An emergency task recommendation model of long-distance oil and gas pipeline based on knowledge graph convolution network

When long-distance oil and gas pipeline accidents take place, a set of requirements and constraints ought to be considered, due to the various environment, multiple accident types and hazard-bearing bodies. Compliance with the response tasks directly influences the subsequent accident development. Furthermore, attribute to geographic location or communication limitations, the accident information may be delayed or unclear. Knowledge graphs are an advantageous representation of the fusion of complex relationships. On this basis, this study aims to establish an emergency task recommendation model that can adapt to several accident information description levels. Combined with the proposed emergency case framework, a knowledge graph of long-distance oil and gas pipeline emergency cases is established for the first time. Realize embedding learning by aggregating neighbor node and relationships. The accuracy of the emergency task recommendation is improved by introducing a variety of relations in the case-feature graph. Apply this model to the 11.22 Qingdao Oil Pipeline Explosion incident. Tasks such as culvert and well boundary detection, combustible gas concentration monitoring, and response upgrading are supplemented. Compare with the accident investigation report conclusions, the recom-mended tasks help to prevent the explosion. The proposed model not only has a case information recording framework but is also not bound by the input format of structured cases. It solves the issue of ambiguous case feature input in the task recommendation and has good practical application value.