<p>Large-scale wind farms with hundreds of megawatts capacity typically show minimal discrepancies in power curves across minute and hour scales due to the power aggregation effect. However, as capacity decreases to the hundreds of kilowatts level in microgrids, this discrepancy increases significantly. Existing literature on two-stage robust planning for wind-powered microgrids has overlooked the substantial differences in fluctuation ratios of small-capacity wind power across different time scales. This oversight results in a critical limitation: Equipment capacities configured based on hourly scale data often fail to adapt optimally during minute-level dispatching. This paper proposes a two-stage distributed robust economic optimal dispatch strategy for microgrids, leveraging empirical mode decomposition (EMD). First, we analyze the phenomenon of overestimation or underestimation of wind power fluctuation uncertainties across different time scales. Subsequently, the EMD method is employed to design a high-frequency flywheel energy storage system, which converts significant minute-scale wind power fluctuations into minor fluctuations at the same time scale. This process promotes the convergence of wind power curves between minute and hourly levels. Finally, using the stabilized hourly scale wind power curves, we conduct two-stage distributed robust planning for low-frequency cold storage tanks and lithium bromide chillers to validate the economic efficiency and robustness of the proposed strategy. A case study on a wind-powered data center microgrid in Jilin Province is performed to verify the strategy’s effectiveness.</p>