Enterprise AI for Improving Electric Submersible Pump Performance

Challenge

An integrated, Latin American hydrocarbon producer engaged Baker Hughes and C3 AI to implement BakerHughesC3.ai (BHC3) technology for improved reliability of Electric Submersible Pumps (ESP) in its upstream production operations to reduce unplanned downtime and deferred production.

As reservoirs age and pressures decline, they require assistance from ESPs to transport hydrocarbons to the surface. When an ESP fails, reservoir production halts for several weeks with a cost of over $300,000 in deferred production, service, and equipment costs. Existing systems produced a high volume of false alarms and could not provide the prescriptive insights required to quickly resolve alerts. Due to the constraints of conventional systems, approximately 20% of ESPs failed on an annual basis. The oil and gas company selected BHC3TM Reliability application to more effectively anticipate and prevent ESP failures.

Approach

The BHC3 team began by creating a unified and federated data image comprised of more than 50 million rows of data including ESP sensor, production, completion, perforation details, deviation surveys, productivity indices, equipment details, and failure data, allowing the company to leverage historical data from all relevant sources.

The team then utilized deep learning and unsupervised algorithms to identify the optimal models to predict ESP failures and diagnose unhealthy ESPs. In addition, the team leveraged asset templates and the BHC3 Reliability Failure Mode Library and recommendation engine, to provide prescriptive actions and root cause insights for identified risks. The team configured a multi-screen user interface to visualize AI-based insights, failure risk scores and time series-based analytics across all 400 ESPs.

Project Objectives

• Create a unified, federated data model integrating data from nine disparate data sources
• Reduce the number of false alarms generated by threshold-based systems
• Apply machine learning to predict ESP failures 20-40 days in advance of failure
• Provide failure mode analysis and recommended actions by embedding BHC3 Reliability Failure Mode Library

About the Company

• $13B annual revenue in 2020
• 1,700+ million barrels of oil equivalent in reserve balance
• Operations in 4 countries
• Investing between $100-150M in innovation and technology to accelerate digital transformation

Project Highlights

• 14 weeks from kick-off to pre-production application completion
• 5 years of historical data integrated, comprising of over 50 million rows from 9 disparate enterprise IT systems
• 16 BHC3 logical objects used to build an extensible data model
• 250+ timeseries analytics developed for machine learning models and application UI
• 14,400+ machine learning model permutations configured and tested to predict pump system failures and detect anomalies
• Configured BHC3 Reliability application user interface

The BHC3 AI Suite provides the necessary and comprehensive services to build enterprise-scale AI applications up to 18x faster than alternative approaches. The BHC3 AI Suite integrates all relevant data sources to rapidly generate predictive insights across the oil and gas value chain. When deployed at enterprise-scale, BHC3 applications can deliver up to $100 million and more in annual economic value to oil and gas organizations.

BHC3 provides wide-ranging Enterprise AI applications for oil and gas companies, including optimizing artificial lift systems operations, selection of drilling targets, and production operations. These pre-configured applications provide insights to automate the well lifecycle design process, allow real-time monitoring of process reliability, and lower costs of maintenance interventions.