Abstract
Power electronic systems are widely used for efficient power processing across various applications. However, due to their nonlinear behavior and susceptibility to disturbances, modeling and simulation are essential for fully understanding their complex interactions. The waveform relaxation (WR) method is employed to solve subsystems independently, with data exchanged until convergence is reached, offering flexibility in selecting solvers and integration techniques. In power electronics, applying the WR method with time windowing improves simulation stability and accuracy by introducing an equation to determine the optimal number of time windows. A time windowing technique is proposed, for the first time, for the co-simulation of a closed-loop power circuit divided into a first subsystem (converter) and a second subsystem (Proportional-Integral controller). These two subsystems are supposed to be designed by two different entities, which are not allowed to exchange their respective models for intellectual property reasons. The time windowing WR method is applied to transfer the data from subsystem 1 to 2. More importantly, an analytical model for the implementation of a time windowing WR technique is developed. This equation allows a deterministic choice of the number of time windows at each iteration, for convergence, based on windowing size and simulation parameters. Results show consistent output within a specified voltage range, with reduced CPU simulation time compared to co-simulation with an arbitrary selection of numbers of time windows, while being comparable to the full system simulation. The proposed technique is implemented with an input voltage range of 1.8-3 $V$, resulting in consistent outputs of 3.3 $V$ in both continuous- and discrete-time simulation modes. The comparative analysis demonstrates that with an increased number of iterations, the elapsed time is 11.1% and 5.6% faster for continuous- and discrete-time simulation modes, respectively, compared to a co-simulation with time windows. Furthermore, the time windowing WR method closely aligns with the full system result, with only a slight difference in elapsed time. Additionally, to validate the proposed analytical model using the time windowing WR technique in a closed-loop circuit, the buck converter is evaluated and compared without the full system. These results highlight the accuracy of WR with time windowing method and illustrate the quicker convergence achieved when used in circuit co-simulation.
Funding Information
  • Agence Nationale de la Recherche (ANR-19-CE05-0016-05)

This publication has 28 references indexed in Scilit: