A Scalable Microservices Architecture for Condition Monitoring and State-of-Health Tracking in Power Conversion Systems

TL;DR

This paper presents a scalable microservices architecture for monitoring and tracking the health of power conversion systems in modern electrical infrastructure.

Contribution

The novel contribution is a containerized, four-layer microservices architecture optimized for real-time condition monitoring and scalability in power electronics.

Findings

The system achieved a Packet Delivery Ratio of 1.0 at ingestion rates up to 100 messages per second per node.

Transmission and processing overheads remained consistently below 5 ms, meeting real-time monitoring requirements.

The modular design supports horizontal scalability and meets Industry 4.0 integration needs.

Abstract

The role of power converters in modern electrical infrastructure (such as electric vehicle charging stations, battery energy storage systems and photovoltaic energy systems) has become critical. Given the high reliability required by these converters, continuous condition monitoring for predictive maintenance is mandatory. Traditional SCADA and HMI systems often face scalability bottlenecks and lack the flexibility in data aggregation and storage scalability required for long-term predictive maintenance. This paper proposes a scalable, containerized microservices-based architecture for degradation tracking and State-of-Health (SoH) monitoring in power conversion systems. The architecture features a decoupled four-layer structure, utilizing dedicated UDP servers for low-latency data ingestion, RabbitMQ (AMQP) for robust message routing, and a NoSQL (MongoDB) storage layer with a FastAPI interface. The proposed system was validated using a Hardware-in-the-Loop (HiL) setup with a Typhoon HIL606 simulator monitoring an Active Neutral Point Clamped (ANPC) power converter. Experimental stress tests demonstrated a Packet Delivery Ratio (PDR) of 1.0 at ingestion rates up to 100 messages per second (msgs/s) per node. The system exhibits transmission and processing overheads consistently below 5 ms, ensuring timely data availability for tracking thermal dynamics and parametric aging trends. This operational performance significantly exceeds the nominal requirement of 2 msgs/s for condition monitoring, ensuring robust data integrity. Finally, this modular approach provides the horizontal scalability necessary for Industry 4.0 integration, offering a high-performance framework for long-term health monitoring in modern power electronics.

Figures