Digital Twin-Based Online Health Monitoring of Power Electronics Systems With Self-Evolving Compensators and Improved Parameter Identification Capability

Abstract

Power electronics systems (PESs) are crucial for energy conversion and control in various sectors, such as aerospace, renewable energy, and electric vehicles. Health monitoring using digital twin (DT) technology is crucial for fault diagnosis, system design, and maintenance in PES, enhancing system reliability and performance. This study first compares the parameter estimation capabilities of three metaheuristic methods: particle swarm optimization (PSO), grey wolf optimization, and the dragonfly algorithm (DA). After that, a two-stage metaheuristics method is proposed, considering physical behavior to enhance the accuracy of estimating parasitic resistances and rapidly identifying PES parameters. Compared to the traditional PSO, the proposed two-stage PSO method improves MOSFET and inductor’s parasitic resistance estimation errors from 31% and 45% to 1.5% and 2.3%, respectively, and reduces computation time by over 60%. Next, parameter identification accuracy and robustness under various external disturbances are also investigated. Next, parameter identification accuracy and robustness under various external disturbances are also investigated. Based on the test results, the proposed method can reduce the error in MOSFET parasitic resistance by up to 31.7% and 11.8% on average. In addition, it can decrease the error in inductor parasitic resistance by a maximum of 44.6% and 16.7% on average. Furthermore, this article introduces a self-evolving compensator that automatically adjusts controller parameters online based on identified component values. This approach addresses the age of PES and improves step response errors by up to 10.6%.