid codes,
and it definitely reduces the workload in the control room.Where the paper says it
Where the paper says it
Where the paper says it
REFERENCE 2212. p. 0 Such controllers do not need a
full model of the system and guarantee constraint satisfaction
even in the presence of model mismatch. There exist several
arXiv:2212.Where the paper says it
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
-
Online Feedback Optimization controllers can be implemented as a decision support tool for operators or directly as a closed-loop controller to derive curative actions in real-time after a contingency occurs.
-
The improved AI system can automate some curative actions in emergency power system operations, such as iteratively adjusting active power and voltage set-points of all generators to minimize the absolute voltage difference over a breaker.
-
The controller's robustness against model mismatch can be tested by calculating sensitivity based on grid topologies that are different from the simulation model, ensuring performance even with severe model mismatch.
Abstract
Curative or remedial actions are the set of immediate actions intended to bring the power grid to a safe operating point after a contingency. The effectiveness of these actions is essential to guarantee curative N-1 security. Nowadays, curative actions are derived ahead of time, based on the anticipated future grid state. Due to the shift from steady to volatile energy resources, the grid state will frequently change and the curative actions would need to be pre-planned increasingly often. Furthermore, with the shift from large bulk production to many small decentralized energy sources more devices need to be actuated simultaneously to achieve the same outcome. Instead of pre-planning, we propose to calculate these complex curative actions in real-time after the occurrence of a contingency. We show how the method of Online Feedback Optimization (OFO) is well suited for this task. As a preliminary demonstration of these capabilities, we use an (OFO) controller, that after a fault, reduces the voltage difference over a breaker to enable the operators to reclose it. This test case is inspired by the 2003 Swiss-Italian blackout, which was caused by a relatively minor incident followed by ineffective curative actions. Finally, we identify and discuss some open questions, including closed-loop stability and robustness to model mismatch.
Sources
- Optimization Algorithms as Robust Feedback Controllers
- Model-Free Nonlinear Feedback Optimization
- An Open Source Power System Simulator in Python for Efficient Prototyping of WAMPAC Applications
Related papers
- Industrial Load Modeling and Optimization for Market-Based Interaction with Power Systems
- Exact and Approximate Convex Reformulation of Linear Stochastic Optimal Control with Chance Constraints
- Certified Set Convergence for Piecewise Affine Systems
- Symbolic Control for Stochastic Systems via Finite Parity Games
- Multiobjective optimization-based design and dispatch of islanded, hybrid microgrids for remote, off-grid communities in sub-Saharan Africa
- Physiology-informed layered sensing for intelligent human-exoskeleton interaction