Digital Solution to Extend the Life of Wells with Continuous Corrosion Monitoring based on Machine Learning Algorithms

Abstract

Well integrity in the oilfield is one of the challenges that petroleum engineers face, as they seek to monitor well corrosion in the field to optimize well performance. Most of these fields can be categorized as brownfields, with some of the wells considered aged and have expected integrity issues. To achieve sustainable production targets with cost-effective and safe operations from these fields requires a close monitoring of the integrity of all elements involved in the production chain. Addressing these challenges requires the engineers to coordinate and analyze several data elements, including casedhole, openhole, reservoir, well, and production data from multiple sources. Another challenge is to create and automate a corrosion workflow that saves the engineers’ time and improves efficiency. In this paper, we introduce an innovative workflow that uses the historical corrosion data while integrating the multiple production and reservoir variables. The innovative approach uses machine learning (ML) algorithms to provide a powerful tool for workover (W/O) candidate selection and for optimizing the corrosion evaluation frequency, which are required in different areas of the fields. Different ML methods (random forest classification and neural net) were applied on training data. Different models were created, and the best model will be used. This offered key insights on the rate of corrosion and corrosion patterns. Further, the developed workflow was designed to be self-sustaining and acting as a surveillance tool for monitoring the integrity of the wells. The first step of the workflow was to start with organizing and auditing the available corrosion data, followed by a review and analysis of existing openhole, casedhole, production, and reservoir engineering data. This approach led us to understand the extent and severity of corrosion in terms of the corrosion rate and the corrosion index. The corrosion logs were digitally interpreted depth-wise in order to explore the maximum metal loss for each interval. New animated conformance corrosion maps were created. The successful diagnosis through data analytics in a modern integrated software platform will assist in corrosion monitoring and decision-making. The multiple corrosion maps can be animated to visualize the current corrosion profile and predict the corrosion over time, in addition to ranking the wells for W/O candidate selection.