Optimal coordinated management of integrated electricity-heat-computation systems in geographically distributed data centers

The rapid development of the data center (DC) industry has made it increasingly important to prioritize energy efficiency in DCs. This paper studies the cooperative dispatch of workload and energy supply equipment for geo-distributed green DCs, in which various forms of energy flow, i.e. electricity flow, heat flow and data flow, are considered. An integrated electricity-heat-computation system model is constructed, which considers diverse flexible operating characteristics of DCs, including waste heat recovery systems (WHRSs), the thermal inertia of indoor air, different types of energy supply equipment, and the spatial-temporal flexibility of computational workloads. To achieve significant environmental and economic performance, a model predictive control (MPC) based two-stage coordinated scheduling strategy is proposed, which consists of a day-ahead scheduling stage and an intra-day online dispatch stage. During the day-ahead stage, the pre-scheduling plan is obtained, while in the intra-day stage, the MPC online rolling optimization method is applied to reduce the negative impact caused by operational uncertainties of computational workloads and renewable energy sources (RESs). The proposed solution is extensively assessed through a case study, and the numerical results confirm its effectiveness and benefits in terms of economic cost, RES accommodation, CO2 emission mitigation and power usage effectiveness (PUE).