Establishing a Sustainable AMI Infrastructure Using a Multi-Agent Q-Learning Framework

Advanced Metering Infrastructure (AMI) plays a critical role in modern Smart Grid systems by enabling the automated collection, transmission, and analysis of energy consumption data. Within AMI networks, smart meter readings are first aggregated by Phasor Measurement Units (PMUs) and then forwarded to Phasor Data Concentrators (PDCs), which subsequently transmit synchronized data to the Meter Data Management System (MDMS). The reliability and efficiency of this communication process are vital, as the failure of any PDC can disrupt data flow and degrade system performance. To address this challenge, we propose a resilient SDN-based architecture for AMI communications that introduces a virtual concentrator mechanism within each SDN controller. This mechanism dynamically assigns PMU data flows to appropriate PDCS based on resource availability, end-to-end latency, and bandwidth constraints. To enhance scalability and adaptability in large-scale deployments, we develop a Multi-Agent Q-Learning framework in which each PMU operates as an autonomous agent that learns to select the most suitable PDC through interaction with the environment. Simulation results demonstrate that the proposed approach significantly improves network reliability, ensures balanced resource utilization, and provides robust fault-tolerance against concentrator failures, thereby enabling efficient and scalable AMI communication under dynamic network conditions.

کلیدواژه ها:

AMI network ، Smart Grid ، Concentrators ، Software Defined Networks (SDN)