An Inverse-Hybrid-Modeling Digital Twin System for Natural Gas Energy Metrology

LIU Bin; ZHONG Lu; FENG Quanyuan; CHEN Yihong

doi:10.11999/JEIT260289

Article Contents

Article Navigation > Journal of Electronics & Information Technology > 2026 >

LIU Bin, ZHONG Lu, FENG Quanyuan, CHEN Yihong. An Inverse-Hybrid-Modeling Digital Twin System for Natural Gas Energy Metrology[J]. Journal of Electronics & Information Technology. doi: 10.11999/JEIT260289

Citation:

LIU Bin, ZHONG Lu, FENG Quanyuan, CHEN Yihong. An Inverse-Hybrid-Modeling Digital Twin System for Natural Gas Energy Metrology[J]. Journal of Electronics & Information Technology. doi: 10.11999/JEIT260289

Citation:

PDF( 5056 KB)

An Inverse-Hybrid-Modeling Digital Twin System for Natural Gas Energy Metrology

doi: 10.11999/JEIT260289 cstr: 32379.14.JEIT260289

LIU Bin^{1, 2},
ZHONG Lu²,
FENG Quanyuan^{1
,
,},
CHEN Yihong¹

1.
College of Information Science and Technology, Southwest Jiaotong University, Chengdu 611700, China
2.
Chengdu Qinchuan IoT Technology Co., Ltd., Chengdu 610100, China

