Modelling, analysis, and improvement of energy consumption in data centres via demand side management

Abstract

The demand for data centers has increased in recent years due to the need for businesses to store, process, and manage vast quantities of data, as well as the swift advancements in information technology and digitalization. As more organizations adopt technologies such as artificial intelligence, big data analytics, cloud computing, and storage capacity, their requirements for computational power, rapid data access, and storage capacity grow. Furthermore, the significance of data centers has been further emphasized due to the growing prevalence of remote working options, online platforms, and other digital services. Data centers typically comprise infrastructure such as high-performance servers, data storage systems, network infrastructure, cooling systems, and power conversion systems. Particularly high-performance servers and the cooling systems employed to dissipate the heat produced by them are characterized by their exceptionally high and relatively flexible energy consumption characteristics. The high energy consumption rate of data centers has become so significant that it constituted 1.7% of global energy consumption by 2022. With such high energy consumption, the flexible load characteristic of data centers allowing energy demand to be adjusted over time makes them valuable and suitable players to participate in demand-side management (DSM). This contributes to sustainability efforts by making the energy use of data centers more effective and economical. To achieve this, it is essential to develop decision-making tools that accurately estimate data center energy demand and create optimal energy management schedules. In this chapter, an analysis of trends in data center energy consumption, power models, and their role in DSM is covered. Additionally, a case study involving the development of a decision support system comprising forecasting and optimization modules is provided. It is concluded that the proposed approach can lead to in a 16.9% decrease in the energy costs of data centers. © 2024 Elsevier B.V., All rights reserved.