Project Pursues Autonomous Waterflooding Operations Driven by AI
Chris Carpenter
DOI: https://doi.org/10.2118/0524-0085-jpt
2024-05-01
Journal of Petroleum Technology
Abstract:_ This article, written by JPT Technology Editor Chris Carpenter, contains highlights of paper SPE 215236, “AI Innovations in Waterflood Management: The Path to Autonomous Operations,” by Sanjoy K. Khataniar, SPE, Shripad S. Biniwale, SPE, and Mohamed A. Elfeel, SPE, SLB, et al. The paper has not been peer reviewed. _ In the S Field described in the complete paper, a major control mechanism applied to optimize waterflooded reservoirs is controlling the water injection and pumping rates of producers. The reservoir surveillance team has been using a simple, spreadsheet-based analytical approach that proved limiting as the number of injection patterns increased. The complete paper presents various innovations in bringing real applications of artificial intelligence (AI) for waterflooding management. The AI-based solution combines cloud technologies, data processing, data analytics, machine-learning algorithms, robotics, sensor and monitoring systems, automation, edge gateways, and augmented and virtual reality. Application of AI to Waterflood-Performance Management The authors devote a subsection of the complete paper to a description of a process they term “Design Thinking.” The main steps in the Design Thinking process are discover, define, ideate, experiment, build prototype, and test. The approach and purpose of each of these steps is detailed in the complete paper, with examples of most steps discussed in this synopsis. Establishing a strategic foundation framework is critical for any business wishing to improve efficiency. A commonly used framework separates the business’s operational, tactical, and strategic aspects. In the complete paper, the authors present a conceptual realization of such a framework for waterflood-performance management. The attributes of this framework include time-horizon integration; user-friendliness; and a decision-focused, fit-for-purpose design. Results The opportunity for the authors to consider introducing AI into waterflooding operations arose in 2019 when a large operator in Ecuador began to consider leveraging digital technologies for all its field operations to generate better business value. Waterflood optimization was defined as encompassing all activities that optimize injectivity, capacity, and quality from water sourcing to injection at the sand face. Requirements were grouped into seven categories, and a high-level implementation road map was created to increase the digital maturity of assets as quickly as possible. Further analysis highlighted the fact that the existing injection-pattern balancing methodology was not robust enough and that addressing it on a priority basis would have the greatest effect. The S Field was identified as a candidate for solution prototyping. The S Field is one of several being developed and produced by waterflooding. The field has been in production since the early 1980s by natural depletion. Reservoir pressure has declined considerably to levels just above bubblepoint pressure. A redevelopment of the field is being undertaken with horizontal wells and pattern waterflooding. Injection has closely followed offtake, maintaining a voidage replacement ratio close to unity and allowing production offtake to increase threefold. The field has faced numerous operational and social challenges that have led to significant production losses and decreased operational efficiency. The following are the primary bottlenecks identified by the operations team: - Remote location - Concerns of the local community - Electrical power shutdowns - Flow-assurance events - Lack of well testing - Lack of measurement at surface