CAS Institute of Coal Chemistry operates AI-based coal conversion platform which covers 740,000 structures

https://www.cas.cn/cm/202611/t20261117_5124320.shtml

Traditional research in coal conversion involves numerous variables, such as reaction processes and material properties; developing a single catalyst formulation can often take a decade or longer. Wen Xiaodong, a researcher at the Institute of Coal Chemistry, Chinese Academy of Sciences (Shanxi), explains that catalyst R&D previously relied heavily on expert experience and extensive, repetitive experimentation. Knowledge scattered across academic literature and data stored on individual computers remained isolated, preventing effective reuse. To address this challenge, my country’s first intelligent, end-to-end research platform for the coal chemical industry—the Intelligent Platform for Coal Conversion (IPCC)—was recently unveiled in Taiyuan, Shanxi. By integrating decades of accumulated research in coal chemistry, the platform transforms the catalyst development process from one of “trial and error” to one of “calculating before executing.”

The Intelligent Platform for Coal Conversion focuses on the core objective of “intelligent coal chemistry driven by both theory and data.” It is built upon four key pillars:

  • a knowledge base,
  • a domain-specific large model,
  • specialized tools, and
  • a verification mechanism.

The knowledge base contains data on the properties of over 38,000 pure substances and more than 740,000 structured corpus entries, alongside over 70 integrated property-calculation models. This converts knowledge—previously scattered across academic papers, reports, and the minds of experts—into searchable, inferable, and reusable research assets. The domain-specific large model interprets the needs of researchers and dispatches the appropriate specialized “sub-agents.” The specialized tool layer encompasses 121 distinct capabilities across 11 categories, enabling the intelligent invocation of common research software and streamlining the entire workflow—from planning and experimentation to analysis, visualization, and reporting—into an automated delivery chain. Meanwhile, the verification mechanism cross-references the sources of evidence and units of measurement, ensuring that results are reliable and the research process remains fully traceable. Furthermore, the platform is complemented by an intelligent laboratory for coal conversion. Equipped with self-developed automated experimental apparatus, it covers the entire lifecycle of catalysts—from preparation and characterization to performance evaluation. Experimental protocols generated by the system’s algorithms can be executed directly by machines, with the resulting data fed back into the platform to refine its models, thereby creating a closed-loop process of “design, preparation, testing, and improvement.”

“We aren’t just using AI as a ‘showpiece’; we are integrating data and robotics directly into the core of scientific research,” says Wen Xiaodong. “We let intelligence flow through the data, breathing new life into coal conversion through transformation.” Previously, developing a single formulation required researchers to stand by the furnace and manually adjust parameters; now, the platform narrows down the search space, machines carry out the experiments, and researchers focus solely on high-level strategic decision-making.

This research paradigm has already proven effective in practical applications. In the development of catalysts for Fischer-Tropsch synthesis, the platform utilizes specialized models to design experimental protocols, while automated systems handle preparation and testing—achieving a carbon dioxide selectivity of less than 1% in the final products. In research on olefin separation, the platform analyzes the distribution of α-olefins, providing actionable insights for optimizing pathways toward high-value, coal-based chemicals.

Most popular posts: