مقالات فنی

Data-driven modeling of outlet temperature in fractured granite cores with implications for geothermal reservoir heat extraction

Heat transport within fractured geothermal reservoirs is controlled by intricate, nonlinear relationships among thermal conditions, hydraulic behavior, and geological characteristics. These coupled processes make traditional prediction approaches less reliable. In ...

Heat transport within fractured geothermal reservoirs is controlled by intricate, nonlinear relationships among thermal conditions, hydraulic behavior, and geological characteristics. These coupled processes make traditional prediction approaches less reliable. In the present research, machine learning techniques were employed to estimate the outlet temperature of water passing through a single fracture in a granite core sample. The models were trained using an experimental dataset consisting of 120 observations. The considered input features include confining temperature, confining pressure, fluid flow rate, fracture aperture, and inlet water temperature. Correlation analysis shows that confining temperature (0.91) and inlet temperature (0.86) have the strongest positive influence on outlet temperature, while flow rate exhibits a moderate negative correlation (−0.38), and fracture aperture (−0.07) and confining pressure (0.06) have minimal effects. SHAP analysis confirms confining temperature as the most influential parameter. Various models, including Decision Tree, Random Forest, AdaBoost, KNN, CNN, SVR, MLP-ANN, and Ensemble Learning, were evaluated. Optimal hyperparameters include depth = 10 for Decision Tree, 41 estimators for AdaBoost, depth = 8 for Random Forest, k = 1 for KNN, and C = 50 for SVR. Ensemble Learning and Random Forest achieved the best performance, with total R2 values of 0.997 and 0.995, respectively, along with low MSE and AARE%. The results demonstrate that such models provide highly accurate and reliable predictions, offering effective tools for evaluating and optimizing heat transfer performance in geothermal reservoir systems.

درباره نویسنده
مهدی عباسی
مدیر پروژه شرکت KPE
مقالات مرتبط