Coordinated planning and operation of SST-based DC data centers with tiered workload flexibility

The rapid growth of artificial intelligence (AI) workloads is making modern data centers increasingly energy-intensive and reliability-critical. This trend not only raises electricity expenditure but also amplifies the economic impact of power interruptions, motivating coordinated planning of power infrastructure and computing resources. Against this background, solid-state transformer (SST)-based DC supply architectures are emerging as a promising direction for next-generation data centers because they provide a controllable interface for integrating diverse energy and computing resources. Under this architectural setting, this paper proposes a coordinated planning and operation framework for SST-based DC data centers with tiered workload flexibility. The framework incorporates nonlinear SST losses into long-term capacity planning, treats tiered workload flexibility as a planning resource, and coordinates energy storage and diesel generators for backup supply during utility outages. An annual bi-level model is formulated to jointly determine the capacities of SST conversion stages, renewable generation, energy storage, diesel generators, and heterogeneous computing servers, together with their scenario-dependent operating decisions. Piecewise linear approximation is used to reformulate the SST loss model, and an enhanced Benders decomposition algorithm is developed to solve the resulting multi-scenario mixed-integer problem. Case study results demonstrate that the proposed framework reduces total annual cost while satisfying workload service constraints. Further analysis shows that nonlinear SST losses influence capacity configuration, whereas tiered workload flexibility reshapes server deployment and backup resource requirements.