A Performance Evaluation of DEIM-Based Models for Natural Gas Pipeline Network Analysis Under Abnormal Measurement Conditions

This study proposes a data-driven reduced-order modeling framework for reconstructing the full pressure field using sparse sensor information. Accurate state estimation of natural gas pipeline systems is a fundamental requirement for ensuring operational safety, maintaining supply reliability, and supporting advanced monitoring and control. The presence of sensor fault and missing measurements, however, often degrades the quality of full-field reconstructions in practical applications. Proper Orthogonal Decomposition(POD) is first applied to extract dominant spatial modes, and the Discrete Empirical Interpolation Method(DEIM) is used to select optimal sensor locations based on the extracted modes. Dynamic Mode Decompositon with control(DMDc) is then employed to model system dynamics while incorporating control inputs. The full state dynamics are recursively reconstructed using only the selected sparse measurements. Quantitative evaluations, based on root-mean-square error(RMSE) metric, demonstrate that the proposed DEIM-DMDc framework enables accurate and robust pessure field reconstrution even under sensor noise and data loss, offering a promising approach for reduced-sensor monitoring of large-scale natural gas infrastructures.