From droop to optimality: The potential of volt/var control for power distribution grid enhancement

arXiv:2305.04666 · eess.SY, cs.SY · Submitted 2023-05-08 · 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: "From droop to optimality".

Dev: Here's the gist of the paper. When high amounts of active power are injected into power distribution grids, the overall power flow is limited because voltages reach their upper acceptable limits.

Rosa: This paper investigates the optimality gap of different volt/var control schemes in a realistic setting using real data and predictions for increasing PV capacity.

Dev: Grid Context and Problem Definition Volt/var control aims to raise the power flow limit without physically reinforcing the grid by controlling voltage using reactive power.

Rosa: The optimality of a volt/var control scheme is assessed by comparing its performance to the solution of an optimal reactive power flow problem which minimizes reactive power used subject to voltage constraints at all buses and reactive power limits of participating devices.

Paper summary: Rosa: Hey Dev, so we're looking at this paper titled "From droop to optimality: The potential of volt/var control for power distribution grid enhancement". Essentially, the big idea is that when you pump a lot of active power into a grid, the voltage limits get hit, and this paper looks at how different ways of controlling reactive power—volt/var control—handle that.

Dev: Right. The core claim seems to be comparing standard droop control against more sophisticated methods to see if we can push the power flow limit higher without needing physical upgrades. It suggests that local droop control is actually not optimal and won't cut it when more solar capacity comes online, which is a key point in this paper from page one of this work <ref:2305.04666#pg1> <ref:2305.04666#pg1>.

Taro: I wonder how they frame the problem when they talk about the ideal reactive power flow solution versus what we are actually doing on the ground? Because that gap between those two is where all the trouble usually starts.

Rosa: Exactly, Taro. They compare local droop control with a few other schemes, including one based on historical data and another called OFO which uses full communication to track the optimal solution of a problem called ORPF.

Dev: And what they found is that coordinated control methods like OFO show performance that is near-optimal compared to the ORPF benchmark, suggesting they can handle more power injection. The paper points out that coordinated volt/var control can enable another ten point four percent of maximum active power injections compared to just droop control.

Taro: That ten point four percent figure is significant when you think about scaling up renewable energy integration across the whole system. So, the implication here is that better coordination matters a lot for grid capacity expansion.

Rosa: It really does. And they show that even advanced local controllers, like one tuned using historical data, can shrink that gap to optimality quite a bit compared to standard droop control. The authors suggest this is an advantage because it doesn't require you to set up massive communication infrastructure just for the local decisions.

Dev: That’s a big deal for deployment because setting up those real-time communication networks isn't always feasible right away. They also highlight that coordinated methods can reach this maximum level of grid enhancement using only voltage magnitude measurements and minimal model information.

Taro: So, if we look at the failure modes, it seems like local controllers are inherently limited by the information asymmetry, which means they are stuck on a Pareto curve that the true optimum dominates. That's what page two of this work <ref:2305.04666#pg2> discusses about how comparing performance to an optimal reactive power flow problem defines optimality <ref:2305.04666#pg1>.

Rosa: Right. And that leads us into what it means for real-world operation, because they point out that local controllers are subject to those strict Pareto curves that the optimum dominates, which shows the limitation of just acting on locally available voltage measurements.

Paper summary: Dev: The paper is quite clear about this trade-off: you have to choose between how much reactive power you use and how well you satisfy all the voltage constraints across every bus in the system. Local controllers can't always balance that perfectly because they don't see the whole picture, which is a constraint mentioned on page two of this work <ref:2305.04666#pg2> <ref:2305.04666#pg1>.

Taro: If we think about what happens when things go wrong, like when you add all that extra PV capacity they are testing in their scenarios, the difference between the droop and OFO controllers in terms of constraint violations grows quite large.

Rosa: That's where the two thousand thirty-five scenario comes in <ref:2305.04666#pg1>. When all that additional PV capacity is installed at R15, the difference in those persistent constraint violations between the droop and OFO controllers jumps from eight point six to one hundred fifteen point two per unit hours when we look at page three of this work <ref:2305.04666#pg3>.

