A Game-Theoretic Approach to Improve Energy-Related Data

With the increase in the energy demand, the magnitude of energy productionoperation increased in scale and complexity and went too far in remote areas. To manage such abigfleet, sensors were installed to send real-time data to operation centers, where subject matterexperts monitor the operations and provide live support. With the expansion of installed sensorsand the number of monitored operations, the operation centers wereflooded with a massiveamount of data beyond human capability to handle. As a result, it became essential to capitalizeon the artificial intelligence (AI) capability. Unfortunately, due to the nature of operations, thedata quality is an issue limiting the impact of AI in such operations. Multiple approaches wereproposed, but they require lot of time and cannot be upscaled to support active real-time datastreaming. This paper presents a method to improve the quality of energy-related (drilling) real-time data, such as hook load (HL), rate of penetration (ROP), revolution per minute (RPM),and others. The method is based on a game-theoretic approach, and when applied on the HL???one of the most challenging drilling parameters???it achieved a root mean square error (RMSE)of 3.3 accuracy level compared to the drilling data quality improvement subject matter expert's(SME) level. This method took few minutes to improve the drilling data quality compared to weeks in the traditional manual/semiautomated methods. This paper addresses the energy data quality issue, which is one of the biggest bottlenecks toward upscalingAI technology into active operations. To the authors'knowledge, this paper is thefirst attempt to employ the game-theoreticapproach in the drilling data improvement process, which facilitates greater integration between AI models and the energy live datastreaming, also setting the stage for more research in this challenging AI-data domain.