Accepted Date: 2026-04-23
Rev Recd Date: 2026-04-23

Available Online: 2026-05-23

Abstract

Abstract

Objective Global natural gas consumption continues to rise at an average annual growth rate of 3.2%. A 0.1% reduction in energy measurement error can reduce trade disputes by approximately $750 million per year. Traditional research primarily adopts indirect methods for energy measurement, with the mainstream chromatographic analysis and acoustic velocity correlation methods facing multiple bottlenecks and certain application limitations. Chromatographic analysis exhibits low anti-interference error but suffers from high flow dynamic response delay and insufficient dynamic calibration capability. Additionally, it has poor adaptability to multi-gas source switching, requires manual calibration, and has high usage and maintenance costs. The lack of interoperability standards for energy networks further exacerbates system integration difficulties. Although acoustic velocity correlation demonstrates low-latency flow dynamic response, it has high anti-interference error, which may increase significantly with single-component content changes (e.g., hydrogen content from 5% to 10%) and even fail under complex operating conditions (e.g., multi-gas source mixing, dynamic pressure fluctuations). To address these challenges, new mechanism modeling-oriented methods have emerged, with the two most representative research directions being “mechanism modeling-driven” and “hybrid modeling”. Both approaches integrate multi-source data fusion with virtual-real interaction to establish mechanism models between flow rate, other parameters, and energy, providing a new paradigm for accurate energy measurement, but new challenges have also arisen. The “mechanism modeling-driven” approach is based on static flow modeling using Computational Fluid Dynamics (CFD), but it has low dynamic parameter update efficiency (delay > 30 seconds), struggles to adapt to real-time operating condition changes, and relies on massive labeled data with insufficient interpretability. The “hybrid modeling” approach has unresolved issues in multi-module collaborative optimization. Furthermore, the core challenge of existing research lies in the lack of industrial-grade verification platform support, addressing the problems of dynamic response delay, parameter identification difficulties, excessive physical simplification, and weak anti-interference capability in traditional natural gas energy measurement methods under complex conditions. Building on the latest research findings of the “mechanism modeling-driven” and “hybrid modeling” approaches, this study innovatively introduces a variational autoencoder (VAE)-based operating condition feature extraction algorithm and a dynamic Bayesian network parameter calibration mechanism, combined with variational expectation-maximization (VEM) algorithm for offline calibration. It proposes a reverse hybrid modeling-driven digital twin system, which effectively solves the aforementioned problems in traditional natural gas energy measurement processes. Methods This study proposes a natural gas energy measurement digital twin system based on reverse hybrid modeling, which centers on a three-tier architecture of “algorithm-system-scenario”. It integrates calorific value, flow rate, and energy mechanism models with multi-source real-time data streams. A variational autoencoder (VAE) is introduced to achieve unsupervised operating condition feature mining, and a parameter self-correction loop is constructed by combining dynamic Bayesian network with variational expectation-maximization (VEM) system calibration. Industrial-grade devices such as ultrasonic flowmeters and gas chromatographs are integrated to ensure real-time data transmission and closed-loop control. The system covers core operating conditions such as dynamic pressure fluctuations, hydrogen-containing gas mixtures, and multi-gas source switching, ensuring a high degree of adaptability between the model and practical applications. Through 25 weeks of continuous verification on a full-scale industrial-grade experimental platform, the results show that the system has an operational delay ≤ 3.8 s, data transmission jitter ≤ 0.5 s, average daily energy consumption per device ≤ 1.2 kW·h, mean time between failures (MTBF) ≥ 4 100 h, energy measurement error ≤ 0.15%, calorific value error ≤ 0.12%, and flow rate indication error ≤ 0.2%. Meanwhile, the system meets security requirements through industrial Ethernet encryption and hierarchical access control, providing engineering support for intelligent pipeline network optimization and standardized integration. Results and Discussions First, a multi-level hybrid modeling framework is established: modular hybrid modeling is achieved through an algorithm-system-scenario three-tier architecture. Numerical methods combined with data are generally more flexible than purely analytical models and can be used to represent complex multi-physical systems because they employ fewer physical lumped parameterizations; moreover, under mechanical, energy, and hydrodynamic effects, these parameters may change during the energy measurement process. Deep integration of mechanism models and real-time data is realized through variational autoencoder (VAE) and dynamic Bayesian network, reducing parameter synchronization delay to 3.8 seconds, which effectively supports fluid-acoustic co-simulation and rapid response for complex working conditions such as hydrogen-containing natural gas. Second, an integrated algorithm for reverse hybrid modeling and system calibration is proposed: by introducing variational autoencoder (VAE), dynamic Bayesian network, and variational expectation-maximization (VEM) algorithm, a reverse hybrid modeling algorithm is constructed to form a self-supervised and adaptive intelligent system with inner closed-loop operation. The VAE encoder compresses high-dimensional operating condition data into low-dimensional feature vectors, enabling unsupervised feature extraction without massive labeled data. It can also automatically generate perturbed data similar to the input data based on learned internal distribution laws of the data, simulating abnormal operating conditions to verify anti-interference capability. Combined with dynamic Bayesian network, a continuous iterative cycle of "prior → evidence → posterior" is constructed to realize system self-correction and adaptive response to operating condition changes. The VEM algorithm specifically compensates for systematic errors that are difficult to cover by dynamic Bayesian networks, overcoming the limitations of traditional static models. Conclusions This study describes and gradually validates a hybrid digital twin system that combines experimental data-driven approaches with physical models, successfully simulating the physical characteristics of natural gas energy measurement. A full-scale test platform was constructed, and all major parameters of the system were rigorously validated through experimental measurement data and compared with industry benchmark data. Each independent module within the "algorithm-system-scenario" three-tier hybrid modeling architecture (including calorific value measurement, flow calculation, and energy conversion) underwent 25 weeks of continuous experimental validation, confirming a high degree of consistency between model predictions and actual measurements.On the established digital twin experimental platform for natural gas energy measurement, systematic validation was conducted on the three core functions: flow measurement under dynamic conditions, multi-component calorific value determination, and energy accumulation. The results demonstrated that the output of the digital twin model matched the physical device measurement data with over 99.5% accuracy. Notably, under complex operating conditions such as pressure pulsations and hydrogen-containing gas mixtures, the system maintained measurement accuracy within 0.5%, significantly outperforming traditional methods and meeting the Class A accuracy requirements for natural gas measurement.By introducing a multi-tier hybrid modeling framework, this study successfully addressed the challenges of parameter identification difficulties and excessive physical simplifications inherent in traditional natural gas energy measurement methods. The integration of Variational Autoencoders (VAEs), dynamic Bayesian networks, and Variational Expectation- Maximization (VEM) algorithms enabled unsupervised feature extraction for complex operating conditions and adaptive model parameter calibration, reducing reliance on prior physical knowledge and massive labeled datasets. Experimental evidence demonstrates that the proposed method maintains high precision and robustness even in complex scenarios, such as pressure pulsations and hydrogen-containing gas mixtures, where traditional models struggle to provide accurate descriptions.
- Natural gas energy metrology,
- Digital twin,
- Inverse hybrid modeling,
- Variational autoencoder (VAE),
- Dynamic bayesian network

FullText(HTML)

References(28)

