Transforming Hydrocarbon Production with Digital Technologies

Industry Context

• Global population and prosperity growth is expected to double electricity demand by 2050, requiring all forms of energy production including traditional hydrocarbon production.
• Hydrocarbon demand is projected to grow by nearly 47% by 2050.
• Key challenges in growing hydrocarbon production:
  • Natural production decline from existing reservoirs
  • High capital costs for drilling new wells (up to $100 million per offshore well)

Collaboration between AWS, ExxonMobil, and Baker Hughes

• AWS, ExxonMobil, and Baker Hughes have formed a strategic collaboration to apply digital technologies to solve the hydrocarbon production growth problem.
• Goals:
  • Maximize productivity of existing assets
  • Minimize operating costs (labor, maintenance, emissions)
  • Improve safety by reducing field operator trips

Challenges in Unconventional Operations (ExxonMobil Permian Basin)

• Large well count (over 10,000 wells) and wide geographic area (larger than Florida)
• Difficulties in operating each well optimally and reducing downtime
• Key challenges faced by production engineers:
  • Disconnected systems and manual processes
  • Siloed decision-making
  • Lack of time for strategic planning
  • Continuous onboarding of new personnel

Addressing ESP (Electrical Submersible Pump) Optimization

• ESPs are a critical artificial lift method used in unconventional operations, but suffer from low run life (around 1 year vs 10 years in conventional wells)
• Downtime and maintenance costs associated with ESP failures are significant
• Opportunity to leverage predictive maintenance models trained on historical data

Leucipa: ESP Optimization Solution

• Developed by Baker Hughes in collaboration with ExxonMobil
• Uses an ensemble of physics-based and machine learning models to:
  • Identify critical conditions affecting ESP run life (e.g., sand production, scale formation)
  • Predict remaining useful life of ESPs
• Provides recommendations to production engineers on actions to improve ESP run life

Key Learnings and Principles for Success

1. Focus on high-value, complex use cases, not just easy problems
2. Prioritize progress over perfection - be iterative and deliver incremental value
3. Carefully select the right technology partners to accelerate innovation
4. Invest in robust MLOps frameworks to handle data and model complexity
5. Actively engage domain experts in the data labeling and model validation process

Results and Future Vision

• Pilot in the Delaware region of the Permian Basin showed 10x return on investment
• Goal to increase ESP run life by at least 10%, which could translate to $25-50 million in annual value
• Vision for the future:
  • System-level optimization beyond individual wells, encompassing the entire production system
  • Expanding scope to include new challenges (e.g., electrification of compression fleet)
  • Democratizing insights through a smart digital platform to enable efficient decision-making by production engineers