The Research on a Knowledge Graph-Based Method for Detecting Version Differences in Metering Foundations

This paper addresses the version management challenges faced by metering automation systems in the context of smart grids by proposing a knowledge graph-based method for version difference detection and localization, named I-OVDL (Index-based Optimized Version Difference Location with Incremental Updates). As metering automation systems evolve toward distributed microservice architectures, independently updated subsystems have led to severe version fragmentation. Traditional text-based version control methods exhibit significant limitations in detecting semantic differences and providing real-time responses. The main contributions of this study are threefold: First, we construct a domain-specific ontology for the metering system version knowledge graph and apply the TransE algorithm for embedding representation, enabling the mapping from code changes to business semantics, with a 28.6% improvement in accuracy. Second, we design a weighted difference measurement model across four dimensions—functionality, interface, performance, and dependency—using the entropy weight method to determine the weighting coefficients for each dimension. Third, we introduce an indexing optimization technique based on locality-sensitive hashing, which reduces the complexity of version retrieval from O(n) to sublinear time. Experimental results on a dataset containing 685 real system versions demonstrate that the I-OVDL method achieves a 93.6% difference detection rate (a 30.1% improvement over traditional methods), 92.47% feature coverage, and reduces version localization time from 120 seconds to 6.8 seconds. In a provincial power grid pilot application, the method helped reduce regression test cases by 37%, significantly improving operation and maintenance efficiency.