Energy Management Of Distributed Internet Data Centers In Smart Grid Environment

With the proliferation of cloud computing and Internet online services, more and more data and computation are migrated to geographical distributed Internet data centers (IDCs), which can provide reliability, management, and cost benefits. However, IDC operators en-counter several major problems in IDC operations, such as huge energy consumption and energy cost, high carbon emission. To deal with the above problems, IDC operators have to efficiently manage the way of energy consumption and energy supply. Considering the potential of smart grid, this dissertation studies the energy management of distributed IDCs in smart grid environment.Firstly, we investigate the problem of minimizing the long-term energy cost for dis-tributed IDCs in smart microgrids considering power outages. We formulate a stochastic optimization problem to minimize the time-averaged expectation of energy cost with the uncertainties in electricity price, workload, renewable energy generation and power grid s-tate. To solve the problem, we design an algorithm without requiring any knowledge about system statistics. Theoretical analysis shows the feasibility and performance guarantee of the proposed algorithm. Numerical results based on practical data show the effectiveness of the proposed algorithm.Secondly, we investigate the problem of minimizing the carbon-aware energy cost for distributed IDCs in smart microgrids. We formulate a stochastic optimization problem to minimize the time-averaged expectation of the weighted summation of energy cost and car-bon emission with the uncertainties in electricity price, workload, renewable energy gen-eration and carbon emission rate. To solve the problem, we design an algorithm without requiring any knowledge about system statistics. Moreover, we analyze the feasibility and performance guarantee of the proposed algorithm. Evaluations based on real-life data show that the proposed algorithm can achieve lower energy cost and carbon emission simultane-ously compared with the carbon-oblivious algorithm.Thirdly, we investigate the problem of minimizing the risk-constrained energy cost for distributed IDCs in smart grid environment. We propose a risk-constrained decision framework and formulate a risk-constrained expected energy cost minimization problem with the uncertainties in electricity price and workload. To solve the problem, we develop a decomposition-based cutting plane algorithm. Extensive evaluations based on real-life data show that the proposed decision framework can achieve the optimal tradeoff between operation risk and expected energy cost according to the risk preferences of IDC operators.Finally, we briefly summarize the main contents in this dissertation and propose several directions for future research.