[Objective] High penetration of renewable energy sources(RESs) and grid-connected electric vehicle(EV) clusters introduces strong stochasticity and volatility on both the source and load sides

which challenge the secure and economic operation of microgrids. Demand-side flexibility provides an effective means to buffer renewable intermittency and improve source-load balance. This study aims to exploit such flexibility to enhance renewable accommodation and economic performance under wind and solar generation uncertainty. [Methods] This paper proposes a bi-level day-ahead economic dispatch strategy that integrates Minkowski-sum-based flexibility aggregation with wind-solar uncertainty quantification. First

Minkowski Sum theory is applied to aggregate feasible dispatchable boundary sets of heterogeneous EV clusters with different battery capacities

arrival-departure times

and charging preferences

and a unified virtual energy storage(VESS) model is established to represent vehicle-side flexibility with charging/discharging power limits

energy capacity constraints

and feasible operating regions. Second

Latin hypercube sampling(LHS) is used to generate wind and photovoltaic trajectories

and kmedoids clustering is employed to extract a small set of typical scenarios that preserves statistical diversity for tractable optimization. The upper level determines system schedules to minimize operating cost

while the lower level updates agent decisions in response to prices and constraints. Based on the scenario set

a dynamic pricing mechanism is designed to maximize the covariance between demand response loads and wind-solar outputs

so that flexible loads are incentivized to follow renewable availability and net-load fluctuations are mitigated. Finally

a multi-agent cooperative game model is formulated

and coalition cooperation is introduced to enable energy sharing among agents and to allocate costs and benefits through a profit-sharing mechanism while satisfying individual rationality and local constraints. [Results] Case studies were conducted on a representative microgrid with high RES penetration and EV integration. Compared with the non-cooperative mode in which each agent operated independently

the proposed strategy reduced the total operating cost by approximately 24 680.5 yuan/d and improved the spatiotemporal matching between load demand and wind-solar output. The optimized schedules also indicated reduced renewable curtailment and smoother net-load profiles while satisfying EV charging requirements and operational constraints

and the coalition mechanism improved participation incentives. [Conclusion] The proposed source-load-storage coordinated dispatch strategy aggregates heterogeneous EV flexibility via VESS modeling

quantifies wind-solar uncertainty via scenario generation and reduction

and strengthens renewable-load coupling through covariance-oriented pricing and coalition-based cooperation. It achieves coordinated improvement in economic performance and operational stability under multi-agent benefit equilibrium

providing a practical dayahead scheduling approach for microgrids with strong source-load randomness.