Application of Big Data Analysis and Deep Neural Network for Oil and Gas Production Enhancement

Abstrak

Most of Indonesia’s major oil field are now entering marginal phase, characterized by lower production rate and rising number of production complications. Performing reservoir simulation in marginal field is not an effective way to increase production, as a lot of efforts has to be dedicated in matching reservoir performance with various attempts done in the wells, such as well stimulation, reperforation etc. Therefore, the abundance of data previously recorded during lifetime of wells in marginal field can be used as an asset to lift production without having to resort to resource-hungry reservoir simulation.

This research is based on an idea that a big cache of data from lifetime of a field can be analyzed using machine learning and deep neural network to pinpoint well problems, determine cause and effects of a certain treatment to a particular well, and mine valuable information from daily reports and other engineering related reports. The algorithm is first trained to recognize  language input from reports, where for older wells many types of languages and idioms exist and has to be properly recognized. The second step is recognizing well problems based on diagrams or treatments, and the last step is determining clusterization or categorization of wells based on problems and estimation of productivity increase after treatment by engineers. It is hoped that this approach can be a solution to improve productivity of marginal fields without having to perform lengthy periods of trial and error in conventional reservoir simulation.