Automatic Generation of GIM Data Audit Rules Based on Sentence Embedding Vectors

The digital design results of power substations establish the foundation for their running and maintaining. Currently, the digital design of substations is delivered in the format of Grid Information Model (GIM), an alternative Building Information Modeling (BIM) format for describing power grid infrastructure in China. Since the correctness, compliance, and consistency of GIM data are necessary conditions for information sharing and business decision support, the GIM data must be audited before sharing among the stakeholders. The traditional manual review of GIM data is too inefficient and costly to execute, and thus the power grid industry seeks automatic review approaches. However, one challenge for automated auditing of GIM data is the lack of auditing rules. In order to establish such a rule base for GIM data auditing, this study first categorizes the audit rules, and proposes an XML encoding method for the audit rules. Meanwhile, the methodology of converting the rules described in natural language into XML is also rules proposed using the SBERT model. The application of the developed tool is demonstrated and verified through case studies.