Multiplicity of equilibria in conjectural variations models of natural gas markets

Tobias Baltensperger, Rudolf Marcel Füchslin, Pius Krütli, John Lygeros

Abstract

Spatial partial equilibrium models are essential tools for analyzing energy markets, yet they often suffer from a "hidden trap": the existence of multiple valid equilibria. This paper translates a natural gas market model into a Linear Complementarity Problem (LCP) to rigorously determine the uniqueness of market variables. Our analysis reveals that while prices and production volumes are unique, trade routes and infrastructure flows often are not. We show that modeling traders with even minimal market power can resolve sales volume ambiguity, providing a critical safeguard for policy analysts.

Introduction

The global natural gas market is a complex, capital-intensive oligopoly. To study its dynamics, researchers frequently employ Spatial Partial Equilibrium models with Conjectural Variations (SPE-CV). These models simulate the strategic interactions between producers, infrastructure operators, traders, and consumers. However, a significant challenge persists: under standard assumptions, these models can produce multiple, equally valid market states for a single set of parameters. This multiplicity poses a risk to infrastructure planning and policy recommendations, as a single model run may yield an arbitrary solution from a solver.

Model Formulation and Methodology

The authors developed a deterministic SPE-CV model of the global gas market, encompassing producers, pipelines, LNG shipping, storage, and consumers. To analyze the uniqueness of solutions, the model was reformulated as a Linear Complementarity Problem (LCP).

The methodology involved:

  1. LCP Translation: By assuming affine (linear) and quadratic cost functions, the complex non-linear system was represented as an LCP.
  2. Matrix Analysis: By examining the core matrix of the LCP, the authors proved that while the overall solution is not globally unique (due to the matrix being positive semi-definite rather than a P-matrix), specific components like wholesale prices and total production are strictly unique.
  3. Bounding the Solution Space: For non-unique variables, the solution space forms a convex polyhedron. Linear Programming was used to determine the extreme bounds (minimum and maximum values) of this space, mapping the full range of possible physical flows.

Key Insights: Market Power and Uniqueness

A central finding is the role of market power in stabilizing model outputs. Under perfect competition, the math cannot definitively decide who sells to whom, leading to significant variance in predicted trade flows. However, when traders are modeled with even a small degree of market power ($\theta > 0$), they are mathematically incentivized to diversify their consumer base, which guarantees a unique equilibrium for sales volumes.

Conversely, infrastructure routing remains inherently ambiguous. If multiple routes (e.g., different pipelines or LNG paths) have identical marginal costs, the model treats them as interchangeable. This means that while the amount sold is certain, the path it takes may not be, creating multiple valid equilibria for physical infrastructure utilization.

Empirical Results

The theoretical framework was applied to a 2012 global natural gas market dataset covering 50 nodes and 291 trade routes.

  • Impact of Competition: In simulations of perfect competition, sales volumes between specific traders and consumers varied by as much as 186 million cubic meters per day (mcm/d)—a massive discrepancy considering Germany's winter consumption is roughly 293 mcm/d.
  • Resolving Ambiguity: Introducing a slight market power parameter reduced this sales volume ambiguity to exactly 0 mcm/d.
  • Infrastructure Flexibility: Even with unique sales, physical flows through infrastructure (pipeline routing, storage injection/extraction) varied by 21 to 171 mcm/d, highlighting the inherent flexibility in the global network.

Limitations and Future Work

The mathematical proofs presented rely on specific structural assumptions, such as affine infrastructure cost functions and quadratic production costs. More complex, non-convex structures might challenge these uniqueness guarantees. Future research could extend this framework to stochastic models or multi-carrier energy systems. The authors suggest that modeling software should evolve to automatically explore and report the full solution space, embracing the inherent ambiguity of infrastructure flows rather than forcing a single, potentially arbitrary, route.

Conclusion

This research highlights the danger of relying on a single model run in energy market analysis. For robust infrastructure planning, analysts must account for the multiplicity of equilibria. The paper recommends that future modeling efforts should either incorporate market power to ensure sales uniqueness or explicitly report the full range of possible infrastructure flows to inform decision-makers of the system's inherent routing flexibility.

Summary updated

BibTeX
@article{baltensperger2016multiplicity,
  author = {Tobias Baltensperger and Rudolf Marcel Füchslin and Pius Krütli and John Lygeros},
  title = {{Multiplicity of equilibria in conjectural variations models of natural gas markets}},
  journal = {Eur. J. Oper. Res.},
  year = {2016},
  doi = {10.1016/j.ejor.2016.01.032},
}
Figure