Dev: One hundred fifteen point two per unit hours sounds like a lot of persistent errors, which means the control scheme is struggling significantly under that high injection load. But the paper says OFO still performs comparably to that ORPF benchmark.

Taro: So the coordination really pays off when you push past the limits where local decisions alone start failing, even if they are pretty good locally. It seems like coordinated control can track that ideal solution and mitigate constraint violations much better than what standard droop can do on its own.

Rosa: That’s the main thrust of this paper, "From droop to optimality: The potential of volt/var control for power distribution grid enhancement". It shows that while droop is the current standard, it’s fundamentally suboptimal when you push for higher injection limits.

Dev: And the implication is that for grid operators, moving toward more advanced local control methods that use historical data might be a viable step to shrink that gap without needing full communication infrastructure right away. But we have to remember what the authors flag: these local controllers are inherently suboptimal because they lack information about the whole system.

Taro: So, looking ahead, it sounds like future research should focus on how local data-driven control methods can further reduce that optimality gap under different operational conditions, rather than just sticking to standard droop tuning.

Rosa: Exactly. The paper ends by suggesting that advanced local control methods have an evident advantage over standard droop because they lower the optimality gap without needing that communication infrastructure we talked about earlier.

Dev: And for those who only listen to the show, it means if you're managing a grid with lots of solar, you should seriously consider moving beyond basic droop settings toward something that learns from historical data to get closer to what's actually possible.

Conclusion: Rosa: So we've been looking at how different methods of controlling reactive power affect how much extra active power you can push through your grid, and this paper is titled "From droop to optimality: The potential of volt/var control for power distribution grid enhancement".

Dev: Yeah, the authors are really digging into comparing standard droop control with these more advanced settings to see if we can find a better way to handle those voltage limits.

Rosa: They're essentially showing that standard droop isn't the best approach when you try to maximize what you can inject into the power system.

Dev: They do look at this trade-off between how much reactive power you use and whether all the voltage constraints are actually met everywhere on those buses. It’s about finding a better balance.

Rosa: The main point is that coordinating your controls, using methods like OFO, can let you get another ten point four percent of maximum active power injection compared to just sticking with droop control.

Dev: That ten point four percent difference is significant when you're talking about integrating more solar capacity into a real system. It suggests that better coordination matters for how much power the grid can handle.

Rosa: And they show that even these advanced local controllers, the ones using historical data, can shrink that gap with optimality quite a bit without needing massive communication setups.

Dev: But they also lay out some real limitations, showing how much the performance difference between droop and those coordinated methods jumps when you add a lot of extra solar capacity at certain points.

Rosa: So the implication is that for grid operators, moving toward smarter control strategies, even local ones based on data, could help them push their grid capacity further.

Dev: That means we need to think about how these advanced local controls can actually run in real time and what the safety case looks like when things get stressful.

Rosa: And that brings us to how those coordinated methods track the ideal solution, which is something we'll explore next.

Jonas G. Matt, Lukas Ortmann, Saverio Bolognani, Florian Dörfler

Automatic Control Laboratory, ETH Zurich · Eastern Switzerland University of Applied Sciences

eess.SY, cs.SY

Submitted: 2023-05-08

Updated: 2026-06-22

Comments: Published in Sustainable Energy, Grids and Networks, Vol. 46, Article 102379 (2026)

Journal ref: Sustainable Energy, Grids and Networks, 2026 47, 102379

DOI: 10.1016/j.segan.2026.102379

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

Importance score: 87/100

The gist: When high amounts of active power are injected into power distribution grids, the overall power flow is limited because voltages reach their upper acceptable limits.

