⚡ electrical engineering

Privacy-Preserving Uncertainty Disclosure for Facilitating Enhanced Energy Storage Dispatch

This paper proposes a novel privacy-preserving framework that releases marginal value function bounds to reduce forecast conservativeness and mitigate excessive withholding, thereby significantly enhancing energy storage dispatch efficiency, system welfare, and storage profitability.

Original authors: Ning Qi, Xiaolong Jin, Kai Hou, Zeyu Liu, Hongjie Jia, Wei Wei

Published 2026-03-17

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Ning Qi, Xiaolong Jin, Kai Hou, Zeyu Liu, Hongjie Jia, Wei Wei

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine the power grid as a massive, bustling marketplace where electricity is bought and sold every second. In this market, Energy Storage (like giant batteries) plays the role of a clever trader. Their job is simple: buy electricity when it's cheap, store it, and sell it back when it's expensive. This is called "arbitrage."

However, there's a big problem: Uncertainty.

The Problem: The Foggy Forecast

Imagine you are a battery trader. You know the price of electricity right now, but the future is foggy. You don't know exactly how much wind or solar power will be generated tomorrow, or how many people will turn on their air conditioners.

Because the future is unclear, traders get scared. To protect themselves, they become too conservative. They think, "What if the price spikes even higher than I expect? I better hold onto my battery and not sell yet, just in case."

This is called "Withholding."

The Result: The battery sits idle. The grid loses a valuable tool to balance itself. Electricity prices become more volatile, and the whole system becomes less efficient and more expensive for everyone.

The Solution: A "Safe Guard" from the Boss

The System Operator (the "Boss" of the grid) has all the data. They know the weather, the demand, and the risks. But they can't just tell the traders, "Hey, the price will definitely be $50 tomorrow," because that would reveal secret market data and ruin the competition.

This paper proposes a clever middle ground: Privacy-Preserving Uncertainty Disclosure.

Think of it like a Weather App with a "Safety Net".
Instead of giving the traders a specific, secret forecast, the Boss gives them a "Safety Ceiling" (a mathematical bound).

The Analogy: Imagine you are driving in fog. You can't see the road ahead. The Boss doesn't tell you the exact location of every pothole (privacy). Instead, they put up a giant, glowing sign that says: "The road ahead will never be worse than this specific bump."
The Effect: Now, the trader knows, "Okay, the worst-case scenario is this bad. I don't need to be terrified of a worse scenario that doesn't exist."

How It Works (The Magic Trick)

The "Rolling Horizon" (The Moving Safety Net):
The Boss doesn't just give one static sign for the whole day. As time passes and the fog lifts (real-time data comes in), the Boss updates the sign.
- Morning: The sign is high and wide (lots of uncertainty).
- Noon: The sign drops lower and gets tighter (we know more now).
- This allows the battery trader to adjust their strategy in real-time, becoming less conservative as the day goes on.
The "Opportunity Cost" Cap:
The paper proves mathematically that this "Safety Ceiling" is a reliable limit on how much money the battery could lose by acting too cautiously. It acts as a truthful anchor.
- If the battery thinks, "I should hold back because the price might go to $100," but the Boss's sign says, "The max opportunity cost is actually only $40," the battery realizes it's being too scared. It starts selling more often.
Stopping the "Hoarding":
Sometimes, traders try to game the system by holding back too much power to artificially drive prices up. The Boss's "Safety Ceiling" acts like a speed limit. If a trader tries to bid too high (withholding too much), the system says, "Nope, your bid is capped at this safe limit." This stops them from hoarding power unnecessarily.

The Results: Why It Matters

The authors tested this on a simulated version of the New England power grid (ISO-NE). Here is what happened when they used this "Safety Net":

Batteries Woke Up: Storage units started responding 39% more to market signals. They stopped sitting idle and started doing their job.
Less Waste: The "gap" between what the system could do and what it actually did shrank significantly.
Money Saved: The total cost to run the grid dropped slightly (about 0.23%), which sounds small but saves millions of dollars across a whole country.
Traders Win Too: Because the batteries were used more efficiently, the battery owners actually made 13% more profit.

The Big Picture

This paper solves a classic dilemma: How do you share important information without giving away secrets?

By giving traders a "reliable limit" on uncertainty rather than a specific prediction, the System Operator helps them make smarter, less fearful decisions. It turns a group of scared, hoarding batteries into a coordinated, efficient team that keeps the lights on and the prices fair.

In short: It's about replacing "paralysis by analysis" with "confidence by bounds," making the power grid safer, cheaper, and greener for everyone.

1. Problem Statement

The integration of high levels of renewable energy has increased the reliance on energy storage for grid flexibility. However, current storage dispatch is largely decentralized, with participants making decisions based on limited information.

The Core Issue: Storage operators lack accurate, real-time system-level uncertainty information. Consequently, they rely on conservative day-ahead price forecasts and interval predictions.
Consequences: This information gap leads to conservative decision-making and strategic capacity withholding (withholding capacity to wait for higher prices). This behavior reduces social welfare, increases system costs, and leads to inefficient storage utilization.
The Challenge: System operators (ISOs) cannot simply disclose raw uncertainty data (load, renewable generation, network constraints) due to privacy concerns and market sensitivity. Existing disclosure mechanisms (e.g., deterministic bid caps) are often static, day-ahead, and fail to account for real-time uncertainty updates.

