Predictive Modelling of RF Energy for Wireless Powered Communications

Abstract

Energy harvesting enables perpetual operation of wireless networks without the need for battery change. In particular, energy can be harvested from radio waves in the radio frequency spectrum. To ensure a reliable performance, energy prediction modelling is a key component for optimizing energy harvesting because it equips the harvesting node with adaptation to energy availability. We use two machine learning techniques, linear regression (LR) and decision trees (DT) to model the harvested energy using real-time power measurements in the radio spectrum. Numerical results show that LR outperforms DT by attaining minimum 85% prediction accuracy. These models will be useful for defining the scheduling policies of harvesting nodes.