Web application

To facilitate centralized monitoring, diagnostics, and control of the hybrid communication system, a full-stack web application was developed using scalable and production-grade technologies. The backend layer was built using Node.js31, a non-blocking, event-driven runtime environment known for its efficiency in handling concurrent data streams with low latency overhead. The Express.js framework32 was employed to structure the backend into modular RESTful APIs, responsible for managing authentication, role-based authorization, and interfacing with the HES to manage data. The modular design enables the platform to efficiently process continuous smart meter data while remaining maintainable and extensible, decoupling ingestion, storage, analytics, and visualization to optimize performance. This architectural approach draws modular inspirations from recent real-time IoT data processing frameworks33, while avoiding the unnecessary communication overhead of a full distributed microservices model. This design supports real-time event-driven processing of PLC and RF data while ensuring interoperability and scalability for integration with future services such as predictive analytics and grid-wide monitoring. Data persistence and retrieval were handled using MongoDB Atlas34. On the client side, the interface was built using React.js35, a declarative JavaScript library for building interactive and component-based UIs. The frontend provides separate views for different user roles. For example, end-users are presented with energy consumption summaries and trend visualizations, while utility administrators and technicians have access to detailed network diagnostics and AI decision logs (which record the channel chosen dynamically by the on-edge AI agent, the network metrics of that time). Real-time visualizations are powered by integrated charting libraries that render consumption data, signal strength trends, and packet loss heatmaps. A notable feature of the application is the network map, which displays the active communication path (PLC or RF) used by each smart meter in real time. This visualization reflects the decisions made by the PPO-based AI model and helps operators understand system behavior and resilience. The web application’s design also includes a framework for notification systems to alert users to abnormal events such as persistent communication failures, unexpected usage spikes, or performance degradation allowing timely intervention. System validation included confirming that the data acquired by the RF DCU, as there are three instances of receiving reading requests, each one from a different sensor, was accurately transmitted to the backend server and correctly rendered on the web application’s dashboard for real-time visualization. This confirmed the end-to-end integrity of the RF communication pathway spanning node-level transmission, centralized data aggregation, backend processing, and real-time visualization through the web interface.

User dashboard

The MeterFlow user dashboard provides a dynamic view of real-time energy usage. Data received via the PLC or RF channel is immediately processed by the backend server and reflected on the dashboard, enabling users to monitor consumption patterns and key metrics. The system also offers a Consumption page that details electricity usage and meter status, including Meter ID, location, and PLC metrics such as latency and packet loss. Configurable thresholds for warning and critical consumption levels allow users to proactively manage their energy usage.

Admins dashboard

The MeterFlow web application offers an administrator dashboard for system-level monitoring and control, presenting aggregated metrics such as total meters, overall energy consumption, and pending notifications. It includes panels for Meter Status (Online, Maintenance, Offline) and Communication Lines (PLC vs. RF), enabling utility providers to assess real-time system performance across the infrastructure. The network map, shown in Fig. 6, highlights the communication channel selected by the local PPO agent embedded in each meter, using distinct node colors (blue for the Central DCU, green for online meters, red for offline) and line styles (solid black for PLC, dashed green for RF). This visualization provides a real-time geographic map of the network, displaying the active communication channel (PLC or RF) and operational status (online, offline) for each meter. This allows operators to identify network performance patterns and correlate them with physical locations for maintenance and optimization.

Fig. 6: Network map showing active links.