Case Study: Enhancing Production with
Well Intelligence AI Model
Overview.
In the pursuit of optimising production from aging assets, a major oil operator faced challenges with their traditional well models. These models had become outdated and difficult to manage, resulting in inefficient use of resources and suboptimal production.
Intelligent Plant’s Well Intelligence tool offered a cutting-edge solution by leveraging historical data to create a digital twin of the well, and using AI to suggest optimisations. Despite initial scepticism from the well engineers, the tool revealed valuable insights and improvements, demonstrating its effectiveness in extending the financial viability of aging assets.
Industry Context and Challenge.
Traditionally, field production optimisation involved individual well models, engineering analysis, and well tests to enhance each well's performance. However, this approach often fails to optimise the entire field or system effectively. Well Intelligence addresses this gap by using existing production data and machine learning to optimise field production as a whole. The technology aims to improve production by up to 1% across the field, leveraging the ability to simulate and adjust multiple wells simultaneously.
Project Overview
Deployment and Initial Findings:
-
Ageing Asset Focus: The Well Intelligence tool was deployed to enhance production from an aging well within the client’s portfolio. Historically, the client had struggled to justify the cost of maintaining physical models and had encountered issues with outdated data and inefficiencies.
-
Historic Data Utilisation: Well Intelligence used existing historical data from the historian to create a digital twin of the well. This allowed the tool to simulate various settings and provide optimisation suggestions without the need for extensive new data collection.
Challenge and Testing:
-
Sceptical Engineers: Initially, the well engineers were sceptical about the AI model’s recommendations. They intentionally chose the third-best suggestion from the model, expecting it to fail and thus disprove the tool's value. They believed that following the AI’s advice would lead to adverse effects on the well.
-
Unexpected Outcomes: Contrary to their expectations, the well did not experience failure. Instead, the results revealed that the valve had failed, leading to inaccurate readings. This unexpected discovery highlighted the valve issue that had previously gone unnoticed.
Historical Data Review:
-
Verification Process: Testing and verification of Well Intelligence were conducted over a four-week period on a field, which produced around 8,000 barrels per day. Due to the field’s stability and infrequent steady-state production periods, only a limited set of stable production data was available. Despite this, the application successfully analysed historical data and recommended valve adjustments.
Results and Impact.
Surprising Insights:
-
Valve Failure Discovery: The unexpected discovery of the valve failure underscored the tool’s value in identifying issues that traditional methods had missed. This finding alone justified the deployment of Well Intelligence.
-
Production Optimisation: While the tool predicted that minor production improvements were possible, the stability of the field and external factors limited the observable gains. However, even small improvements had a significant impact on the financial viability of the aging asset.
-
Operational Benefits: By following the AI’s recommendations, the client achieved improvements in production efficiency, reducing the frequency of shutdowns and enhancing overall asset performance.
Industry Value:
-
Potential Gains: In general, Well Intelligence has the potential to improve production by up to 1% across fields. The case study demonstrated the requirement to get engineers on board with any deployment of technology - it should not be used to undermine but to supplement and to help the engineers, only together will they get the full potential gains.
Well Intelligence uses historic data to provide recommended valve positionings
Lessons Learned.
-
Cultural Shift: Gaining trust in AI recommendations remains a significant challenge, particularly in well control, and in general in the North Sea Oil Industry. The initial scepticism from engineers highlighted the need for effective communication and trust-building.
-
User Engagement: Extensive interaction with users is essential to communicate the benefits of AI, gather feedback, and facilitate successful technology adoption.
-
Future Deployment: Future deployments of Well Intelligence should motivate the engineers and recognise their contribution to the success of the deployment.
Conclusion.
The deployment of OpMode on BP's PU-IC choke valve demonstrated the tool's powerful capability to detect early signs of equipment degradation, even in scenarios where traditional monitoring methods might fail. By identifying changes in behaviour long before a failure occurred, OpMode provided BP with the opportunity to plan for maintenance proactively, reducing the risk of unplanned shutdowns and optimising operational efficiency.
This case study highlights the value of using intelligent monitoring tools like OpMode to enhance the reliability of critical equipment in the oil and gas industry. Intelligent Plant continues to deliver solutions that empower operators to maintain continuous operations, reduce risks, and achieve substantial cost savings.