A Coalitional Stable and Fair Reward Allocation for Dynamic Virtual Power Plants
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: "A Coalitional Stable and Fair Reward Allocation for Dynamic Virtual Power Plants".
Dev: Here's the gist of the paper. This paper establishes crucial cooperation criteria for the operation of Dynamic Virtual Power Plants (DVPPs) and proposes a control design and reward allocation mechanism to enable and incentivize Distributed Energy Resources (DERs) to provide dynamic ancillary services (DAS).
Rosa: Market Framework Introduction Ancillary service markets are generally organized centrally by the system operator, with main market design options including the procurement method, type of price, and remuneration structure.
Dev: The paper assumes a competitive DAS spot market characterized by marginal pricing and capacity-based remuneration.
Rosa: This architecture mirrors the Finnish power system used in the case study for its transparency, economic efficiency, and widespread adoption in DAS markets.
Paper summary: Rosa: So we're looking at this paper titled "A Coalitional Stable and Fair Reward Allocation for Dynamic Virtual Power Plants." Essentially, it’s about how to get distributed energy resources, or DERs, to actually cooperate and provide those dynamic ancillary services.
Dev: Right. The main idea is proposing a control design and a reward allocation mechanism specifically to enable and incentivize these DERs to provide dynamic ancillary services for the grid. It claims their results show that when you aggregate different types of DERs, you can outperform what any single one of them can do on its own <ref:2606.31728#pg1>.
Taro: So, it’s not just about adding more power; it’s about leveraging technical complementarities between the different DERs to achieve better performance than if they were operating separately <ref:2606.31728#pg1>.
Rosa: Exactly. They're looking at how these heterogeneous DERs, like batteries and wind, can work together to provide things like frequency containment reserve for disturbances FCR-D, voltage control, or fast frequency reserve FFR <ref:2606.31728#pg2>.
Dev: And they set up a competitive spot market where there's marginal pricing and capacity-based remuneration, which they think is transparent and efficient because it mirrors systems like the Finnish one <ref:2606.31728#pg2>.
Taro: So, what’s the core of their control design? How do they figure out which DER does what role in that cooperation?
Rosa: Well, they divide the DERs into three types of roles: low-pass filter, band-pass filter, and high-pass filter based on their capabilities <ref:2606.31728#pg1>. They assign each DER a part of the desired transfer function by designing something called an adaptive dynamic participation factor, or ADPF mi s <ref:2606.31728#pg1>.
Dev: And this factor encodes whether the DER is taking on the low-pass, band-pass, or high-pass role <ref:2606.31728#pg1>. The participation condition they set ensures that even with all these different roles, you still achieve the overall desired behavior up to a bandwidth τ c <ref:2606.31728#pg1>.
Taro: That sounds like a complex coordination problem in real time. What about how they handle the money part—the rewards?
Rosa: That’s where they get into game theory and propose five fundamental criteria for cooperation: individual rationality and coalitional stability, incentive compatibility, optimality and efficiency, fairness, and ex-post consistency <ref:2606.31728#pg1>.
Dev: They then suggest two ways to allocate those rewards: either the Nucleolus or the Shapley Value <ref:2606.31728#pg1>. The Nucleolus is used when the game isn't convex because it minimizes dissatisfaction across all possible coalitions to keep things stable.
Paper summary: Taro: But the Shapley Value seems more aligned with fairness, right? It assigns rewards based on a player’s average marginal contribution <ref:2606.31728#pg1>. So they use the Shapley Value when the forecasted reward game is convex to satisfy that criterion of fairness <ref:2606.31728#pg1>.
Rosa: They also check ex-post consistency, which means after you’ve actually done it, the expected value of your reward aligns with what was forecasted plus some systematic bias <ref:2606.31728#pg1>. The Shapley Value helps satisfy that because its linearity ensures the expected value aligns with the forecasted game plus a bias µ <ref:2606.31728#pg1>.
