DC microgrid clusters (DCMGCs) are a network of interconnected DC microgrids that work collaboratively to enhance resilience, reliability and optimize the economic benefits of the systems. However, due to the intricate dynamics of high order and nonlinearities, modeling for controller design can be an overwhelming task. To overcome this issue, this paper proposes a data-driven predictive control approach to regulate the power flow of DCMGCs. The proposed method employs a nonlinear autoregressive network with exogenous inputs (NARX) neural network, which does not rely on analytical modeling. Instead, the prediction model is obtained by fitting and training the system data. Consequently, the control objectives of DCMGCs can be achieved intelligently using this model-free approach. The hardware-in-the-loop (HIL) prototype of the DCMGC based on dSPACE^TM MicroLabBox and Microcontroller STM32 is built, and the proposed method is verified by HIL experiments.

DC microgrid cluster;data-driven;NARX neural network;predictive control tertiary control