Distributionally robust optimization scheduling model for electric power and computing power coordination considering spatiotemporal response

With the advent of the digital era, the data generated by the application of information technology in various industries has grown explosively, leading to a continuous increase in the energy consumption of computing. This trend reshapes the power demand structure and exacerbates power supply and demand imbalances. Therefore, it is necessary to explore the potential of data centers in regulating power systems and propose a data-driven power-computing collaborative distributionally robust optimization scheduling model. Firstly, a spatiotemporal response model for data center computing tasks is established. Subsequently, in addition to wind and photovoltaic power, the uncertainty of computing tasks has also been considered. Scenarios are generated by integrating the Copula function with the Wasserstein Generative Adversarial Network with gradient penalty (WGAN-GP). The model is then improved to a two-stage distributionally robust optimization model through weighted Composite Norm and efficiently solved using the C&CG algorithm. Finally, the simulation analysis and case study demonstrate that the spatiotemporal regulation of computing smooths the net power load curve, effectively alleviating peak power supply pressure, and promoting renewable energy integration. The introduction of norms and confidence levels diminishes the reliance on intuitive experience, making the decisions formulated more comprehensive. The improved distributionally robust optimization model balances the economic performance of stochastic optimization and the safety of robust optimization, and the operability and flexibility of the model are enhanced.