Key concepts

Volt/var Control
This technique aims to increase the power flow limit on a grid by managing voltage using reactive power instead of physically upgrading lines. It seeks to raise voltage levels without needing new hardware, optimizing how much reactive power is used.
Droop Control
This is the standard, local control method where each device sets its reactive power output based only on its own local voltage measurement and a fixed curve (droop). It's simple but lacks coordination across the entire grid.
Optimality Gap
This measures how far a real-world control scheme is from the absolute best possible solution for managing reactive power under voltage constraints. The paper shows that local controllers have a large gap, while coordinated methods can reduce this significantly.
OFO (Optimal Flow Optimization)
A sophisticated control scheme requiring full communication and feedback to track the optimal reactive power flow solution. It can achieve near-optimal performance by coordinating all devices to minimize reactive power usage while satisfying all voltage limits across the grid.

Terminology

Summary

When high amounts of active power are injected into power distribution grids, the overall power flow is limited because voltages reach their upper acceptable limits. p. 0

Where the paper says it

, optimized based on historical data) may also shrink the gap to optimality. A R T I C L E I N F O Keywords: Grid enhancement Volt/var control Reactive power Renewable energy Online feedback optimization Droop control Real data A B S T R A C T When high amounts of active power are injected into power distribution grids, the overall power flow is limited because voltages reach their upper acceptable limits.

This paper investigates the optimality gap of different volt/var control schemes in a realistic setting using real data and predictions for increasing PV capacity. p. 1
Where the paper says it

In this paper, we thus aim to investigate the optimality gap of different volt/var control schemes in a realistic setting. Using an extensive dataset of real household electricity consumption and PV production, combined with predictions of increasing PV capacity over the next two decades, we first demonstrate that overvoltages will occur more frequently during peaks of DER infeed.

Grid Context and Problem Definition

Volt/var control aims to raise the power flow limit without physically reinforcing the grid by controlling voltage using reactive power. The optimality of a volt/var control scheme is assessed by comparing its performance to the solution of an optimal reactive power flow problem which minimizes reactive power used subject to voltage constraints at all buses and reactive power limits of participating devices. p. 1

Where the paper says it

These devices are controlled voltage sources and can actively engage in volt/var control by adjusting their reactive power setpoints according to a prescribed control scheme. The optimality of a volt/var control scheme can be assessed by comparing its performance to the solution of an optimal reactive power flow problem like minimize reactive power used subject to voltage constraints at all buses and reactive power limits of participating devices, (1) which ensures an acceptable voltage profile across the grid whenever available reactive power resources permit it using a minimum amount of reactive power to do so.

The predominant method currently employed for volt/var control is droop control, where each inverter determines its reactive power setpoint based on a predefined function of the local voltage measurement, commonly referred to as a droop curve. p. 1
Where the paper says it

From a grid enhancement perspective, choosing the reactive power setpoints according to (1) maximizes the potential active power flows without compromising voltage stability. The predominant method currently employed for volt/var control is droop control, wherein each inverter determines its reactive power setpoint based on a predefined function of the local voltage measurement, commonly referred to as a droop curve.

Local droop control is a local control scheme because the control actions of participating devices are executed without coordination. p. 1
Where the paper says it

In fact, many modern grid codes already mandate the involvement of these devices in voltage regulation [2–4]. Droop control is a local control scheme because the control actions of participating devices are executed without coordination.

Control Schemes Compared

The paper compares four different volt/var controllers within the simulation framework: (1) local droop control with standard parameter tuning, (2) a local droop controller tuned based on historical data and a system model according to [23], (3) a local droop controller tuned based on historical solutions of the ORPF problem according to [11], and (4) OFO, which relies on full communication and measurement feedback to track the solution of the ORPF problem.

Performance Analysis Across Scenarios

The numerical study over a whole year of real data suggests that coordinated volt/var control can enable another 10.4% of maximum active power injections compared to droop control. p. 0

Where the paper says it

Our numerical study over a whole year of real data suggests that coordinated volt/var control can enable another 10. 4% of maximum active power injections compared to droop control.

In the 2035 scenario, the difference in persisting constraint violations between the droop and OFO controllers increases from 8. p. 3
Where the paper says it

Under this distribution, the difference in persisting constraint violations between the droop and OFO controllers increases from 0. 6 per-unit hours in the 2030 scenario and from 8.