Dev: So they claim that when you look at their case study on the Finnish power grid, the cooperative reward for all DERs surpasses what they would earn acting alone by sixteen percent <ref:2606.31728#pg1>.
Taro: That’s a concrete number. And it’s not just that; they found that for something like a battery energy storage system, the Shapley Value allocation shows it benefits most from cooperation compared to acting alone because of its fast response time, allocating a reward exceeding its standalone value by forty percent <ref:2606.31728#pg1>.
Rosa: That’s interesting because it means the cooperation isn't just theoretical; it has measurable economic benefits for specific assets in the system <ref:2606.31728#pg1>. And they found that the average hourly reward of cooperation is better than acting alone for both the forecasted and realized reward games, GF and GR <ref:2606.31728#pg1>.
Dev: But there’s a limitation mentioned regarding how they solve this. They admit that scaling up to a larger number of DERs gets computationally expensive when you try to evaluate the reward allocations over all twenty N choose two coalitions <ref:2606.31728#pg1>.
Taro: So, for now, they rely on these efficient methods, like computing the Nucleolus for moderately large games or using sampling-based approximation with bounded error for the Shapley Value <ref:2606.31728#pg1>.
Rosa: And they’ve got some interesting proofs too. For individual rationality and coalitional stability, they prove that the Forecasted Reward Game split with the Nucleolus reward allocation fulfills it <ref:2606.31728#pg2>.
Dev: They also proved incentive compatibility for the Nucleolus by showing that minimizing the maximum dissatisfaction over all coalitions incentivizes truthful reporting <ref:2606.31728#pg3>. And for the Shapley Value, they showed that manipulation leads to a decrease in the allocated reward <ref:2606.31728#pg4>.
Taro: So, it sounds like they’ve done a lot of theoretical heavy lifting to back up their criteria before testing it on real-world conditions. What about how they handle when things go wrong?
Paper summary: Rosa: They define a penalty for non-available capacity as three times the potential reward, which is q equals three pi <ref:2606.31728#pg3>. They use a test function h to compare the actual DVPP response to this penalty region ρ of time and bid capacity b <ref:2606.31728#pg3>.
Dev: The test function h evaluates if the action of the DERs intersects that penalty region, and it only passes if they do not <ref:2606.31728#pg3>. They visualize this penalty region as a red shaded area in Figure two <ref:2606.31728#pg3>.
Taro: That’s important for a safety case. So, when the world misbehaves, how does the system recover gracefully?
Rosa: The optimal grand-coalition bid they selected for their test was thirty-three megawatts submitted equally between FFR and FCR-Dup <ref:2606.31728#pg4>. They also look at uncertainty, quantifying it with reliability γ D b equals the probability of success <ref:2606.31728#pg4>.
Dev: And when they optimize that reliability for the grand coalition N, they always pick the most conservative bid selection, meaning gamma N b equals one hundred percent <ref:2606.31728#pg4>. That simplifies the bidding optimization down to a risk-averse strategy <ref:2606.31728#pg4>.
Taro: So, what does this mean for someone just listening to the show? It means that if you’re running a DER, cooperating with others can actually increase your expected reward compared to just trying to operate on your own <ref:2606.31728#pg1>.
Rosa: It changes the economics of distributed energy resources by showing that cooperation isn't just nice; it has a quantifiable positive effect on the system dynamics and the economics of those resources <ref:2606.31728#pg4>.
Dev: The paper concludes that while criteria one through three are always met, they need the convexity of both games, GF and GR, to satisfy the full set of criteria <ref:2606.31728#pg4>. So those two conditions are key for getting the full theoretical payoff <ref:2606.31728#pg4>.
Taro: And they suggest that this approach might actually incentivize DER cooperation to provide those dynamic ancillary services, leading to both better power system dynamics and better economics for the resources involved <ref:2606.31728#pg4>.
