This article, written by Editorial Manager Adam Wilson, contains highlights of paper SPE 150285, ’Decision Support and Workflow Automation for the Development and Management of Hydrocarbon Assets Using Multiagent Systems,’ by Amr S. El-Bakry, SPE, and Michael C. Romer, SPE, ExxonMobil Production Company; Peng Xu, Anantha Sundaram, and Adam K. Usadi, ExxonMobil Research and Engineering; H. Lane Morehead, SPE, ExxonMobil Upstream Research; and Mark L. Crawford, SPE, Bryce A. Holloway, and Charles A. Knight, ExxonMobil Information Technology, prepared for the 2012 SPE Intelligent Energy International, Utrecht, The Netherlands, 27-29 March. The paper has not been peer reviewed.
Asset-management workflow automation poses challenges because these workflows are multidisciplinary, cross functional, and expertise intensive. Advancements in the fields of artificial intelligence, software engineering, and decision sciences have led to the development of multiagent systems (MASs). Each agent functions as an entity of a distributed computing system, performing a broad range of tasks that may include advanced analytics, data-quality checks, and oversight of other agents. Although agents may have competing priorities, they can convey information to one another to broker a decision.
To improve the efficiency and effectiveness of asset management, it is essential to develop an integrated, flexible, and sustainable information-technology system that addresses these challenges. Such a system should integrate heterogeneous, distributed data sources and computer models (e.g., analytical models). It must process data automatically, converting them to high-level information understandable by people. It must be capable of developing asset- management plans, monitoring the execution of those plans, and adjusting them as necessary. Such a system should adhere to best-practice workflows and yet adapt to novel situations, allowing for easy customization to address evolving asset needs.
The MAS technology and design approach can satisfy these requirements in a scalable, flexible, and sustainable way. MASs are composed of multiple interacting intelligent software agents, each of which is a computer entity that is capable of autonomous action within its environment to meet its objectives.
An intelligent agent is an autonomous, goal-driven, problem-solving computational entity capable of effective operation in dynamic and open environments. It is a natural extension of modern software-design and -development best practices. Fig. 1a shows a schematic of an intelligent agent. Fig. 1b illustrates an agent using two methods to plan its actions. For simple scenarios, an agent acts in a reactive mode; it directly maps the state of the environment to an action. For more-complex scenarios, an agent operates in a deliberative mode; it uses more-sophisticated planning algorithms to identify a sequence of actions to achieve its goals.