GS Caltex expands AI applications at Yeosu plant

(GS Caltex)

At the GS Caltex Corp. plant in Yeosu, South Jeolla Province, engineers specializing in artificial intelligence (AI) constantly monitor the entire facility, which spans 6 million square meters. The plant’s operations are comprehensively managed using AI, which processes vast amounts of data transmitted from 200,000 sensors connected to 800,000 pieces of equipment.

AI’s predictive capabilities have already far exceeded initial targets of 70 percent accuracy in both refining and petrochemical processes. In particular, the polymer unit - where plastic raw materials are produced - boasted an 84.6 percent success rate in detecting abnormal signs.

According to GS Caltex on Thursday, the Yeosu plant was selected for the government’s AI Autonomous Manufacturing Leading Project in October 2024, beating a competition rate of 8 to 1, and is now on track to become one of South Korea’s flagship AI factories.

The five-year project, which runs through 2028, has just begun but is already outperforming its original goals.

As part of the process, data from the 800,000 devices and 200,000 sensors - transmitted every second - are automatically stored each minute in a centralized “data lake.”

“While traditional process control relied mainly on mathematical simulations, GS Caltex has built a machine learning operations (MLOps) framework that connects lab and on-site data, improving the quality of AI model training,” the company said.

GS Caltex is now preparing to deploy its operation optimization platform (OOP) - a comprehensive AI system that automatically detects performance degradation, anomalies, or efficiency losses in real time and instantly applies optimized adjustments to the production process.

AI is expected to make operational decisions even before human operators do.

“There is still room for human decisions in the process, but operators will be able to see, in real time, how their choices compare to AI’s optimal values through KPI metrics such as sales performance,” Kang Jae-min, head of GS Caltex’s digital innovation team, said. “Since AI can even optimize nonlinear reaction processes such as catalytic reactions that mathematical simulations could not handle, it will significantly boost productivity.”

GS Caltex expects that once the full AI-factory transition is complete, the system’s prediction accuracy will be as high as 95 percent. This will reduce the plant’s annual carbon emissions by more than 10,000 tons and increase production of high-value-added products by over 10 percent.

2025 marks the first year in which policies promoting AI adoption in manufacturing are being actively implemented in the field.

A total of 213 companies competed to participate when the Ministry of Trade, Industry and Energy opened applications for 26 leading projects in October 2024, a sign of growing industrial interest. GS Caltex was among those selected.

The ministry aims to expand the number of AI-enabled manufacturing sites to 500 by 2030, in line with President Lee Jae-myung’s pledge to make Korea one of the world’s top three AI powers.