Research on production emergency scheduling of refineries based on large language model
GONG Yu
YANG Jingzheng
SONG Ren
CAO Xin
Abstract：The production emergency scheduling of refineries is crucial for smooth operations and responding to national energy conservation and carbon reduction calls.The current production emergency scheduling is mainly manual,which has many problems such as high labor cost and reliance on knowledge and experience.Therefore,it is of great significance to use artificial intelligence technology to assist or replace manual labor.The research on the production emergency scheduling model of refineries based on LLM mainly includes:① The event extraction model uses the contextual learning ability of LLM to extract key information from the emergency event text,and then uses it to build a knowledge base for the production emergency scheduling of refineries.② The scheduling plan generation model intelligently selects LLM or CBR that has been fine-tuned for emergency dispatch of refinery production according to the device,description and impact to generate a scheduling plan.Finally,validation based on an actual refinery emergency scheduling plan shows the LLM-based model accurately extracts key information and generates high-quality plans,confirming the emergency scheduling model's feasibility and practicality.
Keywords：refineriesproduction emergency schedulinglarge language modelcase-based reasoningfine tuning
Publication Date：2025-09-28
Online Publishing Date：2025-11-05(First online date of this platform, not the publication date of the document)
Pages：11( 99-109 )