| Title | 3D Modeling of Terrestrial Temperature and Geothermal Resource Assessment throughout China Using Physics-informed Neural Network |
|---|---|
| Authors | Wei Runqian, Katsuaki Koike, Taiki Kubo |
| Year | 2026 |
| Conference | World Geothermal Congress |
| Keywords | |
| Abstract | Geothermal energy is known to be useful for specific purposes including electricity generation, agriculture, and aquafarming. With the increasing demand for renewable energy sources, geothermal energy is recognized as a promising source of energy. Although the utilization of geothermal heat is popular in China, the use of geothermal electricity is on a much smaller scale due to the lack of detailed modeling of temperature-at-depth, heat source distribution, and potential of available energy. Although well-temperature logging data are the most fundamental for the estimation of temperature-at-depth, available open-source data are extremely limited in China. For these problems, this study aims to clarify the 3D distribution of terrestrial temperature throughout China using a physics-informed neural network (PINN) with consideration of multiple crustal properties such as heat flow, heat conductivity of rock, Curie point depth, gravity anomaly, etc. These datasets can supplement the sparseness of well-temperature logging data. Heat conduction is used as the governing equation in PINN, and the interpolation and extrapolation of the temperature-at-depth are implemented through an ensemble learning method based on a combination of Random Forest, XGBoost, and LightGBM by setting the surface temperature and Curie point depth to boundary conditions. The resultant 3D temperature model specifies high potential zones of geothermal resources and promising areas of resource use. Furthermore, correlations of the temperature model with the geologic units and structures are discussed, and the main control factors on the high and low temperature anomalies are identified by a sensitivity analysis. |