This digest presents the design and experimental validation of a secure, machine learning (ML)-enabled Industrial Internet of Things (IIoT) power supply system for real-time monitoring, control, and predictive diagnostics in oil and gas applications. The proposed platform integrates HTTPS-encrypted telemetry, and deterministic Modbus RTU. Experimental validation under elevated temperature and humidity confirmed uninterrupted operation for 15.9 hours without packet loss or timestamp drift. The MCU-based embedded logistic regression model achieved 93.1% soft-fault classification accuracy with less than 2 ms inference latency, enabling predictive diagnostics under real-time constraints. Compared to conventional designs, this work advances IIoT power supply architectures by unifying secure bidirectional control, supervisory control and data acquisition (SCADA) interoperability, and embedded intelligence for Industry 4.0-compliant deployment
