Scholarlink

<p>Reliable temperature forecasting in Enhanced Geothermal Systems (EGS) is essential for reservoir design and economic assessment, yet petroleum-based decline curves and many machine-learning (ML) surrogates do not enforce geothermal heat transfer. In addition, high-fidelity thermo-hydro-mechanical (THM) simulation remains computationally expensive. This study proposes a unified, physics-consistent framework that advances both decline-curve analysis (DCA) and surrogate modeling for downhole geothermal temperature forecasts. The classical Arps decline family is generalized for geothermal use by introducing an equilibrium-temperature term motivated by Newton-type cooling, ensuring finite late-time temperature limits while reducing exactly to the conventional Arps forms when the equilibrium term is set to zero. The extended decline curves are validated against Utah FORGE downhole temperature measurements and then used to construct learning surrogates on a controlled THM dataset spanning fracture count, well spacing, fracture spacing, host-rock thermal conductivity, and circulation rate. An equation-informed neural network embeds the modified decline equations as differentiable internal computational layers to produce full 0–60&#xa0;month temperature trajectories from design and operational inputs while preserving interpretable decline structure. A probabilistic Gaussian Process Regression surrogate is also developed for direct multi-horizon forecasting with calibrated uncertainty, while a direct XGBoost regression baseline provides a purely data-driven reference. Across the simulation dataset, the extended decline models reproduce temperature trajectories with near-perfect fidelity (median R<sup>2</sup> = 0.999; median RMSE = 0.071&#xa0;°C), and the equation-informed network achieves typical hold-out errors of MAE = 3.06&#xa0;°C and RMSE = 4.49&#xa0;°C. The Gaussian Process surrogate delivers the strongest predictive accuracy across 3–60&#xa0;month horizons (macro R<sup>2</sup> = 0.965; RMSE = 3.39&#xa0;°C; MAE = 2.34&#xa0;°C) with well-calibrated uncertainty, whereas the XGBoost baseline exhibits higher errors, underscoring the value of physical structure or probabilistic modeling for reliable geothermal temperature forecasting.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Physics-based decline curve analysis and machine learning for temperature forecasting in enhanced geothermal systems: Utah FORGE

  • 期刊论文

  • 发表日期：

    2026年6月3日

查看全文
  • Mina S. Khalaf

摘要

Reliable temperature forecasting in Enhanced Geothermal Systems (EGS) is essential for reservoir design and economic assessment, yet petroleum-based decline curves and many machine-learning (ML) surrogates do not enforce geothermal heat transfer. In addition, high-fidelity thermo-hydro-mechanical (THM) simulation remains computationally expensive. This study proposes a unified, physics-consistent framework that advances both decline-curve analysis (DCA) and surrogate modeling for downhole geothermal temperature forecasts. The classical Arps decline family is generalized for geothermal use by introducing an equilibrium-temperature term motivated by Newton-type cooling, ensuring finite late-time temperature limits while reducing exactly to the conventional Arps forms when the equilibrium term is set to zero. The extended decline curves are validated against Utah FORGE downhole temperature measurements and then used to construct learning surrogates on a controlled THM dataset spanning fracture count, well spacing, fracture spacing, host-rock thermal conductivity, and circulation rate. An equation-informed neural network embeds the modified decline equations as differentiable internal computational layers to produce full 0–60 month temperature trajectories from design and operational inputs while preserving interpretable decline structure. A probabilistic Gaussian Process Regression surrogate is also developed for direct multi-horizon forecasting with calibrated uncertainty, while a direct XGBoost regression baseline provides a purely data-driven reference. Across the simulation dataset, the extended decline models reproduce temperature trajectories with near-perfect fidelity (median R2 = 0.999; median RMSE = 0.071 °C), and the equation-informed network achieves typical hold-out errors of MAE = 3.06 °C and RMSE = 4.49 °C. The Gaussian Process surrogate delivers the strongest predictive accuracy across 3–60 month horizons (macro R2 = 0.965; RMSE = 3.39 °C; MAE = 2.34 °C) with well-calibrated uncertainty, whereas the XGBoost baseline exhibits higher errors, underscoring the value of physical structure or probabilistic modeling for reliable geothermal temperature forecasting.