Enterprise AI for Liquified Natural Gas Process Optimization

Project Challenge

A major liquified natural gas (LNG) company harnesses the country’s natural gas resources and produces LNG and natural gas liquids for export. LNG plants turn vapor natural gas into liquified natural gas and are made up of parallel LNG trains—the liquification and purification facilities. This company has a total production capacity of 20+ million tons of LNG per year and holds multiple long-term purchase agreements with international buyers.

The optimization of LNG production was being done manually—the company relied on the knowledge and expertise of key personnel as well as attempting to use manual experimentation.

Although they were able to use these resources, along with an abundance of valuable data across its well-instrumented assets, several challenges remained. The company lacked the capability to analyze and use their large, disparate datasets in the decision-making process, resulting in a gap between current and optimal production. This gap also reduced energy efficiency and increased greenhouse gas emissions per unit LNG production.

To close this gap, the company needed to solve the following problems: siloed expert knowledge with limited documentation and data sharing; inability to update operational setpoints in real time due to frequently changing feed quality and ambient conditions; and manual optimization solutions based on traditional methods are labor intensive and have limited capacity to analyze process variables.

About the Company

  • $2B+ in annual revenue
  • 1,000+ employees
  • 20+ million tons per annum of LNG production capacity

Project Objectives

  • Integrate all data from LNG trains across disparate data sources (e.g., asset hierarchy, sensor tags, failure events, piping and instrumentation design) into a unified federated data image
  • Apply machine learning algorithms with multi-train optimization to understand effects of various independent variables to optimize LNG throughput
  • Deliver intuitive user interfaces that expose machine learning outputs to provide operating set points for optimal production throughput and energy efficiency

Approach

A team of C3 AI, Baker Hughes, and Shell experts collaborated with project managers and subject matter experts from the LNG company to develop a production application that would allow users to configure, run, and manage optimization models for multiple LNG trains on a single platform while automating process data streaming and aggregation.

To develop this application, the team needed to create a unified and federated data image. The company uses a software platform called the OSIsoft PI system to capture, process, and record data in real-time. With the use of 3,500 PI tags across seven LNG trains, the unified data model included time series measurement data, asset framework, and existing metadata from previous machine learning projects. The machine learning models leverage Shell’s built-in LNG expertise and process knowledge as well as integrate Shell’s state of the art optimization frameworks.

The team configured BHC3 Process Optimization for LNG dashboards to visualize the predictive insights generated by machine learning models and optimization algorithms. The intuitive dashboards provide a managerial view of high-level key performance indicators (KPIs) and operating metrics.

BHC3 Process Optimization for LNG, combined with process control services offered by Baker Hughes TPS iCenter Services, enabled the LNG company to increase its annual LNG production by 1%, leading to a $50M annual economic benefit.

Project Highlights

  • 6 months to production
  • Integrated 3,500 data tags for 7 LNG trains
  • Created unified object model to represent asset hierarchy, telemetry data, sensor data, and OSI PI data
  • 14 ML models and optimization pipelines trained
  • 100+ ML models with multi train optimization developed
  • 7 BHC3 Process Optimization for LNG application user interfaces configured

Results

$50M
estimated annual economic impact
6
months to production
100+
ML models with multi-train optimization
3,500
PI tags integrated

Solution Architecture

BHC3 Process Optimization

Proven results in weeks, not years

timeline
Get insights into BHC3 capabilities, enterprise AI best practices and highest-value use cases.
Understand BHC3™ AI Suite's capabilities, its model-driven architecture and test it against your company's sample data set.
Identify a high-impact business problem and collaborate with the BHC3 team to rapidly build an AI application that solves it.
Scale and deploy a tested BHC3 application into production. Incorporate user feedback and optimize algorithms to drive maximum economic value.

Thank you for your interest in BHC3.ai

We'll review your information and a team member will get back to you within 24-48 hours.

Return to home
Loading form...