Formation of a training sample for machine learning

Abstract

In this paper, the Eastern section of the Chutyrsko-Kiengopskoye field was considered as an example. It was determined which input variables affect the oil production process. An algorithm for managing producing wells is described, where producing wells affect a different number of neighboring producing wells. An algorithm for controlling injection well pumps is described, where the operation of injection well pumps can occur in 4 modes. The principle of operation of producing and injection wells is described, where the 1083R producing well was considered as an example. Taking into account the described algorithms for controlling the operation of producing and injection wells, a training sample for machine learning was formed, which consists of 42300 examples.