6 to 115.2 per-unit hours when all additional PV capacity is installed at R15. p. 3
Where the paper says it

Bolognani et al. Table 1 Persisting constraint violations in per-unit hours under the droop and OFO controllers, for uniform PV scaling and for the worst-case distribution concentrating all additional PV capacity at the electrically most isolated bus R15.

The OFO controller exhibits a performance comparable to that of the ORPF benchmark, achieving near-optimal control relative to the ORPF benchmark. p. 5
Where the paper says it

Lastly, the OFO controller (Fig. 6(e)) exhibits a performance that is comparable to that of the ORPF benchmark.

Fundamental Trade-offs and Limitations

Any volt/var control method must fundamentally navigate a tradeoff between reactive power usage and voltage constraint satisfaction. Local controllers are subject to Pareto curves that are strictly dominated by the optimum, illustrating the intrinsic limitation of information asymmetry. p. 8

Where the paper says it

The ORPF is the theoretical optimum. Coordinated controllers such as OFO can in principle achieve this optimum, whereas local controllers are subject to Pareto curves that are strictly dominated by the optimum.

Coordinated controllers such as OFO can in principle track this solution, whereas local controllers act only on locally available voltage measurements which constitutes an information asymmetry. p. 7
Where the paper says it

A centralized controller such as OFO which has access to grid-wide measurements can in principle track this solution, as can be seen in Fig. Local controllers however, act only on locally available voltage measurements, constituting an information asymmetry that is the fundamental reason for their suboptimal performance.

Experimental Validation

The experiment on a real distribution grid feeder demonstrated that when OFO is used for control, the grid can be enhanced to transmit 11.23 kW which is another 10. p. 0

Where the paper says it

We finally demonstrate that coordinated control methods can track the optimal solution and enhance the grid to its full potential if grid-wide communication is available. Our numerical study over a whole year of real data suggests that coordinated volt/var control can enable another 10.

5% on top of the capacity achieved using droop control. p. 8
Where the paper says it

23 kW which is another 10. 5% on top of the capacity achieved using droop control.

The experiment showed that coordinated control methods like OFO can reach this maximum level of grid enhancement using only voltage magnitude measurements and minimal model information. p. 9
Where the paper says it

In the experiment on a real distribution grid feeder, the enhancement was 10. We showed that coordinated control methods like OFO can reach this maximum level of grid enhancement using only voltage magnitude measurements and minimal model information.

The paper concludes that advanced volt/var control can enhance the capacity of a distribution grid by approximately 10% beyond the currently established practice. p. 9
Where the paper says it

Finally, we demonstrated its practical viability on a real distribution grid feeder. Overall, our analysis suggests that advanced volt/var control can enhance the capacity of a distribution grid by approximately 10% beyond the currently established practice, enabling distribution grid operators to mitigate or postpone physical grid reinforcements that may be required in the future.

Future Directions and Implications

The study highlights the importance of moving toward better volt/var control methods that use available reactive power resources as effectively as possible. The authors suggest that advanced local control methods have evident advantages over standard droop control by reducing the optimality gap without needing communication infrastructure. p. 9

Where the paper says it

Advanced local control methods such as the presented optimized droop controllers have evident advantages over standard droop control. They can reduce the optimality gap without the need for a communication infrastructure.

They also encourage future studies to explore how local data-driven control methods can further reduce the optimality gap under different conditions. p. 9
Where the paper says it

While these were not included explicitly in this work, it remains interesting to explore by how much and under what conditions these methods can further reduce the optimality gap. We encourage future studies to replicate our experiments to validate and extend their findings.

The findings motivate distribution grid operators to adopt more advanced volt/var control strategies and potentially invest in real-time communication infrastructure. p. 9
Where the paper says it

Overall, our analysis suggests that advanced volt/var control can enhance the capacity of a distribution grid by approximately 10% beyond the currently established practice, enabling distribution grid operators to mitigate or postpone physical grid reinforcements that may be required in the future. We believe that our study can motivate distribution grid operators to adopt more advanced volt/var control strategies and, potentially, invest in real-time communication infrastructure to enable coordinated control.

