Drilling is a critical operation in the geothermal wells to unlock the potential resources. However, considering the amount of data generated, it is very difficult to track and detect the anomalies in operational domain. In this paper, we have investigated the different data mining algorithms that can capture the pattern in the operational parameters considering the other petrophysical properties. In addition, a suitable framework is proposed for the assessing the patterns in the operational system. We used the FORGE well log data of already drilled wells and synthesize the evolution of dynamic data with 5 second interval. Now, on each segment of the data the suitable algorithms were implemented to identify and remove the effect of operational parameters using a series of digital filtering techniques. Then, the filtered version of well logs was used as input for supervised and unsupervised machine learning algorithms such as Expectation-Maximization Clustering, multivariate linear regression, deep learning models with hidden layers. Finally, hazardous zones are classified using the classifications which can improve the confidence in the operation. Such classifications can be an invaluable tool as raw operational data is difficult to visually identify and classify the anomalies. Overall the proposed framework can significantly improve the drilling operation in geothermal wells and can be further extended for real-time monitoring systems which is highly exhaustive job.

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

Copyright 2020, 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.