References

[1]	SUN Qie, LI Hailong, MA Zhanyu, et al. A comprehensive review of smart energy meters in intelligent energy networks[J]. IEEE Internet of Things Journal, 2016, 3(4): 464–479. doi: 10.1109/JIOT.2015.2512325.
[2]	MONTUORI L, ALCÁZAR-ORTEGA M, VARGAS-SALGADO C, et al. Enabling the natural gas system as smart infrastructure: Metering technologies for customer applications[C]. 2020 Global Congress on Electrical Engineering (GC-ElecEng), Valencia, Spain, 2020: 96–100. doi: 10.23919/GC-ElecEng48342.2020.9286291.
[3]	ULBIG P and HOBURG D. Determination of the calorific value of natural gas by different methods[J]. Thermochimica Acta, 2002, 382(1/2): 27–35. doi: 10.1016/S0040-6031(01)00732-8.
[4]	LANG Xianming, LI Ping, GUO Ying, et al. A multiple leaks’ localization method in a pipeline based on change in the sound velocity[J]. IEEE Transactions on Instrumentation and Measurement, 2020, 69(7): 5010–5017. doi: 10.1109/TIM.2019.2955774.
[5]	SELEZNEV V. Computational fluid dynamics methods for gas pipeline system control[M]. Computational Fluid Dynamics. IntechOpen, 2010: 335–362. doi: 10.5772/7110. (查阅网上资料,未找到本条文献出版地,请确认).
[6]	KUNZ O and WAGNER W. The GERG-2008 wide-range equation of state for natural gases and other mixtures: An expansion of GERG-2004[J]. Journal of Chemical & Engineering Data, 2012, 57(11): 3032–3091. doi: 10.1021/je300655b.
[7]	连远锋, 田天, 陈晓禾, 等. 数字孪生辅助强化学习的燃气站场巡检任务分配算法[J]. 电子与信息学报, 2025, 47(7): 2285–2297. doi: 10.11999/JEIT241027. LIAN Yuanfeng, TIAN Tian, CHEN Xiaohe, et al. Gas station inspection task allocation algorithm in digital twin-assisted reinforcement learning[J]. Journal of Electronics & Information Technology, 2025, 47(7): 2285–2297. doi: 10.11999/JEIT241027.
[8]	刘宁波, 张子豪, 陈宝欣, 等. 海上目标多源数据特征提取与特征相关性分析[J]. 电子与信息学报, 2025, 47(10): 3745–3758. doi: 10.11999/JEIT250200. LIU Ningbo, ZHANG Zihao, CHEN Baoxin, et al. Features extraction and correlation analysis of multi-source data for maritime targets[J]. Journal of Electronics & Information Technology, 2025, 47(10): 3745–3758. doi: 10.11999/JEIT250200.
[9]	胡小方, 杨涛. 基于忆阻循环神经网络的层次化状态正则变分自编码器[J]. 电子与信息学报, 2023, 45(2): 689–697. doi: 10.11999/JEIT211431. HU Xiaofang and YANG Tao. Hierarchical state regularization variational autoencoder based on memristor recurrent neural network[J]. Journal of Electronics & Information Technology, 2023, 45(2): 689–697. doi: 10.11999/JEIT211431.
[10]	杨旗, 薛定宇. 基于双尺度动态贝叶斯网络及多信息融合的步态识别[J]. 电子与信息学报, 2012, 34(5): 1148–1153. doi: 10.3724/SP.J.1146.2011.01012. YANG Qi and XUE Dingyu. Gait recognition based on two-scale dynamic Bayesian network and more information fusion[J]. Journal of Electronics & Information Technology, 2012, 34(5): 1148–1153. doi: 10.3724/SP.J.1146.2011.01012.
[11]	ULBIG P and HOBURG D. Determination of the calorific value of natural gas by different methods[J]. Thermochimica Acta, 2002, 382(1/2): 27–35. (查阅网上资料,本条文献与第3条重复,请确认). doi: 10.1016/s0040-6031(01)00732-8.
[12]	中华人民共和国国家质量监督检验检疫总局, 中国国家标准化管理委员会. GB/T 17747.3-2011 天然气压缩因子的计算第3部分: 用物性值进行计算[S]. 北京: 中国标准出版社, 2012. General Administration of Quality Supervision, Inspection and Quarantine of the People's Republic of China and Standardization Administration of the People's Republic of China. GB/T 17747.3-2011 Natural gas—calculation of compression factor—Part 3: Calculation using physical properties[S]. Beijing: Standards Press of China, 2012.
[13]	国家市场监督管理总局. JJF 1993-2022 天然气能量计量技术规范[S]. 北京: 中国标准出版社, 2022. State Administration for Market Regulation. JJF 1993-2022 Metrological specification for the energy measurement of natural gas[S]. Beijing: Standards Press of China, 2022. <b>(查阅网上资料, 未找到本条文献出版年, 请确认)</b>.