Data and Modeling Assumptions

The simulation framework is based on Python and the open-source package pandapower. The analysis focuses exclusively on voltage constraint violations as the limiting factor for PV integration, not line flow limits. p. 2

Where the paper says it

Our analysis focuses exclusively on voltage constraint violations as the limiting factor for PV integration. Line flow limits are not considered, which is justified for the low-voltage network studied here, where overvoltage becomes the binding constraint before thermal limits are reached.

Modeling assumptions include assuming a box-shaped P-Q feasibility region for reactive power setpoints. p. 2
Where the paper says it

Reactive power capability region. We assume a box-shaped P-Q feasibility region, so that reactive power setpoints can take any value in [𝑞min, 𝑞max] regardless of active power generation.

The study deliberately studied two polar cases: local controllers that operate without any communication, and coordinated controllers with perfect (lossless, delay-free) communication to all grid nodes. p. 3
Where the paper says it

Either no or perfect communication. We deliberately study two polar cases: local controllers that operate without any communication, and coordinated controllers (OFO, ORPF) with perfect (lossless, delay-free) communication to all grid nodes.

AI and Data Utilization

The DNN-based droop control optimizes the trade-off directly via a regularization parameter lambda that balances reactive power usage against constraint satisfaction. The authors used AI tools for spell- and grammar-checking, as well as style improvements during the preparation of this work. p. 9

Where the paper says it

Florian Dörfler: Writing – review & editing, Supervision, Project administration, Funding acquisition, Conceptualization. Declaration of Generative AI and AI-assisted technologies in the writing process During the preparation of this work, the authors used AI tools for spell- and grammar-checking, as well as style improvements.

p. 9
Where the paper says it

Florian Dörfler: Writing – review & editing, Supervision, Project administration, Funding acquisition, Conceptualization. Declaration of Generative AI and AI-assisted technologies in the writing process During the preparation of this work, the authors used AI tools for spell- and grammar-checking, as well as style improvements.

The DNN training objective minimizes a cost function of the form involving the penalty function that acts as a soft approximation of the voltage constraints. p. 4
Where the paper says it

A separate DNN is trained for each bus to determine the optimal inflection points 𝜃ℎ = (𝑣 ℎ 1 , …, 𝑣ℎ 4 ) for the local droop curve at bus ℎ. The DNN training objective is to minimize a cost function of the form arg min 𝜃ℎ ∑ (𝑝ℎ ,𝑞ℎ )∈D ‖𝑞ℎ (𝑣ℎ ; 𝜃ℎ )‖ 2 2 + 𝜆𝑝(𝑣ℎ ), (4) where the penalty function 𝑝(𝑣ℎ ) ∶= ⎧ ⎪ ⎨ ⎪ ⎩ (𝑣min − 𝑣ℎ ) 2 , if 𝑣ℎ < 𝑣min, (𝑣ℎ − 𝑣max) 2 , if 𝑣ℎ > 𝑣max, 0, otherwise (5) acts as a soft approximation of the voltage constraints in (2).

The authors used AI tools for spell- and grammar-checking, as well as style improvements during the preparation of this work.

Conclusion on Control Effectiveness

The experiment and simulations show that volt/var control is an effective means to enhance the capacity of power distribution grids. The current state of the art, droop control, is inherently suboptimal, both with respect to the constraint violations mitigated and the efficient use of reactive power. p. 9

Where the paper says it

will be a crucial component of distribution grids in the near future, as the installed capacity of distributed energy sources such as PV increases even further. However, our experiments demonstrate that the current state of the art, droop control, is inherently suboptimal, both with respect to the constraint violations mitigated and the efficient use of reactive power.

Advanced local control methods have evident advantages over standard droop control by reducing the optimality gap without communication infrastructure. p. 9
Where the paper says it

Advanced local control methods such as the presented optimized droop controllers have evident advantages over standard droop control. They can reduce the optimality gap without the need for a communication infrastructure.

Coordinated control methods can yield near-optimal mitigation of constraint violations using minimal reactive power. p. 9
Where the paper says it

We encourage future studies to replicate our experiments to validate and extend their findings. In contrast to local control, we demonstrated how adding communication and coordinating the volt/var control across the whole grid can yield near-optimal mitigation of constraint violations using minimal reactive power.

The study suggests that coordinated control can track the ORPF solution and enhance the grid to its full potential if grid-wide communication is available. p. 0
Where the paper says it

We finally demonstrate that coordinated control methods can track the optimal solution and enhance the grid to its full potential if grid-wide communication is available. Our numerical study over a whole year of real data suggests that coordinated volt/var control can enable another 10.

Data Availability

The authors do not have permission to share the data. p. 0

Where the paper says it

https://doi. 102379 Received 9 January 2026; Received in revised form 12 June 2026; Accepted 13 June 2026 Sustainable Energy, Grids and Networks 47 (2026) 102379 Available online 16 June 2026 2352-4677/© 2026 The Authors.

The dataset used was identified as Dataport by Pecan Street Inc. p. 1
Where the paper says it

Among the limited range of publicly available datasets suited for this purpose, Dataport by Pecan Street Inc. [26] was identified as the best one according to the following criteria: a large number of different households at a single location with matching load and PV generation data, high temporal resolution and a total duration sufficient to reveal seasonal effects.

[26]. The dataset available under free academic licensing is from the year 2018, and contains data from 25 different households in Austin, USA. p. 1
Where the paper says it

[26] was identified as the best one according to the following criteria: a large number of different households at a single location with matching load and PV generation data, high temporal resolution and a total duration sufficient to reveal seasonal effects. The dataset available under free academic licensing is from the year 2018, and contains data from 25 different households in Austin, USA with a temporal resolution of one second and a total duration of one year.

The authors do not have permission to share the data.

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. A coordinated control system can track optimal solutions by using online feedback optimization (OFO) strategies, which repeatedly update the reactive power setpoints of the DERs based on real-time voltage measurements. This allows for enhanced grid capacity by another 10.4% of maximum active power injections compared to droop control.

  2. AI systems can be trained to determine optimal local droop curve inflection points using deep neural networks (DNN), which train a DNN to identify the optimal inflection points of the droop curves for the inverters at each bus based on historical data and a linearized grid model. This leads to optimized droop controllers that reduce the optimality gap without the need for a communication infrastructure.

  3. AI systems can be trained using multi-layer perceptrons (MLP) to approximate optimal reactive power flow solutions, where they approximate the ORPF solution by fitting droop curves directly. This allows for creating local control rules based on historical ORPF solutions, potentially improving efficiency compared to standard droop control.

  4. A coordinated system can use dual ascent strategies with two dual variables per inverter to integrate constraint violations and update setpoints, allowing it to track the ORPF solution and guarantees constraint satisfaction even under model mismatch. This is achieved by using real-time voltage measurements from the grid.

  5. The improved AI system can perform grid enhancement by finding reactive power setpoints that minimize reactive power used subject to voltage constraints at all buses through an optimization process, directly aiming for the theoretical optimum defined by the ORPF problem.

Abstract

When high amounts of active power are injected into power distribution grids, the overall power flow is limited because voltages reach their upper acceptable limits. Volt/var control aims to raise this power flow limit without physically reinforcing the grid but by controlling the voltage using reactive power. We use real consumption and generation data on a low-voltage CIGRÉ grid model and an experiment on a real distribution grid feeder to analyze how different volt/var methods can enhance the grid. We show that local droop control enhances the grid but underutilizes the reactive power resources. We discuss how this inefficiency can be partly reduced by fine-tuning the droop curves through data-driven techniques but illustrate that inherent trade-off persist for any local control method. We finally demonstrate that coordinated control methods can track the optimal solution and enhance the grid to its full potential if grid-wide communication is available. Our numerical study over a whole year of real data suggests that coordinated volt/var control can enable another 10.4% of maximum active power injections compared to droop control. In a small-scale real-life experiment, coordinated control enhanced the grid by the same amount.

Related papers