The thermal conductivity of rocks is a strategic component to evaluate the potential of a geothermal resource. This rock property represents the rate of the amount of heat transferred by conduction through a cross-sectional area, characterizing the ability of the rock to transmit heat. Thermal conductivity is measured directly in the laboratory on core samples or using numerical methods and correlations from electric logs. Although these methods can provide thermal conductivity information, they are expensive and sometimes difficult to obtain. During drilling geothermal wells, surface sensors are collecting drilling parameters in real-time. That data has been mainly used to provide information about drilling conditions, prevent or determine potential risks, and for drilling optimization. This study focuses on surface drilling data in a machine-learning workflow to predict the thermal conductivity at the bit while drilling. In this study, we have used a public data set from Utah FORGE geothermal wells project. In this workflow, the main objective is to predict the thermal conductivity using various surface-drilling data variables, such as rate of penetration, weight on bit, torque, flow rate, and others that are usually real-time monitored. Actual thermal conductivity values directly obtained from samples were used to train and test the data set. Supervised regression algorithms were used to link real-time drilling data to thermal conductivity. The algorithms predict the thermal conductivity in the test well with a precision above 80%. Although this methodology was built based on information from the Utah FORGE project, it can potentially be extended to other fields. The results of this study are a small step to demonstrate the capabilities of the use of data analytics to obtain valuable information at a relative low cost.

Press the Back button in your browser, or search again.

Copyright 2021, Stanford Geothermal Program: Readers who download papers from this site should honor the copyright of the original authors and may not copy or distribute the work further without the permission of the original publisher.