Rosa: So, we’ve talked about how they set up the control design and the reward allocation rules in this paper, "A Coalitional Stable and Fair Reward Allocation for Dynamic Virtual Power Plants." It really shows that when you look at these complex systems of distributed energy resources, the way you structure cooperation matters a lot <ref:2606.31728#pg1>.
Conclusion: Rosa: So we've been looking at this paper called "A Coalitional Stable and Fair Reward Allocation for Dynamic Virtual Power Plants." Basically, it’s about figuring out how to make different energy resources cooperate so they actually provide those dynamic services we need in the grid.
Dev: Right. The authors are tackling a big problem: how do you structure the rewards so that every resource wants to join the group and everyone gets a fair deal? It assumes a competitive market, like what you'd see in Finland.
Taro: So, they propose a control design for these virtual plants—they split the resources into low-pass, band-pass, and high-pass roles based on their technical ability.
Rosa: Exactly. They use something called an adaptive dynamic participation factor to tell each resource which role it's taking on so the whole group behaves as one thing up to a certain speed limit.
Dev: And then they get into game theory for the rewards, using things like the Nucleolus or Shapley Value. It’s about making sure you have individual rationality and incentive compatibility while keeping it fair.
Taro: The results show that this cooperation actually adds value—they found a sixteen percent gain compared to each resource operating alone in their case study on the Finnish grid.
Rosa: And for things like batteries, they showed that cooperation gives them a forty percent boost over what they'd get if they acted by themselves because of how fast they can respond.
Dev: It seems like the main takeaway is that this mathematical structure can actually incentivize resources to team up for those dynamic services we need. But there are still some limitations with how hard it is to calculate these rewards for a huge number of resources.
Taro: The paper points out that while the theoretical rules work, getting these complex calculations done when you have thousands of resources is still a computational headache.
Rosa: So, they've shown the theory works and it has positive economic impacts, but the next thing we need to look at is how robust this system is when real-world data—like weather forecasts—is actually wrong.
Automatic Control Laboratory, ETH Zurich · Swiss Federal Office of Energy
eess.SY, cs.SY
Submitted: 2026-06-30
Updated: 2026-06-30
Comments: 9 pages, 8 figures
Journal ref: PowerUp Conference, 2026
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Importance score: 91/100
The gist: This paper establishes crucial cooperation criteria for the operation of Dynamic Virtual Power Plants (DVPPs) and proposes a control design and reward allocation mechanism to enable and incentivize…
Key concepts
- Dynamic Virtual Power Plants (DVPPs)
- DVPPs are systems where multiple Distributed Energy Resources (DERs) cooperate to provide dynamic ancillary services like frequency regulation. The control design divides these DERs into roles based on their capabilities—low-pass, band-pass, or high-pass filters—to achieve the desired system behavior.
- Reward Allocation Criteria
- Five criteria govern how rewards are distributed among cooperating DERs: individual rationality, incentive compatibility, optimality/efficiency, fairness, and ex-post consistency. These criteria ensure that the cooperation is stable and equitable for all participating resources.
- Nucleolus vs. Shapley Value
- These are two mathematical methods used to distribute rewards fairly. The Nucleolus minimizes dissatisfaction across all possible coalitions to ensure stability, while the Shapley Value distributes rewards based on each DER's average marginal contribution, aiming for fairness.
- Ancillary Services (DAS)
- DAS refers to essential services provided by energy resources that help maintain grid stability, such as Frequency Containment Reserve (FCR-D) and voltage control. The paper focuses on how DVPPs can incentivize DERs to offer these specific dynamic services.
