📈 economics

The Welfare Gap of Strategic Storage: Universal Bounds and Price Non-Linearity

This paper establishes that while the efficiency loss of decentralized battery storage in electricity markets is tightly bounded by 4/3 under linear prices and general convex constraints, it can become unbounded under general convex price functions, though it remains bounded by 2 for monomial prices and converges to perfect efficiency as the number of competing batteries increases.

Original authors: Zhile Jiang, Xinhao Nie, Stratis Skoulakis

Published 2026-07-03

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Zhile Jiang, Xinhao Nie, Stratis Skoulakis

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 electricity grid as a giant, bustling kitchen. The "demand" is the hunger of the customers (people turning on lights and appliances), and the "supply" is the chefs (power plants) cooking the meals. Sometimes, the kitchen gets chaotic: too many orders at once (peak hours) or too few orders (off-peak hours).

Battery storage is like a smart pantry. It can buy food when the kitchen is quiet and cheap, store it, and then serve it when the kitchen is frantic and expensive. This helps smooth out the chaos.

However, there is a conflict of interest between two managers of this pantry:

The Social Planner (The Idealist): Their only goal is to make the whole kitchen run as cheaply and efficiently as possible for everyone. They want to perfectly smooth out the hunger spikes.
The Profit-Maximizing Operator (The Entrepreneur): Their only goal is to make the most money. They buy low and sell high, but they might not smooth things out perfectly because that wouldn't be as profitable for them.

This paper asks: How much worse does the kitchen run when the Entrepreneur is in charge compared to the Idealist? In economics, this gap is called the "Price of Anarchy."

Here is what the authors discovered, using simple analogies:

1. The "Flat Road" Scenario (Linear Prices)

Imagine the price of electricity is like a flat road that goes up at a steady, straight angle. If demand goes up a little, the price goes up a little, in a perfectly straight line.

The Finding: Even if the Entrepreneur is greedy, they can't mess things up too badly. The worst-case scenario is that the kitchen is only 33% less efficient (a ratio of 4/3) than if the Idealist were in charge.
The Analogy: It's like driving on a straight highway. Even if you drive a bit recklessly to save a few minutes, you can't get lost or crash into a wall. The "straightness" of the road (linear pricing) keeps everyone on track.
The Surprise: This 33% limit holds true no matter how crazy the weather (demand) gets or how big the pantry (battery) is. It's a structural rule of the universe for straight-line prices.

2. The "Curved Road" Scenario (Non-Linear Prices)

Now, imagine the price of electricity doesn't go up in a straight line. Instead, it curves sharply upward, like a rollercoaster. As demand gets high, the price skyrockets exponentially.

The Finding: If the price curve is too steep or weirdly shaped, the Entrepreneur can cause infinite chaos. The efficiency loss can become unbounded.
The Analogy: Imagine a road that suddenly turns into a vertical cliff. If the Entrepreneur tries to drive fast to make money, they might drive right off the cliff, causing a disaster that the Idealist could have easily avoided. The "curvature" of the price breaks the safety net.
The Nuance: However, if the price curve is a specific, simple shape (like a smooth parabola, or a "monomial"), the chaos is still bounded. The worst it gets is twice as bad (a ratio of 2) as the Idealist's plan. But as the curve gets steeper (higher degree), the gap widens, though it never becomes infinite for these specific shapes.

3. The "Crowded Kitchen" Scenario (Multiple Batteries)

What if we don't have just one Entrepreneur, but a whole team of them competing against each other?

The Finding: Competition is the cure. As you add more batteries (more entrepreneurs) to the market, they start fighting for customers. This competition forces them to act more like the Idealist.
The Analogy: If there is one baker, they can charge whatever they want. If there are 100 bakers, they have to keep their prices low and their service high to survive.
The Result: As the number of batteries grows toward infinity, the "Price of Anarchy" drops to 1. This means the market becomes perfectly efficient. The greed of the individuals cancels out, and the system runs exactly as well as if a single Idealist were in charge.

Summary

The paper tells us that the shape of the price tag matters most.

If prices rise in a straight line, the market is safe; the worst inefficiency is capped at 33%.
If prices rise in a wild curve, the market can break completely.
If you have many competitors, they naturally fix the problem, driving the market toward perfect efficiency.

The authors prove these limits mathematically, showing that while individual greed can cause some waste, the structure of the market (linear prices) or the presence of competition can keep that waste from becoming a catastrophe.

Technical Summary: The Welfare Gap of Strategic Storage

1. Problem Formulation

This paper investigates the efficiency loss in electricity markets caused by the misalignment between social welfare objectives and the profit-maximizing behavior of strategic battery storage operators. The problem is formulated within a generalized continuous-time stochastic setting over a normalized periodic cycle $T = [0, 1]$ .

