研究目的
To minimize the time average expected energy cost of cloud data centers in smart microgrids considering three practical factors: the ramping constraints of backup generators, the charging and discharging efficiency parameters of batteries, and two kinds of data center workloads.
研究成果
The paper presents a real-time algorithm for minimizing the long-term energy cost of cloud data centers operating in smart microgrid environments, considering practical factors such as ramping constraints of backup generators, battery efficiency parameters, and two kinds of workloads. The algorithm's feasibility and performance are analytically proven, and simulation results demonstrate its advantages over baseline approaches in reducing total energy cost under various parameter configurations.
研究不足
The study assumes that the cost functions for conventional generators and battery depreciation are continuously differentiable and convex, which may not capture all practical scenarios. Additionally, the algorithm's performance is analyzed under the assumption that uncertain parameters are i.i.d. over slots, which may not hold in all real-world conditions.
1：Experimental Design and Method Selection:
The study employs a stochastic programming approach to model the energy management problem, integrating constraints related to workload allocation, electricity buying/selling, battery management, backup generators, and power balancing. An online algorithm is designed under the framework of Lyapunov optimization technique to solve the problem without requiring future parameters' statistical information.
2：Sample Selection and Data Sources:
Real-world workload traces and dynamic electricity price traces are used in simulations to model the workload and electricity prices, respectively.
3：List of Experimental Equipment and Materials:
The study considers cloud data centers with homogeneous servers, renewable generators, conventional generators, and energy storage systems (ESS) within smart microgrids.
4：Experimental Procedures and Operational Workflow:
The algorithm involves observing system states, solving a convex optimization problem (P5) to make control decisions, adjusting solutions to satisfy nonlinear constraints, and updating virtual queues for ESS.
5：Data Analysis Methods:
The performance of the designed algorithm is analyzed through extensive simulations, comparing it with two baselines under varying parameters such as charging and discharging efficiency, ramping coefficient, and battery depreciation parameter.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