Terminology
Summary
This paper establishes crucial cooperation criteria for the operation of Dynamic Virtual Power Plants (DVPPs) and proposes a control design and reward allocation mechanism to enable and incentivize Distributed Energy Resources (DERs) to provide dynamic ancillary services (DAS). p. 0
Market Framework Introduction
Ancillary service markets are generally organized centrally by the system operator, with main market design options including the procurement method, type of price, and remuneration structure. The paper assumes a competitive DAS spot market characterized by marginal pricing and capacity-based remuneration. p. 0 This architecture mirrors the Finnish power system used in the case study for its transparency, economic efficiency, and widespread adoption in DAS markets. p. 1 Prominent DAS include Frequency Containment Reserve for Disturbances (FCR-D), voltage control, and Fast Frequency Reserve (FFR). p. 1
DVPP Control Design
The DVPP control scheme divides the DERs into low-pass filter (LPF), band-pass filter (BPF) and high-pass filter (HPF) roles based on their capabilities. Each DER is assigned a portion of the desired transfer function by designing an adaptive dynamic participation factor (ADPF) mi(s), encoding the LPF-, BPF- or HPF-role. p. 1 The participation condition ensures that the overall desired behavior is achieved up to bandwidth τc, specifically P i∈D mi(s)! p. 1= 1/τcs+1. p. 1
Reward Allocation Criteria
The paper proposes five fundamental criteria governing DVPP cooperation: C1 Individual rationality and coalitional stability, C2 Incentive compatibility, C3 Optimality and Efficiency, C4 Fairness, and C5 Ex-Post Consistency. Criterion C1 ensures the viability of the DERs by ensuring every DER benefits from joining the grand coalition. p. 3 Criteria C2–C3 guarantee performance, while C4–C5 establish fairness. p. 3
Reward Allocation Methods
To satisfy the criteria, the paper proposes using either the Nucleolus or Shapley Value for reward allocation. The Nucleolus is employed when the game is non-convex, as it minimizes dissatisfaction over all coalitions to maximize coalitional stability. p. 5 The Shapley Value is used when the Forecasted Reward Game is convex, as it assigns rewards based on a player’s average marginal contribution, satisfying C4. p. 4
Case Study Results
The case study in the Finnish power grid illustrates that the cooperative reward vR(N) surpasses the summed individual values of the DERs by 16%. The Shapley Value allocation shows that the BESS benefits most from cooperation compared to acting alone due to its fast response time, allocating a reward ϕ(BESS) exceeding its standalone value v(BESS) by 40%. p. 5 The average hourly reward of cooperation exceeds the value of acting alone for all DERs and for both the Forecasted and Realized Reward Game, GF and GR. p. 5
Conclusion on Cooperation
The work determines that while criteria C1–C3 are always fulfilled, the convexity of GF and GR is required to satisfy the full set of criteria. The paper concludes that this approach may incentivize DER cooperation to provide DAS, resulting in both improved power system dynamics and DER economics, as shown by the cooperative gain of 16% compared to standalone operation. p. 6
Scalability Considerations
A known limitation of computing the reward allocations is scalability for larger N, as evaluating v(S) over all 2N coalitions is computationally expensive. Recent algorithms compute the Nucleolus efficiently for moderately large games, and the Shapley Value admits sampling-based approximation with bounded error. p. 5 Scaling the mechanism to large sets of DERs via these methods is left for future work. p. 5
Game Theory Proofs
The proof for C1 is provided by Proposition 1, which states that the Forecasted Reward Game GF split with the Nucleolus reward allocation fulfills C1. The proof for incentive compatibility of the Nucleolus is given in Appendix-C, showing that it incentivizes truthful reporting by minimizing the maximum dissatisfaction over all coalitions. p. 4 The proof for C2 of the Shapley Value is presented in Appendix-D, demonstrating that manipulation leads to a decrease in the allocated reward x(j). p. 4
Ex-Post Consistency
Ex-post consistency (C5) is satisfied because the linearity of the Shapley Value ensures that E[vR] aligns with vF plus a systematic bias µ. The expected value of vR(11) aligns with the forecasted game, i. p. 4e., E[vR] = vF + µ.
