Research Risk Factors in Monitoring Well Drilling—A Case Study Using Machine Learning Methods

This article takes an approach to creating a machine learning model for the oil and gas industry. task is dedicated most up-to-date issues of artificial intelligence. One goals this research was build predict possible risks arising in process drilling wells. Drilling wells production highly complex expensive part reservoir development. Thus, together with injury prevention, there goal save cost expenditures on downtime repair equipment. Nowadays, companies have begun look ways improve efficiency minimize non-production time help new technologies. To support decisions narrow frame, it valuable early warning system. Such decision system will engineer intervene prevent high expenses unproductive equipment due problem. work describes comparison algorithms anomaly detection during well drilling. In particular, make when determining geometry grid wells—the nature relative position injection at facility. Development systems are often subdivided into following: placement along symmetric grid, non-symmetric (mainly rows). The tested models classify problems based historical data from previously drilled validate algorithms, we used logs 67 large brownfield Siberia, Russia. Wells were selected analyzed. It should be noted that out wells, 20 without time. experiential results illustrate gradient boosting can complications better than other models.