Computing-Electricity-Water Synergy-Based Expansion Planning of Sustainable Data Center

Abstract

This paper proposes a stochastic-robust computing-electricity-water planning strategy for the cost-effective deployment of sustainable data centers (DCs). This strategy not only considers the tight coupling interactions of energy, workload and water flow to meet dynamic service requirements of DCs, but also can co-optimize the investment decisions related to electricity and water delivery network expansions to mitigate long-term siting risks. Moreover, a temperature-aware power and water demand model is proposed to enhance operational flexibility in DC expansion planning, and a modified thermodynamic formulation of DCs is formulated to explore the interdependencies between electricity and cooling water demands of DCs on ambient temperature fluctuations by integrating a variable resistor. To address the fact that higher computing capacities of DCs will attract more workloads, a stochastic-robust programming approach is employed, which considers both decision-independent uncertainties from renewable generation, electricity load and water consumption, as well as demand-induced uncertainties related to investment behaviors. Furthermore, a Parametric Column-and-Constraint Generation (PC&CG) algorithm with multiple linearization techniques is presented to alleviate the computational complexity of the original DC planning problem and quantitatively assess demand-driven workload uncertainties. Simulation results validate the effectiveness of the proposed strategy in achieving an economical investment portfolio of sustainable DCs.