The machine learning method in artificial intelligence is widely applied in various fields. In this study, three typical Imachine-learning algorithms are employed to mine and analyze drilling data of earlier drilled wells of an oilfield in Iraq where lost circulation is severe and then establish the risk prediction model of lost circulation while drilling. Geological characteristics and operational drilling parameters are both taken into consideration, and the risk level corresponds to loss rate. After collecting numerous data related to lost circulation, BP neural network, support vector machine and random forest algorithms are employed to conduct supervised learning. In addition, multiple trainings are conducted to meet the accuracy, and then the accuracy of different algorithms is compared. The prediction results of testing data indicate that the prediction model of lost circulation based on random forest algorithm has the best performance. Moreover, the importance estimation of variables based on random forest indicates that depth, mud density, filtration, pump pressure, flow rate and geostress have a significant influence on lost circulation.

This content is only available via PDF.
You can access this article if you purchase or spend a download.