The encyclopedia · Software & IT · Technical decision · 1980–1989
Texaco's OMEGA blended gasoline by nonlinear model, saving $30M a year
OMEGA used detailed nonlinear models of gasoline attributes to plan and schedule blending across Texaco's refineries, worth about $30 million a year.
Texaco Inc.
the move
A refinery blends dozens of streams into gasoline grades that must simultaneously meet octane, volatility and a stack of other specs. Each attribute is a nonlinear function of the blend, which is why recipes were largely built by experience and trial.
Texaco began developing OMEGA in 1980: an optimization-based decision-support system for planning and scheduling blending, built on detailed nonlinear models of gasoline attributes, refinery databases and on-line data acquisition. It ran on personal computers and larger systems.
Texaco used OMEGA in all seven US refineries plus its Canadian and Welsh refineries. Benefits were estimated at about $30 million annually, along with better quality control, improved planning and marketing information, and the ability to run what-if studies.
why it works
- Nonlinear attribute models captured what linear approximations missed.
- Optimization found cheaper component mixes that still met specs.
- Online data kept the model matched to what the refinery was doing.
- What-if studies turned the system into a planning tool, not just a scheduler.
what transfers
If the constraint that matters is nonlinear chemistry, a linear shortcut loses money: model the real attribute relationships, and the optimizer finds blends that meet specs with cheaper components.
what came after
OMEGA became Texaco's standard blending tool across its refineries, and the Interfaces paper documenting it became a reference case for refinery optimization. The estimated $30 million annual benefit was attributed directly to the improved blending decisions.
references
- OMEGA: An Improved Gasoline Blending System for Texaco
- OMEGA: An Improved Gasoline Blending System for Texaco
spotted an error? The archive wants to know.