🔢 mathematics

Steady Flow of Natural Gas in Pipeline Networks via Solution of a Nonlinear Differential-Algebraic System of Equations

This paper presents a methodology for modeling steady-state natural gas flow in pipeline networks by retaining previously neglected gravitational and inertial effects within a nonlinear differential-algebraic system, which is solved using a Newton-Raphson algorithm enhanced with sensitivity ODEs and a collocation-based initialization scheme, ultimately demonstrating that while gravity is significant, inertial effects are negligible across various network scales.

Original authors: Shriram Srinivasan, Kaarthik Sundar

Published 2026-02-23

📖 4 min read🧠 Deep dive

CC BY 4.0

Original authors: Shriram Srinivasan, Kaarthik Sundar

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 a massive, invisible river of natural gas flowing through a sprawling web of steel pipes that crisscrosses the entire United States. Engineers need to know exactly how much pressure is at every point in this web to ensure the gas gets to your home safely and efficiently.

For decades, engineers have used a "shortcut" to calculate this. They assumed the gas flow was perfectly steady, the pipes were perfectly flat, and they ignored two tricky forces: gravity (pulling the gas down hills) and inertia (the gas's tendency to keep speeding up or slowing down). This shortcut turned a incredibly complex math problem into a simple algebra puzzle that computers could solve easily.

The Problem with the Shortcut
The authors of this paper asked a simple question: "What if the pipes aren't flat? What if they go up mountains and down valleys? Does ignoring gravity and inertia actually matter?"

They realized that when pipes go up and down, gravity acts like a heavy hand pushing the gas down or holding it back. Ignoring this is like trying to calculate how fast a car is going on a steep hill while pretending the hill doesn't exist.

The New, Harder Way
To get the real answer, the authors had to stop using the simple "shortcut" equations. Instead, they had to solve a much harder type of math problem called a Differential-Algebraic System.

Here is a simple way to visualize the difference:

The Old Way (Algebra): Imagine you are trying to guess the total distance of a trip. You just multiply speed by time. It's a single, static calculation.
The New Way (Differential-Algebraic): Imagine you are driving that trip, but the road is constantly changing slope, and the wind is pushing you. You can't just do one multiplication. You have to calculate your speed and position at every single tiny inch of the road, while also making sure you don't run out of gas at the junctions where roads meet.

How They Solved the Impossible
Solving this "inch-by-inch" problem for a network with thousands of pipes seemed impossible because the math gets "stiff" (like trying to steer a giant ship with a tiny rudder). The authors came up with three clever tricks to make it work:

The "Sensitivity" Map (Forward Sensitivity):
Imagine you are trying to find the perfect setting on a thermostat. You turn the dial a tiny bit and see how the temperature changes. The authors built a "map" that tells them exactly how a tiny change in pressure at the start of a pipe changes the pressure at the end. This map helps the computer navigate the complex math without getting lost.
The "Magic Lens" (Variable Transformation):
The numbers in their equations were behaving badly—some were huge, some were tiny, making the computer dizzy. They put on a "magic lens" (a mathematical transformation) that squashed the huge numbers and stretched the tiny ones so everything fit on the same scale. Suddenly, the computer could see the solution clearly.
The "Rough Sketch" (Two-Point Collocation):
When you try to solve a hard puzzle, you don't start with the final piece; you start with a rough sketch. The authors created a "rough sketch" of the solution by only checking the math at the very beginning and very end of each pipe (ignoring the middle for a moment). This gave them a great "first guess" to feed into their main computer program, which then refined the answer perfectly.

What They Found (The Big Surprise)
After running simulations on real-world data, including a massive network spanning eight US states, they discovered two things:

Inertia is a Ghost: The gas's tendency to speed up or slow down (inertia) was so weak compared to the friction of the pipe walls that it didn't matter at all. Engineers can safely ignore this force to save time.
Gravity is a Giant: However, gravity was a huge deal. If a pipe goes up a mountain, the pressure drops significantly more than the old "flat pipe" models predicted. If you ignore gravity, you might think you have plenty of pressure when you actually don't, which could lead to safety issues or supply shortages.

The Takeaway
This paper is like a new, high-definition GPS for gas pipelines. It proves that while we can ignore the gas's "momentum," we absolutely cannot ignore the hills and valleys it travels over.

The authors didn't just solve a gas problem; they built a new toolkit that can be used for any fluid network—whether it's oil, hydrogen, or hot water for heating cities—whenever the pipes go up and down and the math gets too messy for simple shortcuts. They showed us that with the right mathematical tricks, even the most complex, winding networks can be understood and managed safely.

1. Problem Statement

The simulation of steady-state gas flow in pipeline networks is critical for operations optimization, capacity estimation, and contingency planning. Traditionally, this problem is simplified by neglecting gravitational effects (due to elevation changes) and nonlinear inertial effects (convective acceleration). Under these assumptions, and often assuming an ideal gas equation of state (EoS), the governing partial differential equations reduce to a system of nonlinear algebraic equations.

However, this simplification limits the accuracy of simulations for:

Non-horizontal pipelines: Real-world networks span mountainous terrain and flat plains, where gravity can accelerate or resist flow significantly.
Non-ideal gases: Natural gas often requires non-ideal EoS (e.g., CNGA) for accurate density-pressure relationships.
Complex interactions: When gravity and non-ideal EoS are combined, the governing equations cannot be integrated analytically to yield closed-form algebraic expressions. Instead, they result in a nonlinear system of first-order Ordinary Differential Equations (ODEs) coupled with algebraic boundary conditions (mass balance at nodes and pressure continuity).

