Probabilistic assessment method for demand response potential of 5G base stations considering user spatiotemporal distribution

5G base stations (BSs), with flexible communication power and backup energy storage, are emerging demand-side resources for power-system regulation. Accurately quantifying their demand response (DR) potential is essential for effectively exploiting this flexibility. Existing studies that assess BS flexibility under prescribed communication conditions may yield inaccurate DR-potential boundaries, as the actual DR potential varies with communication states and backup-energy requirements that are shaped by users' spatiotemporal distribution. This paper proposes a probabilistic DR potential assessment framework for 5G BSs considering user spatiotemporal distribution. First, an Evolving Graph Convolutional Network is used to probabilistically predict grid-level user distributions, and spatially correlated scenarios are generated using a Gaussian covariance function. Second, cluster-level communication traffic is mapped to users' quality of service (QoS) requirements. Finally, the DR potential of each scenario is evaluated by comparing the maximum-energy-efficiency baseline plan with the minimum-power optimization plan, while incorporating backup energy storage flexibility. A case study in Minhang District shows that the DR potential assessed for a typical day accounts for 43.61% to 54.02% of the baseline power, demonstrating substantial flexibility of 5G BSs.