AI-Driven Well Timelines for Well Optimization

Abstract

Artificial intelligence and machine learning algorithms provide energy companies with the possibility to digitally re-construct well histories, using both public and company specific historical well data. In this paper we discuss how oil and gas companies are creating a digital knowledge layer for oil and gas wells that provide a timeline of significant well events. Examples of key timeline events include, when drilling problems such as kicks happened, when blowout preventers were tested, when bottom hole pressures were taken, and when well interventions were done. This new generation of AI-driven applications are powered by a combination of a computational knowledge graphs and AI algorithms. These AI algorithms encode the expertise of subject-matter experts such as Petro-technical engineers and combine their experience with decades of historical well-events data extracted from databases, documents, and sensors to automatically create well event timelines. This technology enriches and combines companies’ internal siloed well data with public well data to create an integrated digital knowledge layer for wells. Engineers can optimize the life cycle of the wells by visually exploring this interactive timeline to understand and make decisions about the well. Petro-technical engineers have easy access to knowledge related to people, equipment, vendors, wells and more, so they can make better, more informed decisions faster. We show how we train the application's machine learning algorithms to read hundreds of thousands of historical reports to harvest knowledge about the well and store the extracted knowledge in an enterprise digital knowledge layer. By using the knowledge harvested and captured by this AI-driven application, experienced engineers can make better decisions that optimize the operations of their upstream assets.