Artificial neural network based prediction of reservoir temperature: A case study of Lindian geothermal field, Songliao Basin, NE China

Abstract

Reservoir temperature is a key parameter in geothermal researches. Existing geothermometers are based on equilibrium of water-rock interactions. Among them, the use of machine learning to predict reservoir temperature has attracted a lot of attention. In order to explore its practicality, the Lindian geothermal field was taken as the research area, 29 hot water samples were collected from the Lindian geothermal field and the temperature logging data from 11 geothermal wells at the time of geothermal well completed were collected for research. several methods and ANN methods were used to estimate reservoir temperature, and the prediction error was tested by the temperature logging data. The results show that the prediction error of artificial neural networks is the smallest, followed by Na/K geothermometer, and chalcedony and integrated multicomponent geothermometry approach are the largest. The reservoir temperature of the Lindian geothermal field is 40–85 °C. This suggests that artificial neural networks can be used as an accurate method for estimating reservoir temperature.