A hybrid PSO-LSSVM framework for enhanced intelligent recognition of Vapor-Liquid Two-Phase flow patterns in geothermal production wellbores

Abstract

Accurate flash evaporation flow pattern recognition is critical for optimizing industrial processes in geothermal energy and chemical engineering. This study integrates signal processing with machine learning to advance this task. First, complementary ensemble empirical mode decomposition (CEEMD) was applied to extract intrinsic mode function (IMF) energy spectra from differential pressure signals, capturing dynamic phase-change features. Subsequent analysis revealed sample type and size significantly impact model performance: multi-parameter datasets (combining inlet temperature, flow velocity, and IMF energy) yielded optimal training, while increased sample volume improved accuracy across all models. Three machine learning models, support vector machine (SVM), least squares SVM (LSSVM), and particle swarm-optimized LSSVM (PSO-LSSVM), were compared. SVM showed high sensitivity to sample quality; LSSVM enhanced stability but remained limited. PSO-LSSVM, however, outperformed both by leveraging PSO to optimize hyperparameters, achieving 97% test accuracy with robust generalization. It effectively distinguished flow patterns (single-phase, bubble, slug-churn, annular) using multi-feature inputs, even for complex regimes. This work provides a reliable technical framework for real-time flash evaporation monitoring and severity assessment in production wells, bridging lab insights with industrial application.

Document Type: Original article

Cited as: He, S., Li, J., Guan, R., Wang, Z., Gong, L., Zhang, T., Sun, S. A hybrid PSO-LSSVM framework for enhanced intelligent recognition of Vapor-Liquid Two-Phase flow patterns in geothermal production wellbores. Computational Energy Science, 2025, 2(3): 98-110. https://doi.org/10.46690/compes.2025.03.03

References

Caraghiaur, D., Anglart, H. Drop deposition in annular twophase flow calculated with Lagrangian Particle Tracking. Nuclear Engineering and Design, 2013, 265: 856-866.

Camargo, T. F. B., Paladino, E. E. Characterization of transient differential pressure signal features and flow pattern identification in horizontal two-phase flow through a constriction with machine learning models. Flow Measurement and Instrumentation, 2025, 106: 102985.

Dong, Y., Xue, L., Song, X., et al. Multi-physics coupling mechanisms and key development factors in enhanced geothermal systems for hot dry rock. Energy. 2025, 337: 138533.

Elgaddafi, R., Ahmed, R., Kiran, R., et al. Modeling and experimental studies on high-velocity slug and churn gasliquid upward flows in vertical pipes. Geoenergy Science and Engineering, 2025, 257: 214210.

Gorjaei, R. G., Songolzadeh, R., Torkaman, M., et al. A novel PSO-LSSVM model for predicting liquid rate of two phase flow through wellhead chokes. Journal of Natural Gas Science and Engineering, 2015, 24: 228-237.

Huang, N., Shen, Z., Long, S., et al. The empirical mode decomposition and the Hilbert spectrum for nonlinear and non-stationary time series analysis. Proceedings of the Royal Society of London. Series A: Mathematical, Physical and Engineering Sciences, 1998, 454(1971): 903-995.

Liu, Y., Zhao, Q., Chong, D. Review of flow regime transition criteria for adiabatic co-current upward gas-liquid flow in vertical multi-scale channels. Measurement: Energy, 2025, 8: 100066.

Liu, W., Zhang, S., Chen, Z., et al. A fully coupled thermohydro-mechanical model for investigating heat extraction mechanisms in enhanced geothermal systems. Applied Thermal Engineering, 2025, 280: 128459.

Liu, W., Xu, Q., Guo, L. Pressure fluctuation characteristics and wavelet analysis of gas-liquid two-phase flow in gathering riser. Journal of Engineering Thermophysics, 2019, 40: 1820-1825. (in Chinese)

Marcel, C. P., Rohde, M., Van Der Hagen, T. H. Experimental investigations on flashing-induced instabilities in one and two-parallel channels: A comparative study. Experimental Thermal and Fluid Science, 2010, 34(7): 879-892.

Ma, T., Wang, T., Wang, L., et al. A hybrid deep learning model towards flow pattern identification of gas-liquid two-phase flows in horizontal pipe. Energy, 2025, 320: 135141.

Singh, M., Mahmoodpour, S., Singh, M., et al. Impact of thermo-hydro-mechanical process coupling for CO2-water multiphase flow during geothermal energy extraction. Energy, 2025, 334: 137831.

Singh, S., Chakraborty, P. R., Kothadia, H. B. Flash evaporation in a superheated liquid pool using water as medium. International Journal of Thermofluids, 2024, 24: 100939.

Tong, W., Zhu, G. Flow pattern recognition method of gasliquid two-phase flow based on multilayer perceptron. Thermal Power Engineering, 2020, 35: 116-122. (in Chinese)

Yan, Y., Li, X. Ito, K. Numerical investigation of indoor particulate contaminant transport using the Eulerian-Eulerian and Eulerian-Lagrangian two-phase flow models. Experimental and Computational Multiphase Flow, 2020, 2: 31-40.

Wang, H. Prediction of gas-liquid two-phase holdup based on SOA-optimized LSSVM. Petroleum Engineering Construction, 2023, 49: 25-30. (in Chinese) Wang, L., Li, H., Bu, X. Thermo-economic investigation of binary flashing cycle for enhanced geothermal system. Geothermics, 2021, 89: 101951.

Wu, Z., Huang, N. E. Ensemble empirical mode decomposition: A noise-assisted data analysis method. Advances in adaptive data analysis, 2009, 1(1): 1-41.

Yuan, S., Zhang, L., Li, T., et al. Study of Gas-Liquid TwoPhase Flow Characteristics at the Pore Scale Based on the VOF Model. Energies, 2025, 18(2): 316.

Zaborowska, I., Grzybowski, H., Mosdorf, R. Boiling flow pattern identification using a self-organizing map. Applied Sciences, 2020, 10(8): 2792.