An analysis of toe-to-heel air injection for heavy oil production using machine learning
Wei Wei, Alan Rezazadeh, Jingyi Wang, Ian D. Gates
Abstract
With strong potential for improving productivity, economics, and efficiencies, data analytics and machine learning are rapidly emerging in the oil and gas industry. However, their application in heavy oil production is scarce. In this study, we use clustering methods to analyze the Kerrobert Toe-to-Heel Air Injection (THAI) project to understand the inter-relationships of production variables and seek for optimal operating strategy to maximize production rate. More specifically, we use the K-means, normal mixtures and hierarchical clustering methods to determine how operating parameters contribute to oil production. The results reveal that, at current operation mode, air injection rate is constrained for production maximization due to the balance between heat generation and cooling of combustion system by excess air supply. The results also provide insight for improved productivity from THAI suggesting that cyclic injection may improve process performance. The new insights demonstrate that data analytics can provide new understanding of recovery processes as well as potential process improvement.