Aiming at the voltage flicker induced by the coupling of illumination fluctuation and energy storage state switching in PV (photovoltaic)-storage NES (new energy system), an adaptive suppression strategy of hierarchical dynamic modeling and model prediction is proposed. In the outer layer, a unified equivalent circuit of PV-MPPT, energy storage bidirectional DC-DC and grid-connected inverter is constructed based on Simscape. In the inner layer, the cloud cover is described by Markov chain, and a second-order RC battery model dependent on state of charge (SOC) is introduced to realize high-precision flicker feature prediction with an error of ≤5%. In the control layer, rolling time domain model predictive control (MPC) is designed, which takes flicker power, energy storage cycle loss and SOC deviation as multi-objective cost functions and embeds power balance, SOC 20%–90% and power change rate constraints. Furthermore, three adaptive mechanisms are proposed, namely, virtual inertia based on frequency change rate, damping coefficient based on SOC partition and power change rate penalty, to achieve more than 90% flicker suppression rate and 40% reduction of charging and discharging switching times. The dSPACE HIL experiment verifies the effectiveness and robustness of the proposed method in typical irradiation sudden change and periodic fluctuation scenarios.
Zhang et al. (Sun,) studied this question.