Lower operating costs by predicting unscheduled downtime with machine learning.

This publication analyzes the loss of lost profits from downtime during the cessation of oil production due to equipment failure. Purpose of work is to reduce the number of downtime of drilling equipment during oil production and optimize operating costs; to develop a method for predicting changes in the reliability of downhole equipment; to identify and eliminate the reasons for the increase in operating costs when changing technological indicators of well operation. To solve this problem, the random forest machine learning algorithm was used in conjunction with predictive regulation of the wells using a digital double of the field and expert assessment. The implementation of the developed algorithms at oil producing bushes will reduce operating costs for the purchase of electric centrifugal pumps, reduce downtime of mining equipment and increase the life of working pumps. [ABSTRACT FROM AUTHOR]