Application of machine learning models to geothermal groundwater temperature prediction

Geothermal energy as a kind of clean energy has broad application prospects. The temperature assessment of geothermal water in sustainable development and utilization of geothermal resources is an important research topic. Artificial intelligence technology has become a hot spot and frontier direction in the exploration and development of mineral and oil and gas fields, but in the field of geothermal field development, there are few relevant studies. This paper first analyzes the important value of large data and artificial intelligence application in oil and gas field development, and then introduces the application of artificial intelligence in geothermal field development at present. Taking Xianyang geothermal field in Shaanxi province as an example, the single well geothermal water temperature time series model was constructed by using long and short term memory neural network (LSTM) under the predestined production mode. Random forest and XGBoost algorithm were used to predict the groundwater temperature of multiple geothermal wells. The accuracy of the three models was above 95%, and the running speed is fast. The depth at the top of the water intake section is the primary influencing factor of geothermal water temperature in this area. The model verifies that the fault zone plays an important role in heat storage.The application of the example verifies the superiority of machine learning in solving complex problems in geothermal field development, and the reasonable application of artificial intelligence technology can provide more effective decision-making basis for the efficient development of geothermal field and scientific cost reduction, quality improvement and efficiency improvement.