Market Dynamics: Demand $D(t)$ is modeled as a stochastic process with periodicity. The electricity price $P(N(t))$ is determined by an increasing function of net demand $N(t) = D(t) - B(t)$ , where $B(t)$ is the battery discharge rate. The social generation cost is the integral of the price function, $G(N) = \int_0^N P(x)dx$ .
Operational Constraints: The battery policy $B(t)$ must be non-anticipating (causal), satisfy periodicity (net energy change over the cycle is zero), and adhere to a general convex feasible set $\Omega$ . This set $\Omega$ encompasses physical limitations such as power ratings, energy capacity, and ramp rates.
The Conflict:
- Centralized (Social Planner): Maximizes the expected reduction in social generation cost ( $\text{WEL}$ ).
- Decentralized (Private Operator): Maximizes expected arbitrage revenue ( $\text{REV}$ ).
Metric: The efficiency loss is quantified by the Price of Anarchy (PoA), defined as the ratio of the optimal welfare gain (centralized) to the welfare gain achieved under profit maximization (decentralized).

2. Methodology

The authors employ a mix of variational analysis, game theory, and approximation theory to derive bounds on the PoA across different pricing regimes.

Variational Inequalities: For linear price functions, the proof utilizes a function-space inner product $\langle X, Y \rangle = \mathbb{E}[\int_0^1 X(t)Y(t)dt]$ . By applying first-order optimality conditions (variational inequalities) to the concave objective functions of both the planner and the operator, the authors isolate the gap between the two objectives.
Potential Game Theory: In the multi-battery extension, the authors construct an exact potential function $\Phi$ for the game. They demonstrate that the game is strictly concave, ensuring the existence of a unique pure Nash equilibrium.
Bernstein Approximation: To analyze general convex price functions, the paper utilizes Bernstein polynomials to approximate convex functions, showing how unbounded efficiency losses in specific convex cases can be preserved in polynomial approximations.
Canonical Benchmarks: For non-linear pricing analysis, the authors restrict the setting to a "canonical benchmark" ( $\mathcal{I}_{\text{step}}$ ): deterministic, two-level step demand with a two-piecewise constant battery response and no structural constraints. This simplification allows for closed-form analysis of non-linear effects.

3. Key Contributions and Results

A. Linear Price Functions

The paper establishes that for linear price functions $P(z) = az + b$, the PoA is a structural invariant of 4/3.

Result: $\text{PoA} = 4/3$ .
Significance: This bound is tight and holds for any stochastic demand distribution and any convex operational constraint set $\Omega$ . It generalizes previous results that were limited to discrete two-period models. The bound arises solely from the misalignment of objectives, independent of demand volatility or constraint geometry.

B. General Convex Price Functions

The authors demonstrate that linearity is a sharp boundary for bounded efficiency.

Result: For general convex price functions, the PoA can be unbounded, even in the simplest deterministic step-demand setting.
Implication: Convexity alone is insufficient to guarantee market efficiency. The paper constructs a specific convex price function where the profit-maximizing operator shifts demand by a negligible amount compared to the social optimum, causing the welfare ratio to diverge. This result extends to convex polynomials via Bernstein approximation.

C. Monomial Price Functions

To bridge the gap between linear and general convex regimes, the paper analyzes monomial prices $P(z) = \alpha z^d$ .

Universal Bound: For any degree $d$ , the PoA is bounded by 2.
Specific Bounds:
- For quadratic prices ( $d=2$ ), the tight bound is 27/19.
- As $d \to \infty$ , the lower bound approaches $e/(e-1)$ .
Insight: While efficiency loss increases with the degree of non-linearity (curvature), it remains bounded for monomials, unlike general convex functions.

D. Multi-Battery Competition

The analysis is extended to $n$ identical competing batteries under linear pricing.

Equilibrium: The game admits a unique, symmetric pure Nash equilibrium.
Convergence: The PoA for $n$ batteries is given by $\frac{(n+1)^2}{n(n+2)}$ .
Result: As $n \to \infty$ , the PoA converges to 1. This indicates that competition fully restores social efficiency in the limit, providing a theoretical justification for encouraging competition among storage operators.

4. Significance and Claims

The paper claims to resolve open questions regarding the robustness of efficiency bounds in strategic storage markets.

Structural Invariance: The primary claim is that the $4/3$ PoA for linear pricing is not an artifact of simplified models (like discrete time or unconstrained operations) but a fundamental property of linear markets that persists in continuous time with arbitrary stochasticity and constraints.
Sharp Boundaries: The work delineates a precise boundary between pricing regimes that guarantee bounded efficiency (linear and monomial) and those that do not (general convex).
Competition as a Remedy: The paper provides a rigorous proof that in linear markets, increasing the number of competing batteries monotonically reduces efficiency loss, converging to the social optimum.
Limitations: The authors explicitly note that for general convex prices, efficiency guarantees are lost. They do not claim that convexity is always bad, but rather that without the specific structure of linearity or monomial forms, bounded efficiency cannot be guaranteed.

The paper concludes by identifying future directions, including extending the multi-battery analysis to non-linear pricing and incorporating transmission losses, but does not propose specific experimental implementations or policy interventions beyond the theoretical justification for competition.