PipeClaw: An operation support system for natural gas pipelines with operation history, transient prediction, and verified dispatch

Natural gas pipeline dispatch is a continuous engineering process in which current decisions depend on recent operating history, short horizon response, and equipment feasibility. Existing studies have improved scheduling, transient prediction, and decision support, but they usually treat these functions separately and provide limited support for continuous operation across shifts. This study presents PipeClaw, an operation support system for natural gas pipelines that integrates historical operation context, transient prediction, and verified dispatch within one traceable workflow. The system retrieves relevant history for each request, calls PipeFormer when candidate actions require forward consequence analysis, and applies an equipment aware verification loop before returning recommendations. External capabilities such as document processing and data retrieval are integrated through a unified skill mechanism, so all external calls remain visible in the traceable record. Experiments on PipeFD-40 and PipeOpsCycle-40 show that PipeClaw achieves the best overall results, and on PipeOpsCycle-40 it also attains the lowest hard failure rate. The results indicate that the integration of persistent context, short horizon prediction, and explicit verification improves the continuity, safety, and traceability of pipeline dispatch support.