Penalty Region Analysis
The penalty for non-available capacity is defined as three times the potential reward, i.e., q = 3π. The test function h(D, b) compares the DVPP response to ρ(t, b) to assess compliance with the service obligation. p. 2 The penalty region is visualized as a red-shaded area in Figure 2. p. 2
Transfer Function Modeling
The DER time constants are modeled using a first order transfer function 1/ (τis + 1). The specific transfer functions for FFR and FCR-Dup are defined in equations (21) and (22). p. 7 The penalty region functions are given by equations (23) and (24). p. 7 The parameters of the case study include power ratings for Wind, PV, and BESS. p. 7 The procurement period is set to H hourly over the time period 2025-04-06 to 2025-04-12. p. 7 The market service is defined as 50% FFR + 50% FCR-Dup for the entire week. p. 4
Historical Data and Forecast Errors
Meteoblue history+ was employed for the archived historical weather data. Historical forecast errors are approximated by an AR(1) model where ei(Hl) = ϕi · ei(Hl − 1) + σi(Hl). p. 7 The Root Mean Square Error for Wind and PV is 4. p. 75% and 6.5%, respectively. p. 1
Final Findings on Uncertainty
Ex-ante uncertainty is quantified by the reliability (forecasted probability of success) γD(b) = P k∈M(D,b) ηk. The optimal bid b∗(D) is chosen to maximize Rγ(D, b, ID), which represents the reward D can expect ex-ante. p. 2 The optimization of Rγ (7) for the grand coalition N always yields the most conservative bid selection, i. p. 5e., γN (b) = 100%, due to the large penalty q = 3π. p. 5 This simplifies the bid optimization (8) for N to a risk-averse bidding strategy. p. 5 The study shows that the reward of cooperation exceeds the value of acting alone for all DERs and for both the Forecasted and Realized Reward Game, GF and GR. p. 5
Case Study Specifics
The optimal grand-coalition bid selected is 33 MW, submitted equally between FFR and FCR-Dup. The BESS benefits the most from cooperation compared to acting alone due to its fast response time, allocating a reward ϕ(BESS) exceeding its standalone value v(BESS) by 40%. p. 5 Wind captures a higher share of GR compared to GF because it has a higher probability of non-delivery (decreasing GF) but could deliver reliably at this specific time (increasing GR). p. 5 The total output must not enter the red-shaded area to avoid a penalty. p. 5 The Shapley Value determines the rewards in the left Table II. p. 5 The hypothetical Nucleolus reward allocation is listed in greyscale in Table II. p. 5
Future Work
Future work can focus on optimizing the ADPF design, integrating multiple services (e.g., FFR, voltage control), investigating larger case studies involving more agents (DERs), and validating larger DVPP configurations. p. 6 The paper also notes that increasing DER capacity impacts rewards non-linearly. p. 5 The work may incentivize DER cooperation to provide DAS, resulting in both improved power system dynamics and DER economics. p. 6
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
-
Optimization of reward allocation using Shapley Value or Nucleolus allows AI systems to achieve criteria C1 through C5, ensuring individual rationality, coalitional stability, incentive compatibility, optimality, fairness and ex-post consistency in a dynamic virtual power plant setting.
-
The proposed control design enables heterogeneous Distributed Energy Resources (DERs) to overcome inherent technical constraints by dividing them into low-pass filter (LPF), band-pass filter (BPF) and high-pass filter (HPF) roles, where each DER is assigned a portion of the desired transfer function via an adaptive dynamic participation factor.
-
The AI system can maximize its DAS market value by employing a scenario approach to quantify the probabilistic reward R gamma D b ID, selecting an optimal bid b star (D) to maximize this expected reward.
-
The system can determine the optimal grand coalition formation P star (N) and the corresponding forecasted reward vF(N), which represents the aggregated expected market value by bidding optimally in a competitive DAS spot market.
-
The AI system can implement a risk-averse bidding strategy when uncertainty is high, as optimization of R gamma D b ID for the grand coalition N always yields the most conservative bid selection, i.e., gamma N (b) = 100%, due to the large penalty q = 3 pi.
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