Summary

Reinforcement learning for waterflood optimization is one of our flagship applications of artificial intelligence technologies in petroleum engineering. This project utilized reinforcement learning algorithms with a custom-built environment to interact with based on a simple reservoir model to optimize waterflood operations. The end result was a 300 bopd increase in production for a field originally producing 1000 bopd resulting in a $9 million increase in annual revenue.

Project Highlights

• Implemented a custom RL environment based on a capacitance resistance model to reflect the reservoir
• Implemented state-of-the-art RL algorithms for decision-making in waterflood operations including injection rates and producer status
• Insights from RL recommendations resulting in a $9 million increase in revenue

Technical Innovation

The project introduced several technical innovations:

• Custom environment design for petroleum reservoir simulation based on a capacitance resistance model
• Advanced reward function engineering for production optimization
• Implementation of a RL model for the first time in our company’s history

Impact

This project demonstrated the successful application of modern AI techniques to traditional petroleum engineering challenges, paving the way for broader adoption of reinforcement learning in the energy sector.