A Deep-Neural-Network-Based Surrogate Model for DC/AC Converter Topology Selection Using Multi-Domain Simulations

The selection of optimal DC/AC power converter topologies for specific applications is often a time-consuming and complex task, which can lead to suboptimal choices. This paper proposes an AI-assisted methodology to identify the most efficient DC/AC converter based on a set of input design parameters. Separate deep-neural-network-based surrogate models are developed for each considered topology, trained by a large dataset of simulation results obtained from MATLAB/Simulink and PSIM, so that the efficiency of each converter can be determined without performing additional simulations. The proposed methodology allows for quick and accurate efficiency estimation, significantly reducing the analysis time for topology selection. A case study for the two-level converter is also presented, demonstrating that additional parameters, such as the semiconductors junction temperature and output current distortion, can also be predicted using a similar methodology. Results are presented to demonstrate the feasibility of the proposed method.