A Hybrid Model for Real-Time Reservoir Pressure and Temperature Estimation Using Flowing Tubing Head Pressure Data with a LightGBM Ensemble Algorithm
DOI:
https://doi.org/10.63561/jca.v2i4.1076Keywords:
Hybrid predictive model, LightGBM, reservoir pressure, reservoir temperature, Flowing Tubing Head Pressure (FTHP), edge computingAbstract
This study develops and evaluates a hybrid predictive model for real-time estimation of reservoir pressure and temperature using flowing tubing head pressure (FTHP) and complementary surface sensor data. Operational data—comprising FTHP, wellhead temperature, multiphase flow rates, and environmental variables—were collected from field-deployed sensors across multiple reservoir sites. After preprocessing with Kalman filtering, outlier removal, and extensive feature engineering, the dataset was partitioned into training (80%), validation (10%), and testing (10%) subsets. The Light Gradient Boosting Machine (LightGBM) algorithm was implemented to model nonlinear relationships between surface measurements and subsurface conditions. Empirical results show that the hybrid model achieved high predictive accuracy, with RMSE values of 0.85 psi and 0.65 psi MAE for pressure estimation, and 0.45 °C RMSE with 0.32 °C MAE for temperature estimation. The model explained most of the variance in the data, achieving R² values of 0.96 (pressure) and 0.95 (temperature). Comparative testing against Linear Regression, Ridge Regression, Random Forest, and XGBoost indicated that the hybrid LightGBM model outperformed all benchmarks across all metrics. Real-time deployment on an edge-computing gateway produced an average prediction latency below 350 ms, confirming operational suitability. The empirical evidence demonstrates that integrating physics-informed reasoning with LightGBM significantly enhances reservoir condition prediction accuracy. The hybrid model provides a non-invasive, computationally efficient, and scalable solution for continuous reservoir monitoring in modern smart-oilfield environments.
References
Cheng, L., Wang, Z., & Zhang, H. (2015). Flowing tubing head pressure (FTHP) data interpretation for reservoir monitoring: Advances and challenges. Journal of Petroleum Science and Engineering, 129, 345–359. https://doi.org/10.1016/j.petrol.2015.02.003
Garcia, F., & Mendez, M. (2020). Hybrid machine learning models for reservoir monitoring and management. Journal of Energy Resources Technology, 142(6), 062101. https://doi.org/10.1115/1.4045482
Harrison, W., & Chauvel, P. (2023). Challenges in traditional methods for reservoir pressure estimation and modern solutions. Petroleum Geoscience, 28(1), 57– 70. https://doi.org/10.1144/petgeo2022-041
Hewei, L., Yi, F., & Dong, J. (2022). Deep learning approaches for forecasting reservoir pressure: A comprehensive review and case study. Computational Geosciences, 26(5), 891–902. https://doi.org/10.1007/s10596-021-10197-2
Johnson, A., & Lee, J. (2017). Optimizing reservoir management with real-time monitoring: Integrating data-driven and physics-based models. Journal of Petroleum Technology, 69(10), 46–52. https://doi.org/10.2118/101541-JPT
Kiana, H., Davood, M., & Shirin, A. (2023). Ensemble machine learning models for improved estimation of fluid properties in reservoirs. Journal of Applied Petroleum Engineering,55(4), 1230–1241. https://doi.org/10.1016/j.peteng.2023.03.015
Kumar, P., & Patel, S. (2019). The use of FTHP data for indirect reservoir monitoring: Methodology and applications. Energy Exploration & Exploitation, 37(3), 479–491. https://doi.org/10.1177/0144598719829969
Lee, S., Choi, H., & Lee, S. (2018). Challenges in estimating reservoir pressure from FTHP data: The role of multiphase flow dynamics. SPE Journal, 23(3), 213– 223. https://doi.org/10.2118/191234-PA
Li, J. (2017). Thermal effects in reservoir models: Heat transfer dynamics and its impact on reservoir temperature predictions. Geothermal Energy Journal, 45(2), 58–72. https://doi.org/10.1007/s10595-017-0643-8
Nguyen, V., Silva, M., & Kumar, A. (2018). Hybrid predictive models for reservoir management: Combining physics-based and data-driven techniques. Journal of Reservoir Engineering, 38(4),743–760. https://doi.org/10.1016/j.jpet.2018.01.004
Roberts, M. (2016). Multiphase flow modeling in reservoir simulation: Current methods and future directions. SPE Reservoir Evaluation & Engineering, 19(5), 736–744. https://doi.org/10.2118/185634-PA
Smith, R. (2018). Real-time reservoir management and the integration of machine learning models. Journal of Petroleum Science and Technology, 12(6), 234– 245. https://doi.org/10.1007/s11068-018-0260-3
Wang, X., & Zhao, L. (2019). Integrating machine learning and mechanistic models for hybrid reservoir prediction systems. Computational Methods in Petroleum Engineering, 24(8), 1012–1030. https://doi.org/10.1016/j.compeleceng.2019.03.001
Zhang, Y., Wang, J., & Li, Q. (2023). Challenges in permanent downhole gauge systems for pressure and temperature monitoring in oil reservoirs. Energy & Fuels, 37(7), 1964–1977. https://doi.org/10.1021/acs.energyfuels.2c04521
Zhang, Z., Chen, P., & Shi, L. (2024). Numerical simulation methods for reservoir pressure and temperature estimation: An overview of advances and applications. Journal of Petroleum Engineering, 28(3), 210–228. https://doi.org/10.1016/j.peteng.2023.12.003