2. Methodology

The paper proposes a Privacy-Preserving Uncertainty Disclosure Framework that allows system operators to release real-time marginal value function bounds without revealing sensitive system data.

A. Risk-Averse Storage Arbitrage Model

Formulation: The authors model storage arbitrage using Stochastic Dynamic Programming (SDP).
Participants: The model covers both price-takers (self-scheduling) and price-makers (strategic bidding).
Uncertainty Handling: Real-time prices (RTP) are modeled as a Markov process with a mean equal to the Day-Ahead Price (DAP) and a standard deviation ( $\sigma$ ) representing uncertainty. The value function is trained explicitly accounting for these uncertainty intervals.
Theoretical Properties: The paper proves that the marginal value function is:
- Monotonically decreasing with the State of Charge (SoC).
- Monotonically increasing with price uncertainty (under normal price regimes), explaining why storage tends to withhold capacity when uncertainty is high.

B. Privacy-Preserving Uncertainty Disclosure Framework

Mechanism: The system operator solves a Rolling-Horizon Chance-Constrained Economic Dispatch (RT-CED) problem.
Output: Instead of raw data, the operator derives and publishes Marginal Value Function Bounds ( $\theta^\epsilon_{s,t}$ ). These bounds represent the opportunity cost of storage from the system's perspective.
Privacy: The bounds are derived from dual variables of the dispatch problem. They cap the true opportunity cost without revealing load forecasts, renewable generation profiles, or network topology.
Dynamic Convergence: Unlike static day-ahead bounds, these bounds are updated in real-time. As forecast errors vanish, the bounds dynamically tighten and converge to the "hindsight" (perfect information) value.

C. Adjusted Storage Dispatch Algorithm

The framework enables two mechanisms for storage participants:

Indirect Improvement: Storage participants use the disclosed bounds to refine their own uncertainty interval predictions ( $\sigma_{s,t}$ ), leading to less conservative control policies.
Direct Mitigation: For price-makers, the bounds serve as a cap on bidding strategies. If a storage unit attempts to withhold capacity by bidding above the calculated bound, the bid is capped, preventing excessive market power abuse.

3. Key Contributions

Novel Framework: Introduction of a privacy-preserving mechanism where ISOs release real-time marginal value function bounds rather than raw uncertainty data.
Theoretical Rigor:
- Proof that the proposed bounds reliably cap the true marginal value function with a specific confidence level ( $1-\epsilon$ ).
- Proof of monotonicity: Bounds decrease with SoC and increase with uncertainty, providing a theoretical basis for risk-averse behaviors.
- Proof of dynamic convergence: Bounds tighten over time as real-time data becomes available.
Algorithm Design: Development of an adjusted dispatch algorithm that utilizes these bounds to either update internal prediction intervals or cap strategic bids.
Comprehensive Validation: Extensive agent-based simulations on the ISO-NE test system demonstrating scalability and effectiveness.

4. Results

The framework was validated using an agent-based simulation on a modified 8-zone ISO-NE system with 50% renewable capacity and 35% storage capacity.

Enhanced Storage Response:
- Storage response increased by 38.91% during peak hours.
- The optimality gap (difference between actual dispatch and perfect hindsight) was reduced to 3.91%.
- This was achieved by allowing storage to dynamically adjust their prediction intervals based on the disclosed bounds.
Mitigation of Excessive Withholding:
- By capping bids, the framework reduced the average system cost by 0.23%.
- It increased average storage profit by 13.22% (by preventing inefficient withholding that leads to missed arbitrage opportunities).
Scalability and Sensitivity:
- Uncertainty: Benefits scale with higher system uncertainty. At a high uncertainty scale (3.0x), storage profit increased by over 140% in some scenarios.
- Capacity: Benefits increase with higher renewable and storage penetration.
- Computational Efficiency: While calculating bounds for 10,000 storage units without relaxation takes >300s, applying a relaxation (allowing simultaneous charge/discharge for calculation) reduces this to ~250s with negligible accuracy loss (0.03% gap).

5. Significance

Bridging the Information Gap: This work solves the critical trade-off between data privacy and market efficiency. It allows ISOs to guide decentralized storage without compromising proprietary data.
Market Efficiency: By reducing the conservativeness of storage operators, the framework unlocks latent storage capacity, lowering overall system costs and improving renewable integration.
Regulatory Tool: The "bound" mechanism offers a new regulatory tool for ISOs to prevent market manipulation (excessive withholding) without resorting to rigid, static price caps.
Theoretical Foundation: The proofs regarding the monotonicity of value functions with respect to SoC and uncertainty provide a robust mathematical foundation for future risk-averse storage designs and SoC-dependent control strategies.

In summary, the paper presents a mathematically rigorous and practically viable solution to enhance energy storage dispatch in high-renewable grids by leveraging privacy-preserving uncertainty disclosure to align individual storage incentives with system-wide social welfare.