The Core Challenge: There is a lack of computational frameworks capable of solving these coupled nonlinear differential-algebraic systems (DAEs) for large-scale networks while rigorously accounting for gravity, inertia, and non-ideal gas behavior. Furthermore, the relative significance of inertial vs. gravitational effects in steady-state regimes has not been quantitatively delineated.

2. Methodology

The authors propose a computational framework to solve the steady-state flow problem without neglecting gravitational or inertial terms, even when a closed-form first integral does not exist.

A. Mathematical Formulation

Governing Equations: The flow is modeled using steady-state mass and momentum balance equations. The momentum equation includes friction, convective acceleration (inertia), and gravitational components ( $\rho g \sin \theta$ ).
Non-dimensionalization: The equations are non-dimensionalized using characteristic scales (Mach, Euler, and Froude numbers) to analyze the relative magnitude of terms.
Network Topology: The network is modeled as a graph $G=(V, E)$ $G = (V, E)$ where edges represent pipes and compressors, and nodes represent junctions.
- Pipes: Governed by ODEs relating pressure gradient to flow and elevation.
- Compressors: Modeled as algebraic constraints boosting pressure by a fixed ratio.
- Nodes: Enforced via mass balance (algebraic constraints).

B. Solution Strategy: Newton-Raphson with Forward Sensitivity

Since the ODEs cannot be integrated analytically for non-ideal gases with gravity/inertia, the authors treat the problem as finding the roots of a nonlinear system where the "residuals" are defined by the mismatch between the ODE solution and boundary conditions.

Residual Construction: Instead of a closed-form integral $F(p_i, p_j, f_{ij}) = 0$ , they define a residual function $F$ based on the numerical solution of the ODE from one end of the pipe to the other.
Jacobian Evaluation: To use the Newton-Raphson algorithm, the Jacobian matrix (partial derivatives of residuals with respect to pressures and flows) is required. The authors compute these derivatives by solving forward sensitivity ODEs alongside the main flow equations. This allows the method to preserve the continuous structure of the physics while using a discrete iterative solver.

C. Numerical Enhancements

Variable Transformation: To address poor scaling and stiffness in the ODEs during iterations, a variable transformation is proposed: $\pi = p^3$ . This transforms the pressure variable to handle potential negative values during intermediate Newton steps and improves convergence.
Two-Point Collocation for Initial Guess: Newton-Raphson convergence is highly sensitive to initial guesses. The authors propose a two-point collocation scheme (a coarse approximation of the ODE) to generate high-quality initial guesses. This scheme enforces the ODE to be satisfied only at the pipe endpoints, providing a robust starting point for the full ODE-based Newton iterations.

3. Key Contributions

Unified Framework for DAEs: Demonstrated how to solve coupled nonlinear differential-algebraic systems for pipeline networks using Newton-Raphson and forward sensitivity, bypassing the need for analytical integration of the ODEs.
Quantitative Assessment of Effects: Provided a definitive evaluation of the importance of gravity and inertia in steady-state gas flow, delineating regimes where these effects are negligible or critical.
Robust Numerical Techniques: Introduced a variable transformation ( $p \to p^3$ ) to alleviate scaling issues and a two-point collocation strategy to generate reliable initial guesses for large-scale networks.
Generalizability: Showed that the methodology is applicable beyond natural gas to other thermal-fluid networks (hydrogen, district heating, oil) with nonlinear constitutive relations.

4. Results

The methodology was validated through three case studies:

Case Study 1 (Validation): Applied to standard GasLib instances (horizontal pipes, ideal gas). The results confirmed that the two-point collocation scheme provides excellent initial guesses (residuals $< 10^{-2}$ ) and that inertial effects are negligible in horizontal, subsonic flows.
Case Study 2 (Single Pipe Analysis): Simulated a single 122 km pipeline (Yamal-Europe parameters) with varying inclinations ( $\pm 4^\circ$ $\pm 4^{\circ}$ ).
- Inertia: Found to be negligible for all inclination angles and both ideal/non-ideal EoS (pressure changes $< 0.1\%$ ).
- Gravity: Found to be significant. Pressure differences due to gravity were nonlinear functions of inclination, reaching up to $\sim 5\%$ for non-ideal gases at steep angles.
Case Study 3 (Large-Scale Network): Simulated the "Northwest" US pipeline network (9,228 km, 977 nodes, 41 compressors) with real elevation data.
- Gravity Impact: For $\sim 80\%$ of nodes, gravity caused a pressure difference of $< 2\%$ . However, for nodes near capacity limits, differences reached 7–8%.
- Conclusion: While inertia can be safely ignored, gravity must be included for high-fidelity simulations, especially in networks with significant elevation changes operating near capacity.

5. Significance

This work bridges a critical gap in pipeline simulation by moving beyond the restrictive assumption of algebraic simplifications.

Operational Safety: By quantifying the impact of gravity, the framework helps operators avoid unsafe pressure regimes in mountainous terrain that simplified models might miss.
Computational Efficiency: The proposed method proves that complex, stiff differential-algebraic systems can be solved efficiently for large networks without resorting to oversimplified physics.
Future Applications: The framework is directly applicable to emerging energy infrastructure, such as hydrogen transport and CO2 sequestration, where fluid properties and network geometries often defy simple algebraic modeling.

In summary, the paper establishes that while inertial effects are negligible in steady-state gas flow, gravitational effects are non-negligible and must be modeled using a robust differential-algebraic approach to ensure accuracy in large-scale, topographically complex pipeline networks.