[Objective] Due to their energy storage capabilities， electric vehicles （EVs） can serve as flexible resources for grid interaction. The large-scale integration of EVs into microgrids creates new opportunities for optimal scheduling.This paper proposes a two-layer stochastic model predictive control （SMPC）-based optimal scheduling strategy for microgrids.[Methods] In the upper-layer， scenario analysis is employed to handle uncertainties in photovoltaic output， load demand， and the number of EVs. A set of representative scenarios is generated， and an optimization model is established to achieve economic dispatching through rolling optimization. In the lower layer， a dynamic power allocation strategy for charging piles is developed based on a broadcast control method. This strategy enables efficient allocation of upper-layer dispatch instructions while considering the charging and discharging regions of EVs.[Results] Case studies indicate that the proposed optimal scheduling method enhances robustness while maintaining economic efficiency， outperforming both deterministic models and robust optimization （RO） approaches. Moreover， the proposed power allocation strategy for charging piles reduces communication burdens compared with traditional centralized power allocation strategies and demonstrates satisfactory tracking performance.[Conclusions] The proposed optimal scheduling strategy improves the economic efficiency and flexibility of microgrid operation while satisfying EV charging demands. The broadcast control algorithm balances communication efficiency and system scalability， making it suitable for large-scale， plug-and-play scenarios.