[Objective] Driven by the carbon peaking and carbon neutrality goals and the strategy of national computing network to synergize east and west， data centers have become crucial high-density flexible loads， making the coordinated optimization of their green transition and power systems increasingly vital. Focusing on the policy requirement for new data centers at hub nodes to achieve 80% green electricity consumption， this paper aims to explore the impact mechanisms of diverse green electricity consumption modes and computing load flexibility on the operational costs and carbon emissions of data centers， providing a decision-making basis for resource planning and coordinated dispatching in computing hubs. [Methods] A bidirectional coordinated optimization model for computing-power nodes is constructed， comprehensively considering the temporal and spatial flexibility of computing tasks alongside the physical and contractual constraints of diverse green electricity consumption modes， including direct green energy connection， power purchase agreements （PPA）， and grid power supplementation. The levelized cost of electricity （LCOE） and consequential carbon emissions are introduced as dual evaluation metrics. Using the integrated computing-power project group in the Ganzi region as a case study， simulation scenarios are designed with a gradient decrease in the direct green energy connection ratio from 80% to 20%， comparing PPA and grid supplementation strategies while exploring the operational characteristics of data centers under four schemes： no flexibility， temporal flexibility， spatial flexibility， and spatio-temporal synergistic flexibility. [Results] Simulation results indicate that under the 80% green electricity consumption constraint， the LCOE of data centers exhibits a significant "U-shaped" evolution as the direct green energy connection ratio decreases. The dispatching under the spatio-temporal synergistic flexibility achieves a "Pareto improvement" in both economic and environmental benefits through full-dimensional resource optimization， serving as the optimal dispatching paradigm for the low-carbon and economic operation of large-scale integrated computing-power clusters. [Conclusions] The proposed model and evaluation framework quantify the cost boundaries between physical connection and virtual matching and effectively reveal the impact laws of flexibility strategies， offering valuable reference for optimizing the configuration of data centers at computing hubs.