Efficient Gas Source and Composition Tracking Method for Multi-Source Natural Gas Pipeline Networks

As natural gas pipeline networks continue to grow in scale, multiple subnetworks are interconnected to form large-scale networks, leading to a complex topology with multiple sources, sinks, and loops. Especially with the increasing attention on hydrogen energy, tracking and tracing gas sources and compositions has become a critical task in the situation of mixed transportation and the application of hydrogen and natural gas, posing greater challenges for solving simulation models that involve multifunctional calculations. In this study, a steady-state simulation model for natural gas pipeline networks is developed that integrates both hydraulic–thermal calculations and gas source/composition tracking functionalities. A rapid solution is achieved based on a three-layer decoupling strategy. To achieve more efficient and faster solutions for complex multidimensional simulation problems, an improved method combining Singular Value Decomposition (SVD) to enhance the robustness of the solution and logistic regression to accelerate the solution has been proposed. Robustness validation results indicate that compared to the conventional Newton–Raphson algorithm, the improved method achieves iterative convergence even when the relative deviation between initial guesses falls below 1 × 10–13. Furthermore, it accommodates a wider range of initial values farther from the true solution, substantially reducing the model’s sensitivity to initial guess precision and enhancing the robustness of the iterative solution process. The results of the solution acceleration indicate that the improved method demonstrates strong applicability across verification cases involving networks of varying scales. The acceleration efficiency is dependent on the scale of the pipeline network. While maintaining solution accuracy, the improved method achieves a 7.53% to 14.22% increase in solving efficiency, effectively accelerating the simulation calculation process.