Distributed Real-Time Energy Management in Data Center Microgrids
Liang Yu,Tao Jiang,Yulong Zou
DOI: https://doi.org/10.1109/tsg.2016.2640453
IF: 10.275
2018-07-01
IEEE Transactions on Smart Grid
Abstract:Data center operators are typically faced with three significant problems when running their data centers, i.e., rising electricity bills, growing carbon footprints, and unexpected power outages. To mitigate these issues, running data centers in microgrids is a good choice since microgrids can enhance the energy efficiency, sustainability, and reliability of electrical services. Thus, in this paper, we investigate the problem of energy management for multiple data center microgrids. Specifically, we intend to minimize the long-term operational cost of data center microgrids by taking into account the uncertainties in electricity prices, renewable outputs, and data center workloads. We first formulate a stochastic programming problem with the considerations of many factors, e.g., providing heterogeneous service delay guarantees for batch workloads, interactive workload allocation, batch workload shedding, electricity buying/selling, battery charging/discharging efficiency, and the ramping constraints of backup generators. Then, we design a realtime and distributed algorithm for the formulated problem based on Lyapunov optimization technique and a variant of alternating direction method of multipliers. Moreover, the performance guarantees provided by the proposed algorithm are analyzed. Extensive simulation results indicate the effectiveness of the proposed algorithm in operational cost reduction for data center microgrids.
engineering, electrical & electronic