Enterprise AI for Oil & Gas Software and Services
An electrical engineering and software company provides simulation software and consulting services for the oil and gas industry in more than 20 countries.
The company’s simulation solution is adopted across refineries globally and helps refinery operators perform planning and scheduling tasks to meet production objectives. However, unplanned production changes, changing market conditions, and system siloes often result in plans that are no longer optimal.
To address these challenges, the company chose the BHC3™ AI Suite, a comprehensive platform to build, deploy, and operate enterprise AI applications, to augment the current solution with scalable, enterprise AI capabilities. The AI-enabled solution would have an end-to-end view of refinery operations, dynamically recalibrate based on real-time data and production changes and increase the robustness of current planning and schedule systems with ML models.
In less than 8 weeks, the Baker Hughes and C3 AI team delivered a fully interoperable, AI-enhanced simulation solution. The development team integrated 6 years of historical data, from 6 disparate data sources, into a unified data image. The team of 5 modelled over 900 unique objects and created over 100 machine learning configurations on the BHC3 AI Suite, demonstrating the speed and scale of the end-to-end machine learning lifecycle on the BHC3 AI Suite.
The AI-enabled simulation solution has the potential to unlock $1-2 billion in annual economic value for oil and gas companies by eliminating system siloes and allowing operators to quickly adapt to production changes and dynamic market conditions.
About the Company
- ~$100M in 2019 revenue
- 600+ employees
- Operations in 20+ countries
The company’s leading simulation software helps operators meet production goals and maximize profiting by providing decision support for the planning and scheduling stages of refinery production.
Operators uses the simulation solution to analyze various input and output scenarios to determine the profit maximizing volume and mix of crude oil to purchase and petroleum products to produce.
In addition to modelling complex refining processes, operators must flexibly adapt the production to dynamic market conditions of crude oil and petroleum products that can be sensitive to geopolitical forces and other external factors.
When actual production deviates from simulated outputs, the simulation model needs to be recalibrated, a process that involves deep subject matter expertise, and gathering inputs from siloed systems across the refinery.
Disparities between planned and actual output – due to system siloes and limitations of legacy systems – negatively impact profitability, and refineries’ ability to respond to dynamic market conditions. Multiple challenges remained:
- Development teams lacked a comprehensive set of development tools that were interoperable with, and could be used to embed AI capabilities quickly into current solutions
- Development teams did not have an efficient way to configure the existing user interfaces to reflect new AI-enabled predictions and analysis
In less than 8 weeks, a team of Baker Hughes and C3 AI experts demonstrated bi- directional interoperability between the current simulation solution and the BHC3 AI Suite. The project focused initially on the solution’s preheat train module, which represents a crucial step in the overall refining process.
First, the team created a unified, federated data image that represented the entities and asset hierarchies of the preheat train simulation model. The team ingested 6 years of data from 6 disparate data sources, including sensor and meter data, simulation baselines and simulation outputs. More than 900 unique preheat train objects were modelled with 8 logical BHC3 Types and unified in the data model, demonstrating the extensibility of the BHC3 AI Suite to any energy-related source system and refinery configuration.
With the unified data image as the foundation, the team created more than 100 supervised machine learning models using multiple algorithms, features, and automated hyperparameter optimization to illustrate how development teams can easily design, train and deploy ML workflows through pre-integrated open-source and third-party tools and services on the BHC3 AI Suite.
Lastly, the team created a web-based UI with readily available BHC3 UI components to visualize AI-enabled simulation and model outputs. Users have a comprehensive and granular view of the assets in-scope and AI-enabled analytics and predictions.
- 6 years of data from 6 separate sources ingested and unified on a federated data image
- Unified, federated data image created
- 900+ unique objects modelled with the BHC3 Type System
- 100+ supervised machine learning models configured
- 2 user interfaces configured with BHC3 UI components and open-source frameworks
By embedding the AI-augmented simulation software across planning, scheduling and production operations, the company can potentially help oil and gas companies achieve $1-2 billion of annual economic value.