Comparison of different Machine Learning techniques for PCE estimation of MMWave Energy Harvesting Rectifier devices

Power conversion efficiency (PCE) is one of, if not, the most important metric in energy conversion systems, it predicts the energy output of a device, given its input. In radio frequency (RF) energy harvesting (EH) devices, the energy in electromagnetic (EM) waves is captured and turned into a stable low power energy source for various uses, a key component in this process is the harvester's rectifier, of which, its non linear behavior greatly increases the complexity of PCE estimation, making approximate models inaccurate, and precise models extremely complex. This is why, in this work, a set of simpler solutions, for rectifier's device PCE estimation are proposed. Through the use of data science and Machine Learning, Support Vector Regressor (SVR), Random Forest Regressor (RFR), and Neural Network (NN) models are trained and compared, using mean square error and absolute mean error as metrics for validity, obtaining a viable model and defined steps for improvement.