Dual fuel-cell sourced wireless power transfer and load power estimation with Machine Learning

In this study, a power circuit with wireless power transfer feature is proposed for stand-alone applications where more than one fuel cell can serve as power source. In order to harvest maximum power from fuel cells, the perturb&observe $(\mathrm{P} \& \mathrm{O})$ algorithm is used and a 5 -level cascade inverter connection providing low harmonic distortion at high frequencies is modeled. Load power estimation is performed with the data obtained from the modeled system (temperatures, currents, voltages) by using Machine Learning (ML) algorithms (Support Vector Machine (SVM), Artificial Neural Networks (ANN), and Gaussian Process (GP)). The aim of this study is to create a circuit structure for the simultaneous use of multiple hydrogen fuel cells as alternative energy sources in the feeding of industrial loads or vehicle charging stations and at the same time to make high-accuracy load power estimation suitable for smart power systems by using machine learning algorithms. In the proposed system, the computer application of the prototype created with two fuel cells, each with a maximum power of 2 kW, under variable temperature $\left(55^{\circ} \mathrm{C}-65^{\circ} \mathrm{C}\right)$ conditions is realized. Matlab/simulink simulation program is used for system modeling and analysis.