Physics-Constrained Multi-Sensor Fusion and Uncertainty-Aware Error Propagation for Industrial Photovoltaic Measurement Systems Using Jetson Orin Nano

Accurate measurement and uncertainty management are essential for reliable monitoring of industrial photovoltaic (PV) systems operating under heterogeneous sensing conditions. This study proposes a novel edge-based measurement framework that integrates physics-constrained multi-sensor fusion, an online uncertainty propagation graph, and uncertainty-aware Kalman filtering to ensure physically consistent and reliable PV measurements. Unlike conventional approaches that treat uncertainty as a static or post-processing parameter, the proposed method explicitly embeds physical PV relationships into the fusion process while dynamically modeling uncertainty propagation across the entire measurement chain. The framework is implemented on a Jetson Orin Nano edge platform, supported by ESP32-based acquisition nodes and MQTT communication via a Raspberry Pi local broker. Real measurement data collected from industrial PV installations in a real-world industrial zone in Türkiye are used for validation under realistic operating conditions. Experimental results demonstrate that the proposed approach significantly enhances measurement reliability, reducing RMSE from 2.43 V to 1.12 V (53.9% improvement) and variance from 6.12 V² to 2.48 V² (59.5% reduction). The system maintains real-time performance with an average latency of 32 ms at a 50 Hz sampling rate. The key novelty lies in transforming the edge device into an intelligent measurement instrument by jointly integrating physics-constrained fusion, dynamic uncertainty propagation, and adaptive filtering within a unified architecture. The results demonstrate that embedding uncertainty-aware processing directly into the measurement pipeline substantially improves consistency, robustness, and real-time reliability in industrial PV monitoring systems.