Average Modelling of State-of-the-Art Ultra-low Power Energy Harvesting Converter IC

Micro-energy harvesting from environmental sources is a promising technology for mobile or unobtrusive battery powered Internet of Things devices with constrains in terms of size and weight. Special integrated circuits aim to manage the voltage matching and power control in between the energy harvesting transducer and the storage. Hence, models to simulate the energy behaviour of these systems are important tools for system engineers. In this paper, a neuronal network model is presented to predict the performance of two state-of-the-art integrated circuits. Required model inputs are voltage and current from the energy harvesting transducer and the voltage level of the storage element. Model provides conversion efficiency as its output. Three different datasets have been collected with in-field experiments to build and evaluate the proposed model. Evaluation of the developed model on all datasets shows efficiency prediction root mean square error of less than 2 %.