Comparative Data-driven Enhanced Geothermal Systems Forecasting Models: A Case Study of Qiabuqia Field in China

Zhenqian Xue


This work shows that the injection rate and temperature of the circulating water, the number and half-length of fractures, and the well spacing affect the geothermal electricity. The production pressure has little impact on production. The KNN is impracticable in predicting geothermal development due to the occurrence of overfitting. The XGBoost is not highly acceptable since it presents worse stability. The SVM is not highly recommended because it illustrates bad adaptability. The ANN is highly recommended because of its best prediction accuracy, most stable performance and most adaptable ability. The machine learning algorithms significantly improve the efficiency of prediction by 2,700 times compared to the numerical simulation method.