Sparse Identification for Data-Driven Dynamics and Impedance Modeling of Power Converters in DC Microgrids

Confidentiality and variability of the parameters of power converters motivate applying system identification in power electronic systems. In this article, a novel identification framework based on the sparse regression technique is proposed for dc–dc converters operating in a microgrid in order to derive the nonlinear dynamics of the converter under test (CUT) for data-driven control and to estimate the small-signal impedance of the CUT for online stability analysis purposes. The CUT is perturbed by swept-sine and pseudorandom binary sequence excitation signals via an external source in order to collect dynamically rich time- and frequency-domain measurements during the online operation of the CUT. The measurements are then fed to an optimization problem solved by the sequential thresholded least-squares algorithm to discover the nonlinear averaged dynamics and the sparse parameter-varying (SPV) impedance model of the CUT. The data-driven controllers designed using the identified dynamics are shown to exhibit close performance to their physics-based counterparts. Furthermore, the SPV impedance model is also shown to accurately track the measured impedance of the CUT and obviates the need for continuous perturbation of the system at each operating point. Real-time simulation results and frequency-domain analyses are presented to verify the effectiveness of the proposed framework.