Optimizing RF Energy Harvesting in IoT: A Machine Learning Estimation Considering Polarization Effects

Abstract

The rapid evolution of wireless technology has led to the proliferation of small, low-power IoT devices, often constrained by traditional battery limitations, resulting in size, weight, and maintenance challenges. In response, ambient radio frequency (RF) energy harvesting has emerged as a promising solution to power IoT devices using RF energy from the environment. However, optimizing the placement of energy harvesters is crucial for maximizing energy reception. This paper employs machine learning (ML) techniques to predict areas with high power intensity for RF energy harvesting. Five supervised ML algorithms are compared across four scenarios using antennas with circular and linear polarization. The impact of noise filtering on accuracy is also assessed. Results show that random forest outperforms other ML algorithms, demonstrating the effectiveness of ML in estimating optimal energy harvesting locations and providing insights for sustainable energy network development.