Self-Evolving Digital Twin-Based Online Health Monitoring of Multiphase Boost Converters

Abstract

Component degradation in power electronic converters severely threatens the system's reliability. These components degrade over time due to switching action, and this phenomenon is further aggravated with wide band gap devices. For ensuring system reliability and accurate degraded component identification, the development of a real-time noninvasive health monitoring mechanism is desired. This article develops and validates a real-time digital twin (DT)-based condition monitoring for multiphase interleaved boost converters. The DT model is based on an actual state-space modeling approach which is solved numerically using Runge–Kutta fourth to mimic the physical system. Then, the output signals from physical hardware and the DT model are compared to find the least squared error-based multiobjective optimization problem. A metaheuristic approach like particle swarm optimization and genetic algorithm is used to estimate the health of components of the converter. The proposed methodology is extendable to different inductor coupling strategies under continuous-conduction-mode and discontinuous-conduction-mode operations. The idea is to generalize the DT modeling concept for condition monitoring. Moreover, the article proposes decoupling and hybrid approaches to improve estimation accuracy by 9.4% and reduce embedded computational requirements by 22%, respectively. A 75 kW, 60-kHz SiC IBC hardware prototype is built and tested for concept validation. Notably, the challenges and impact of various sensing integrity errors encountered during condition monitoring are also discussed. Finally, the article discusses novel pre and postprocessing steps for improving estimation accuracy and robustness in the case of control, sensing, and operating condition variability.