<p>Renewable energy, especially wind power, is playing an increasingly significant role in the power system. However, due to the intermittency of wind, the uncertainty of wind power generation posed significant challenges for microgrid scheduling. Therefore, this paper proposed a robust optimization model driven by wind power uncertainty set, where the uncertainty set was adjusted according to the prediction results. Firstly, based on historical wind power data, a Conditional Normal Copula (CNC) model was established using Copula theory to describe the dependency structure between wind power prediction results and errors. Then, a large number of wind power samples were generated using Quasi Monte Carlo (QMC) based on the prediction results, and wind power confidence intervals were derived using the imprecise Dirichlet Model (IDM). These intervals were then converted into uncertainty sets to reduce the redundancy of uncertainty. The set considered the conditional correlation between wind power prediction results and errors, reducing reliance on historical data and more accurately describing the uncertainty of wind power. Based on the aforementioned wind power uncertainty set, a two-stage robust scheduling model was established. The simulation results demonstrated that the day-ahead operating costs were 6829.9 yuan, and electricity purchases were 265.8 kWh, indicating optimal performance overall.</p>