To address the complex challenges in active power balance control across multiple timescales, coordination of diverse dispatchable resources, and management of multiple uncertainties in urban power grid with high-penetration green power, this study establishes an optimal day-ahead risk-coordinated dispatch model. This model coordinates pumped hydro storage, large-scale battery storage, clustered small-scale battery storage, and conventional units for active power regulation. To minimize total daily operation costs, operation mechanisms leveraging the complementary strengths of centralized and distributed energy storage are proposed. This mechanism simultaneously addresses second-level step disturbances, minute-level non-step disturbances, and hourly peak shaving uncertainties, achieving optimal coordinated day-ahead reserve capacity allocation for diverse regulation resources. For refined probabilistic characterization of wind power uncertainty, we employ a high-precision forecasting method combining long short-term memory networks with indirect quantile regression, providing more reliable theoretical support for stochastic management. To precisely measure potential active power imbalance risk, conditional value-at-risk theory quantifies wind curtailment and load shedding losses across multiple time scales. Simulation on a modified IEEE 39-bus system with 71% wind penetration validates the effectiveness and superiority of the proposed optimal coordinated scheduling strategy. Results demonstrate that the proposed coordination approaches of centralized and distributed storage ensure secure and stable frequency operation in urban grids.