案例库 · 工程与运营 · 技术决策 · 1980–2014
这条还没译成中文，下面是英文原文。
Chevron's refineries run on a linear program that adds about $1 billion a year.
Seven Chevron refineries use one LP-based modeling tool to pick crude, products and investments — worth roughly $10 billion over three decades.
Chevron
那一手
Chevron built a software modeling tool that its seven company-owned refineries use to select the most profitable crude, evaluate product options, optimize processes and screen capital projects.
The tool is a linear program with distributive recursion, used in both operations and strategic planning; over 30-plus years the company kept improving it until it was embedded in the fabric of the downstream business.
Chevron estimates the effort now contributes close to $1 billion a year and roughly $10 billion cumulatively. INFORMS named the work a finalist for its 2013 Franz Edelman Award.
为什么管用
- A single modeling standard lets refineries compare the same options on the same math.
- Distributive recursion handles the nonlinear detail plain linear programming misses.
- Deep integration with business processes made the model the default, not a special study.
可以搬走什么
A decision model kept and improved for decades compounds: embed it in core processes, feed it trusted data, and its value grows far beyond the first deployment.
后来呢
Chevron's downstream business still runs on the model and its successors; the 2014 Interfaces paper documenting the system became a reference case for linear programming in process industries, and the Edelman finalist citation is used in OR teaching.
资料来源
- Edelman 2013 Announcement (finalists, via Internet Archive)
- Optimizing Chevron's Refineries (Interfaces 44(1), 2014)
- Introduction: 2013 Franz Edelman Award (Interfaces 44(1), 2014)
发现哪里写错了？告诉我们。