[14]	胡晶晶, 杨照明, 苏怀. 基于深度学习的天然气热值动态预测[J]. 能源, 2023, 16(2): 799. doi: 10.3390/en16020799. HU Jingjing, YANG Zhaoming, and SU Huai. Dynamic Prediction of Natural Gas Calorific Value Based on Deep Learning[J]. Energies, 2023, 16(2): 799. (查阅网上资料,未找到本条文献中文信息,请确认). doi: 10.3390/en16020799.
[15]	王池, 李春辉, 王京安, 等. 天然气能量计量系统及方法[J]. 计量学报, 2008, 29(5): 403–406. WANG Chi, LI Chunhui, WANG Jing’an, et al. The system and method for energy measurement of natural gas[J]. Acta Metrologica Sinica, 2008, 29(5): 403–406.
[16]	VASKOVSKII S and BROKAREV I. Analysis of methods and systems for natural gas composition and energy characteristics determination[C]. 2023 5th International Conference on Problems of Cybernetics and Informatics (PCI), Baku, Azerbaijan, 2023: 1–6. doi: 10.1109/PCI60110.2023.10325991.
[17]	魏伟, 鲜平, 彭刚, 等. 天然气能量计量装置研究[C]. 2011 年机电科学、电气工程与计算机国际会议, 吉林, [s. n. ], 2011: 685–688. (查阅网上资料, 未找到本条文献信息, 请确认).
[18]	ZHAO Huichao, PENG Lihui, STEPHANE S A, et al. CFD aided investigation of multipath ultrasonic gas flow meter performance under complex flow profile[J]. IEEE Sensors Journal, 2014, 14(3): 897–907. doi: 10.1109/jsen.2013.2290863.
[19]	ZHAO Yuting, WEI qiang, LI Meng, et al. Design and implementation of an external clamped ultrasonic flowmeter based on time difference method[C]. 2021 IEEE International Conference on Computer Science, Electronic Information Engineering and Intelligent Control Technology (CEI), Fuzhou, China, 2021: 685–688. doi: 10.1109/CEI52496.2021.9574588.
[20]	LI Xiaobin, XIAO Bo, CHEN Xuejiao, et al. MSDF-VAE: A cloud–edge collaborative method for fault diagnosis based on transfer learning[J]. IEEE Internet of Things Journal, 2025, 12(12): 22393–22403. doi: 10.1109/JIOT.2025.3550916.
[21]	ZHANG Haizheng, SESHADRI R, PRAKASH A A, et al. Towards dynamic Bayesian networks: State augmentation for online calibration of DTA systems[C]. 2018 21st International Conference on Intelligent Transportation Systems (ITSC), Maui, USA, 2018: 1745–1750. doi: 10.1109/ITSC.2018.8569926.
[22]	RUAH C, SIMEONE O, HOYDIS J, et al. Calibrating wireless ray tracing for digital twinning using local phase error estimates[J]. IEEE Transactions on Machine Learning in Communications and Networking, 2024, 2: 1193–1215. doi: 10.1109/TMLCN.2024.3448391.
[23]	ZHANG Yan, SUN Wen, and ALCARAZ C. Editorial CFP: IEEE transactions on industrial informatics—special section on digital twin for industrial internet of things[J]. IEEE Transactions on Industrial Informatics, 2023, 19(5): 7214–7216. doi: 10.1109/tii.2023.3261167.
[24]	TYAGUNOV M G and ANDREEV V N. The use of digital twins in the energy sector: Selection and verification of tools for modeling hybrid energy supply systems based on renewable energy sources[C]. 2025 International Russian Smart Industry Conference (SmartIndustryCon), Sochi, Russian Federation, 2025: 572–577. doi: 10.1109/SmartIndustryCon65166.2025.10986090.
[25]	KABIR R, HALDER D, and RAY S. Digital twins for IoT-driven energy systems: A survey[J]. IEEE Access, 2024, 12: 177123–177143. doi: 10.1109/ACCESS.2024.3506660.
[26]	SARGENT R G. Verification and validation of simulation models: An advanced tutorial[C]. 2020 Winter Simulation Conference (WSC), Orlando, USA, 2020: 16–29. doi: 10.1109/WSC48552.2020.9384052.
[27]	JIA Wenlong, WANG Xiujuan, WU Xia, et al. A stable solution method for natural gas density across a wide temperature range using the GERG-2008 equation of state[J]. Fluid Phase Equilibria, 2025, 593: 114328. doi: 10.1016/j.fluid.2024.114328.
[28]	李军, 于波. 天然气性质、基本状态方程与相态关系[M]. 可持续天然气藏与开采工程. 北京: 海湾专业出版社, 2022: 1–28. (查阅网上资料, 未找到本条文献信息, 请确认).