Accurate multistep-ahead prediction of natural gas pipeline-network states is essential for intelligent dispatching, yet such networks contain physically heterogeneous components (gas sources, pipelines, compressors, valves), and historical states and future dispatching commands are decoupled in both temporal scale and physical semantics. This causes conventional data-driven models to suffer from semantic entanglement and cumulative error during long-horizon forecasting. This study proposes a deep spatiotemporal surrogate model with three coordinated designs: (i) type-specific feature encoding combined with global latent-graph mapping and a shared graph convolutional network (GCN) to disentangle heterogeneous-equipment attributes and represent network-wide topological coupling; (ii) a residual-gated temporal coupling mechanism that adaptively fuses historical operating inertia with future external disturbances; and (iii) a temporal-gradient multi-objective loss with a 12-step autoregressive rolling strategy over a 6 h horizon to suppress cumulative divergence. On 85,248 samples built from field monitoring data and commercial mechanistic simulations, the model attains median relative errors of 1.15% for nodal pressure and 2.10% for pipeline flow, capturing macroscopic pressure decay and high-frequency transient flow induced by valve and compressor switching without noticeable delay, providing an efficient tool for online simulation, real-time warning, and decision support in complex natural gas pipeline networks.
Accurate healthy-state performance modeling under multiple operating conditions is essential for the condition assessment of industrial centrifugal compressors. However, conventional healthy-state baselines often struggle to meet the requirements of real-time condition assessment for centrifugal compressors operating under complex real-gas and multi-condition environments. To address this issue, this study proposes a two-layer framework combining a CFD-based physical baseline with data-driven residual correction using limited factory data. A three-dimensional full-machine CFD model was reconstructed and validated under real-gas conditions, then used to generate 1440 healthy-state operating points. XGBoost, LightGBM, Random Forest, and multilayer perceptron were evaluated as baseline surrogate models. A residual-correction model was subsequently trained to compensate for systematic discrepancies between CFD predictions and actual machine performance. Ablation tests compared the CFD baseline, a factory-data-only model, and the proposed hybrid model. Online computation requires only surrogate inference and residual correction, achieving an inference latency of 0.6585 ms per operating point on Intel64 Family, compatible with the 60 s SCADA sampling interval. After correction, the maximum errors in power, polytropic head, and polytropic efficiency were 0.611%, 0.481%, and 0.899%, respectively. Post-overhaul field validation yielded maximum errors of 1.650%, 3.048%, and 1.708% for outlet pressure, power, and polytropic efficiency. The framework provides a physically grounded and computationally efficient healthy-state reference, although its transferability requires validation using additional station-specific data.
This study investigates emergency scheduling optimization for natural gas networks under unexpected abnormal conditions, such as unplanned valve closures or equipment failures, which may trigger pressure alarms. Unlike normal planned operations, emergency scheduling requires rapid response and accounts for the dynamic lag of gas flow. A mixed-integer programming model is formulated to ensure system safety, satisfy pressure constraints, and minimize compressor energy consumption. To solve the high-dimensional nonlinear problem, a variable neighborhood search algorithm is proposed, which constructs initial feasible solutions via a greedy proximity-based approach and iteratively improves them using four “delete-compensate” neighborhood operators with the Metropolis acceptance criterion. Validation on a real natural gas network containing 645 stations, 69 pipelines, and 26 compressor stations, together with repeated computational experiments under two representative abnormal scenarios, demonstrates the effectiveness and applicability of the proposed method.