Abstract
Many previous works have treated the conversion efficiency of an energy harvester as a constant or a non-linear function of input power, ignoring the non-linear effect of the operating frequency. Such assumption may lead to erroneous performance analysis or improper resource allocation. In this article, a novel model is proposed to capture the non-linearity of the conversion efficiency as a function of the frequency and power. The paper first exploits the practical harvesters in the literature to model the conversion efficiency as a function of frequency at fixed power. Based on the shape of the efficiency curve, two categories of harvesters are referred to as Type-I and Type-II. The best-fit model for each category is selected based on the adjusted R-square test. Next, these models are extended to jointly consider both power and frequency dependencies. Using these models, the performances of broad-band simultaneous wireless information and power transfer and OFDM hybrid access point are analysed and optimized. Iterative algorithms are used to find low-complexity solutions. Moreover, the impact of the non-linear conversion efficiency on the average battery life in LoRaWAN is analysed. Our results highlight the significant impact of using the joint model of conversion efficiency.