Case study · Chemicals

ML surrogate models improve air separation unit energy efficiency 10%

Air separation units are electricity-intensive, and full first-principles models are too slow for routine scheduling, demand response and control.

Manufacturer
Linde (Praxair)
Use cases
1
Evidence
Independently assessed
Year
2022

Independently assessed  This result was assessed or published by a party other than the manufacturer or its vendor (a World Economic Forum Lighthouse assessment, a government or institute report, or a peer-reviewed paper).

The challenge

Air separation units are electricity-intensive, and full first-principles models are too slow for routine scheduling, demand response and control.

The approach

Texas A&M and partners built data-driven surrogate models from plant data for planning and scheduling, demand response and MPC, plus data-driven fault detection.

Reported improvements

Percent change as reported by the source; hover for the exact wording

Model size and computational cost↓ 98%
Energy efficiency↑ 10%

1 use case in this implementation

  1. Production planning & scheduling

    Data-driven surrogate models of the air separation unit used for production planning, scheduling and demand response.

    • Surrogate Modeling for Planning & Scheduling resulted in 10% improvement in energy efficiency
    • 98% Reduction in model size & associated computational costs
    • $200k/yr Potential savings for each ASU in operating costs

Outcomes as reported by the source. We do not adjust, combine or extrapolate reported figures.

Related case studies

Put AI to work

Find the AI opportunities hiding in your plant.

Genvo, our implementation partner, works with manufacturers to pick the right first project, connect the systems you already run and put AI agents and tools into daily use, with people approving what goes out.

Assessments and consultations are provided by Genvo. How our partnership works.