Quantum Power Electronics: From Theory to Implementation

Abstract

While impressive progress has been already achieved in wide-bandgap (WBG) semiconductors such as 4H-SiC and GaN technologies, the lack of intelligent methodologies to control the gate drivers has prevented exploitation of the maximum potential of semiconductor chips from obtaining the desired device operations. Thus, a potent ongoing trend is to design a fast gate driver switching scheme to upgrade the performance of electronic equipment at the system level. To address this issue, this work proposed a novel intelligent scheme for the control of gate driver switching using the concept of quantum computation in machine learning. In particular, the quantum principle was incorporated into deep reinforcement learning (DRL) to address the hardware limitations of conventional computers and the growing amount of data sets. Taking potential benefit of the quantum theory, the DRL algorithm influenced by quantum specifications (referred to as QDRL) not only ameliorates the performance of the native algorithm on traditional computers but also enhances the progress of relevant research fields like quantum computing and machine learning. To test the practicability and usefulness of QDRL, a dc/dc parallel boost converter feeding constant power loads (CPLs) was chosen as the case study, and several power hardware-in-the-loop (PHiL) experiments and comparative analysis were performed.