AGENT-BASED HEALTH MONITORING ARCHITECTURE FOR POWER SYSTEMS

摘要

Power system management has become increasingly important due to the fact that power grid systems are subject to frequent failures because of new additions to the grid and emerging power demands. A power failure usually requires real-time actions to be taken so that the failure effects can be mitigated and the system can maintain its safe operating conditions. In this paper, we introduce a three-level hierarchical architecture which combines fault detection and identification methods together with intelligent software agents. The architecture performs decentralized reasoning, and can be integrated into maintenance and logistics scheduling systems to provide fully automated end-to-end solutions.At the lowest level of the architecture, the sensor agents evaluate raw sensor signals to detect and in some cases to diagnose the cause of anomalies. The mid-level agents combine the detection results from the multiple lower level agents to diagnose possible faults that exist in the components or subsystems. Different mid-level agents incorporate algorithms, specifically, a novel Maximum Entropy (ME) method and a Transferable Belief Model (TBM) method. Finally, at the highest level, a fusion agent is utilized to make a final decision based on the diagnostic results provided by the mid-level agents. The architecture enables communication and interaction between the agents, thus allowing an optimal solution to the fault detection and identification problem.We have successfully applied this agent-based health monitoring architecture to data obtained from a simulated model of International Space Station (ISS) Electric Power Systems (EPS). The applicability of the proposed technology has been validated by promising simulation results.