Resource Allocation in Electricity Markets with Budget Constrained Customers

1. Problem Statement

The paper addresses the challenge of modeling electricity markets where customers face budget constraints. In traditional resource allocation models, social welfare is maximized by balancing user utilities against system costs, with prices emerging as dual variables (marginal costs) of power balance constraints.

However, when users have a hard budget limit $b_i$ on their expenditure, the problem becomes ill-defined in a standard primal formulation. The budget constraint takes the form $\lambda x_i \leq b_i$, where $\lambda$ is the market price. Since $\lambda$ is a dual variable determined by the equilibrium solution itself, the constraint is self-referential. This creates two main difficulties:

1. Non-convexity and Equilibrium Existence: It is unclear if a competitive equilibrium exists or is unique, as standard convex optimization assumptions may not hold when constraints depend on the solution.
2. Algorithmic Complexity: Standard distributed algorithms (like dual ascent) cannot be directly applied because the user's optimization subproblem depends on a price that is not yet known.

2. Methodology

The authors propose a two-pronged approach to resolve these issues: an iterative algorithmic solution and a primal reformulation.

A. Iterative Algorithm (Dual-Ascent Extension)

The authors extend the standard distributed dual-ascent algorithm used in unconstrained markets.

• User Subproblem: Given a price $\lambda$, each user $i$ solves:
  $$x_i^*(\lambda) = \arg \max_{x_i \geq 0} \{ u_i(x_i) - \lambda x_i \} \quad \text{s.t.} \quad \lambda x_i \leq b_i$$
  Due to strict concavity of $u_i$, this yields a closed-form solution:
  $$x_i^*(\lambda) = \max \left\{ 0, \min \left\{ (u_i')^{-1}(\lambda), \frac{b_i}{\lambda} \right\} \right\}$$
  This represents the minimum of the unconstrained demand and the budget-limited demand.
• System Operator Subproblem: The operator solves the standard supply problem:
  $$y^*(\lambda) = \arg \max_{y \geq 0} \{ \lambda y - C(y) \} = (C')^{-1}(\lambda)$$
• Price Update: The price is updated via a gradient-like step:
  $$\lambda_{k+1} = \lambda_k + \alpha_k \left( \sum x_i^*(\lambda_k) - y^*(\lambda_k) \right)$$

B. Primal Reformulation (Modified Utility)

To provide a rigorous theoretical foundation, the authors prove that the budget-constrained equilibrium is equivalent to solving a standard convex welfare maximization problem with modified utility functions $\hat{u}_i$.

• Construction of $\hat{u}_i$: The modified utility is constructed piecewise.
  • In regions where the budget is slack (unconstrained), $\hat{u}_i(x) = u_i(x)$.
  • In regions where the budget is binding, the marginal utility is replaced by the derivative of a logarithmic function: $\hat{u}_i'(x) = b_i/x$.
  • The function is smoothed at the crossover points (where $u_i'(x) = b_i/x$) to ensure continuous differentiability.
• Resulting Problem: The equilibrium is the unique solution to:
  $$\max_{x,y} \sum \hat{u}_i(x_i) - C(y) \quad \text{s.t.} \quad \sum x_i = y, \quad x_i \geq 0$$

3. Key Contributions

1. Existence and Uniqueness of Equilibrium: The paper proves (Theorem 1) that under standard assumptions (strictly concave utilities, strictly convex costs), there exists a unique competitive equilibrium price $\lambda^*$ that clears the market, even with budget constraints.
2. Convergence Proof: It is shown that the proposed iterative dual-ascent algorithm (Algorithm 1) converges to this unique equilibrium from any initial positive price, utilizing Lyapunov stability analysis.
3. Primal Equivalence (Theorem 2): The most significant theoretical contribution is the demonstration that the complex, self-referential budget-constrained problem is mathematically equivalent to a standard convex optimization problem with modified utility functions. This allows the application of existing convex optimization theory and distributed algorithms to budget-constrained markets.
4. Explicit Construction: The paper provides an explicit, piecewise construction of the modified utility functions, showing how they splice the original utility with a logarithmic function ($b \log x$) in the binding region.

4. Results and Validation

The authors validate their theory through numerical simulations:

• Quadratic Utilities: Simulations with two users having quadratic utility functions ($u(x) = \beta x - \alpha x^2$) demonstrate the convergence of the price iterate to $\lambda^* \approx 2.743$. The results show that budget constraints reduce consumption for constrained users and lower the equilibrium price compared to the unconstrained case.
• Square Root & Mixed Utilities: Simulations involving square root utilities ($u(x) = \alpha \sqrt{x}$) and a mix of five users with varying budgets confirm the framework's robustness.
• Visual Analysis: The paper illustrates that the modified utility functions effectively "cap" the growth of utility in the budget-binding region, replacing the original curve with a slower-growing logarithmic curve. This geometric interpretation explains why the equilibrium price adjusts to clear the market.

5. Significance

• Theoretical Rigor: The paper resolves the ambiguity of defining budget constraints in general equilibrium models by proving that a unique equilibrium exists and is computable.
• Practical Applicability: By transforming the problem into a convex primal form, the authors enable the use of efficient, scalable distributed algorithms for real-world electricity markets where large loads (e.g., AI data centers) or residential users face strict budget limits.
• Policy Insight: The modified utility framework offers a way to model consumer behavior without explicitly tracking the price in the constraint, simplifying market design and tariff analysis.
• Future Directions: The authors identify open problems, including incorporating network constraints (transmission limits), addressing fairness (proportional fairness), and modeling aggregate behavior of small budget-constrained agents.

In summary, this work bridges the gap between theoretical resource allocation and practical budget-constrained consumer behavior, providing both a computational algorithm and a rigorous mathematical equivalence that simplifies the analysis of such markets.