Abstract: We are interested in numerical schemes for the simulation of large scale gas networks. Typical models are based on the isentropic Euler equations with realistic gas constant. The numerical scheme is based on transformation of conservative variables in Riemann invariants and its corresponding numerical discretization. A particular novelty of the proposed method is the possibility to allow for an efficient discretization of the boundary and coupling conditions at nodal points of the network. The original discretization is analysed in view of its property to correctly recover steady states as well as to resolve possible analytic solutions. Comparisons with existing methods show the advantage of the novel method.
Introduction
The simulation of large-scale natural gas pipeline networks is a critical task for energy infrastructure management. These systems are typically modeled using the isentropic Euler equations, which describe the conservation of mass and momentum. However, a significant challenge arises at the "nodes"—the junctions where multiple pipes intersect. Standard numerical methods often struggle to maintain stability and physical accuracy at these points, frequently producing unphysical oscillations or failing to preserve steady-state conditions. This paper proposes a novel discretization scheme based on Riemann invariants to address these instabilities and improve computational efficiency.
Methodology
Transformation to Riemann Invariants
The core of the proposed method involves a change of variables. Traditional models work with conservative variables: density ($\rho$) and mass flux ($q$). The authors transform these governing equations into Riemann invariants ($w^\pm$). This transformation shifts the mathematical perspective to a "transport" framework, where quantities are viewed as being transported along the pipeline at specific characteristic speeds ($\lambda^\pm$).
By discretizing these Riemann invariants, the authors can employ an Upwind scheme. This approach is inherently more stable because the numerical stencil aligns with the physical direction of information flow. The paper mathematically demonstrates that this scheme is "well-balanced," meaning it can maintain a steady state exactly without introducing artificial numerical fluctuations.
Nodal Coupling and DAE Formulation
At the intersections of the network, the authors treat the system as a set of Differential-Algebraic Equations (DAEs). This formulation allows for the direct enforcement of physical coupling conditions, such as the conservation of mass and the requirement of equal pressure at junctions. By integrating these conditions directly into the solver framework, the method avoids the boundary-related instabilities common in traditional midpoint or endpoint discretization schemes.
Experimental Evaluation
The performance of the Riemann invariant-based scheme was evaluated using three distinct benchmarks:
- Single Pipe Scenario: A 3 km pipe used to validate basic pressure drops and mass flow transitions.
- Diamond-Shaped Network: A 9-pipe benchmark designed to test the scheme's ability to handle loops and multiple intersecting junctions.
- Large-Scale Topology: A realistic 46-node network (comprising 1 supply node and 23 demand nodes) to assess performance in a complex, real-world configuration.
Key Results and Discussion
Computational Efficiency
The proposed method demonstrates a massive improvement in computational speed. In a dynamic "wave scenario" simulating fluctuating demand, the novel method completed the simulation in 0.50 seconds. In contrast, a standard midpoint discretization required 202.57 seconds. This represents a speedup of approximately 400x, making the method highly suitable for real-time monitoring and optimization of gas networks.
Numerical Stability and Accuracy
One of the primary advantages of the Riemann invariant approach is the elimination of unphysical artifacts. While traditional midpoint schemes often produce jagged, oscillating results during abrupt pressure changes, the proposed method remains smooth and matches physical reality. In steady-state tests, the new method showed zero fluctuations, whereas both endpoint and midpoint methods exhibited persistent, non-decaying numerical noise. The scheme also successfully recovered analytic solutions across complex topologies like the Diamond and 46-node networks.
Conclusion
By leveraging the transport properties of Riemann invariants, this paper introduces a robust and highly efficient framework for gas network simulation. The method effectively resolves the long-standing issue of numerical instability at pipeline junctions and provides a significant reduction in computational overhead. While the current model assumes isothermal flow—a common simplification for friction-dominated long-distance pipelines—future work may involve extending this framework to non-isothermal flows and incorporating more complex real-gas laws.
Summary updated