Ambient RF energy harvesting using voltage doubler rectifier for battery-free agricultural sensor networks: Design, statistical analysis, and machine learning validation

The growing adoption of precision agriculture requires sustainable power sources for wireless sensor networks (WSNs) deployed in remote and resource-constrained environments. Conventional battery-powered systems are constrained by limited-service life, maintenance requirements, and environmental concerns. This study presents the design, experimental validation, statistical evaluation, and machine learning (ML)-based modeling of an ambient radio-frequency (RF) energy harvesting system for agricultural monitoring applications. The proposed system harvests RF energy from AM/FM broadcasting (558 kHz–108 MHz) and cellular communication bands (800–2100 MHz) using a frequency-selective broadband antenna, an integrated diplexer, switchable L-section impedance-matching networks, and a Villard voltage-doubler rectifier based on low forward-voltage OA79 germanium diodes. Experimental measurements from 30 independent repeated trials yielded a mean open-circuit output voltage of 4.07 ± 0.08 V (95% CI: 4.04–4.10 V) and a loaded output voltage of 2.11 ± 0.11 V, corresponding to a harvested power of approximately 44.5 μW across a 100 kΩ load. One-way analysis of variance (ANOVA) demonstrated significant differences among diode types (F (2,87) = 687.4, p < 0.001), while Tukey’s post hoc test confirmed the superior performance of OA79 diodes compared with 1N4148 and 1N5819 alternatives. Among the evaluated ML models, Gradient Boosting regression achieved the highest predictive accuracy, with R² = 0.963 and RMSE = 0.071 V. SHapley Additive exPlanations (SHAP) analysis identified diode forward voltage and ambient RF power density as the most influential factors affecting output voltage prediction. The results indicate the potential applicability of ambient multi-band RF energy harvesting as a supplementary energy source for low-power agricultural sensing applications operating under duty-cycled conditions. By combining broadband energy harvesting, statistical validation, and predictive modeling, the proposed framework establishes a systematic methodology for evaluating self-powered agricultural IoT systems operating under variable RF environments.