Abstract
This paper establishes a theoretical framework for proving the existence of solutions to complex stochastic equilibrium problems, specifically stochastic quasi-variational inequalities (SQVIs) and complementarity problems (SCPs). The authors develop "integration-free" sufficiency conditions, allowing researchers to guarantee solution existence based on the properties of individual random scenarios rather than solving often-intractable expected-value integrals.
Introduction
In fields such as economics, energy, and engineering, competition and equilibrium are frequently modeled using Variational Inequalities (VIs). When these models incorporate uncertainty—such as fluctuating fuel prices or demand—they become Stochastic VIs (SVIs), where the underlying functions involve expectations.
The primary bottleneck in analyzing SVIs is the "expectation" term. Standard existence theorems typically require the analytical form of the expectation to verify properties like coercivity or monotonicity. However, in most real-world applications, these integrals cannot be solved in closed form. This paper addresses this gap by extending existence theory to quasi-variational inequalities (where constraints depend on decision variables), multi-valued mappings (handling nonsmoothness), and complementarity problems defined over cones.
Theoretical Framework: Integration-Free Conditions
The authors propose a shift from traditional "integration-based" proofs toward "integration-free" or "almost-sure" sufficiency conditions. Instead of analyzing the expected map $F(x) = \mathbb{E}[G(x, \omega)]$ directly, they examine the properties of scenario-specific maps $G(x, \omega)$.
By utilizing tools from variational analysis, such as Fatou’s Lemma and Lebesgue Convergence Theorems, the authors show that if certain properties (e.g., coercivity) hold for almost every random scenario $\omega$, these properties are inherited by the expected-value map. For multi-valued maps, they employ Aumann’s Measurable Selection Theorem to represent elements of the expectation as integrals of extremal selections.
Applications and Validation
The theoretical results are validated through two sophisticated application models:
- Nonsmooth Stochastic Nash-Cournot Games: A model of firm competition with piecewise smooth price functions, representing regulatory features like price caps.
- Networked Power Markets: A strategic equilibrium model involving multiple generation firms, an Independent System Operator (ISO), transmission constraints, and uncertain costs.
Key Results and Insights
The core insight of this work is the "Power of Almost-Sure": one does not need to compute the integral to know it is well-behaved. If the model's logic holds in almost every individual scenario, the expected-value version of the problem is guaranteed to have a solution.
Key results include:
- Generalized Existence Theorems: Formal proofs (Propositions 8, 9, 12, and 14) establishing existence for SQVIs with both single-valued and multi-valued maps.
- Complementarity Breakthrough: Theorem 16 provides a verifiable condition for the solvability of Stochastic Complementarity Problems over cones.
- Practical Verification: The authors successfully apply these theorems to power market models, providing a "green light" for practitioners to use numerical methods like Sample Average Approximation (SAA) with theoretical confidence.
Conclusion and Future Work
While the provided conditions are sufficient, they are not necessarily required; a problem might still possess a solution even if it fails these "almost-sure" tests. Future research directions include developing stability and sensitivity analysis for these stochastic regimes and exploring stochastic differential variational inequalities for time-dependent equilibrium problems.
Summary updated