Privacy-Preserving Operation Approach for Electricity-Computing Power Systems Considering Workload Migration Capability

The rapid growth of data centers and their consequent power consumption have highlighted the importance of harnessing their potential for spatial load shifting. However, the dispatching of the workload is limited by the migration capability and strictly hindered by privacy barriers of both the computing power system and electricity system. In this paper, we propose a coordinated operation approach for electricity-computing power systems in a privacy-preserving distributed manner.

The co-optimization is formulated as a mixed-integer linear programming (MILP) problem that minimizes the generation and migration costs. Specifically, the computing power system is modeled based on a flexible optical network. By incorporating joint path routing and elastic bandwidth allocation for each workload migration request, this model accurately characterizes workload migration capabilities of the computing power system to ensure reliable spatial load shifting.

Furthermore, to overcome the computational challenges posed by the inherent non-convexity and non-smoothness of the formulated MILP problem, we propose the surrogate absolute-value Lagrangian relaxation (SAVLR) algorithm to solve the problem in a privacy-preserving distributed manner. Numerical results demonstrate that geographical workload migration can alleviate the spatial mismatch between renewable generation and workload requests, thereby improving renewable energy accommodation.