Optimizing Oil & Gas Field Development
A leading unconventional gas producer deployed BHC3™ Well Placement & Completion to optimize well development and to reduce the number of uneconomic wells drilled annually.
The company’s upstream business operates more than 2,000 coal seam gas wells, with plans to grow by 200-300 wells per year through field development. In just 12 weeks, the company deployed BHC3 Well Placement and Completion on the BHC3™ AI Suite to enable this growth.
The application applies BakerHughesC3.ai’s (BHC3) advanced machine learning capabilities to prioritize only high-producing and economically favorable wells for development. Using BHC3 Well Placement & Completion, field development planners can predict the output of an individual well—before commencing drilling—with 70% precision in identifying low-potential wells.
This allows producers to optimize capital expenditure by eliminating drilling uneconomic wells. In addition, field development planners are able to identify key parameters to maximize well output, improving the engineering assessment for future analyses.
About the Gas Producer
Natural gas exploration and production, power generation and retailing divisions
- $10+ billion annual revenue
- 2000+ coal seam gas wells in operation
- Upstream operations across 3 countries
- 6,000+ employees
Building and deploying an AI-based application
Over the course of 12 weeks, the team worked with data scientists and subject matter experts from the unconventional gas producer to deploy AI algorithms to optimize field development. BHC3 built a unified, federated cloud image of all relevant data for assets in scope, integrating 10 disparate source systems. The joint team configured the BHC3 Well Placement & Completion application using data from over 300 wells in operation for over one year. Using only data collected prior to pre-drilling, the application identified nearly 80% of the low-producing wells, providing a 3x increase in predictability over the baseline physical reservoir model. The joint team developed machine learning models to predict the output of each individual well, present and future, with 70% precision. The team built three different algorithms for each well development stage: planning, drilling and completion, and early operations.
- 12 weeks to implement production-ready tool
- 300+ wells analyzed
- 12 months of production data assessed
- Data integrated from 10 disparate source systems
- 340 features tested