Edge-Cloud Energy Management for AC/DC Hybrid Building Microgrids: A Knowledge-Graph-Enhanced Optimization Approach

Abstract

AC/DC hybrid building microgrids require an energy management system (EMS) that coordinates distributed energy resources while maintaining fast local responses to communication and device faults. This study proposes a knowledge-graph-enhanced edge-cloud EMS for an AC/DC hybrid building microgrid. A 24 h linear-programming scheduler coordinates multi-resource dispatch in the cloud, while edge controllers enforce local safety constraints, correct setpoints and maintain fallback operation during link interruptions. The knowledge graph provides semantic context by linking assets, constraints, faults and admissible actions. Six operating scenarios, controlled V2G ablation tests and workday–weekend validation were used to assess the framework. Compared with rule-based EMS, the proposed method reduced peak demand by 28.0% and increased PV self-consumption from 78.4% to 95.4%. Compared with cloud-only MPC, it reduced the simulated mean control-loop latency from 7.54 s to 0.81 s. The eight-rule fault evaluation achieved a mean trigger accuracy of 94.6%. Under normal operation, however, the higher PV self-consumption was accompanied by a modest increase in operating cost and peak demand relative to non-KG edge-cloud MPC. These results support the complementary use of cloud scheduling, edge autonomy and semantic context within the tested simulation scope.