Power system reliability evaluation from a computing-electricity coupling perspective

With the advancement of technologies such as artificial intelligence and cloud computing, data centers are witnessing a continuous surge in energy consumption and scale. Simultaneously, computing resources can migrate across regions through containerization and virtualization technologies. However, current studies on power system planning and operation have not adequately accounted for the dynamic impacts of computing power migration on power system reliability. To bridge this gap, this paper develops a computing-electricity coupling model and introduces a decision-dependent uncertainty (DDU) based method for reliability evaluation. By establishing an explicit correlation model between the device’s force outage rate and computing power migration decisions, we enable a closed-loop evaluation framework integrating workload distribution, power flow, and reliability evaluation. Case studies demonstrate that the spatiotemporal dynamic scheduling capability of computing resources significantly enhances power system operational flexibility and stability. We conduct robustness testing under extreme scenarios to ensure that computational resource reallocation maintains reliability metrics within safe thresholds during extreme load fluctuations and line constraints.