Achieving Dispatchability in Data Centers: Carbon and Cost-Aware Sizing of Energy Storage and Local Photovoltaic Generation

Enea Figini, Mario Paolone · 2024

Read Achieving Dispatchability in Data Centers: Carbon and Cost-Aware Sizing of Energy Storage and Local Photovoltaic Generation by Enea Figini, Mario Paolone (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2412-13853.txt.

Introduction

Enea Figini and Mario Paolone, two authors writing in 2024, size green power gear for data centers. Data centers draw huge amounts of power for servers and cooling. Their demand is set to grow with new trends in computing, raising both costs and emissions.

The authors offer a method to size solar panels and batteries for an existing center. The plan aims to cut cost, cut emissions, and make demand steady enough to schedule ahead. It counts the full life cycle of the gear and the shifting emissions of the grid. Case studies in two countries show the best sizes depend strongly on place.

Catalog record

CREATOREnea Figini, Mario Paolone
YEAR2024
COLLECTIONPAPERS
FORMATTXT
FILE SIZE73 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2412-13853

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Achieving Dispatchability in Data Centers: Carbon and Cost-Aware Sizing of Energy Storage and Local Photovoltaic Generation Enea Figini, Mario Paolone Abstract Data centers are large electricity consumers due to the high consumption needs of servers and their cooling systems. With the rapid growth of crypto-currency and artificial intelligence, their electricity consumption is expected to increase substantially. With the electricity sector being responsible for a large share of global greenhouse gas (GHG) emissions, it is important to lower the carbon footprint of data centers to meet GHG emissions targets set by international agreements. Moreover, uncontrolled data center integration into power distribution grids increases the stochasticity of electricity demand, thus increasing the need for reserve capacity and leading to operational inefficiencies and higher emissions. This work provides a method to size a PhotoVoltaic (PV) system and an Energy Storage System (ESS) for an existing data center looking to reduce both its carbon footprint and demand stochasticity through day-ahead dispatching. A scenario-based optimization framework is developed to jointly size the PV and ESS, minimizing the expected operational and capital expenditures and the carbon footprint of the data center complex. The model considers the life cycle assessments (LCA) of the systems and the dynamic carbon intensity of the upstream electricity supply. Case studies in different Swiss cantons and regions of Germany emphasize the need for location-aware sizing processes since the obtained optimal solutions strongly depend on the local electricity carbon footprint and on the irradiance conditions. The maximum carbon footprint reduction reaches approximately 50% in Germany and 4% in Switzerland. Installed power generation and energy storage capacities vary by up to 36 times across regions. 1 Introduction Data centers consumed 460\text{\,}\mathrm{TW}\text{\,}\mathrm{h} of electricity worldwide in 2022, representing almost 2% of the total electricity demand [ 1 ] . The International Energy Agency (IEA) expects the global annual electricity demand from data centers to range between 650 and 1050\text{\,}\mathrm{TW}\text{\,}\mathrm{h} in 2026, potentially accounting for up to 4.5% of the global electricity demand. In Ireland and Denmark, the data center electricity consumption represents respectively 13% and 8% of the country’s demand and it is expected to grow to 30% and 20% , respectively, by 2026. With the electricity and heat sector accounting for 14.8 billions of tons of \text{\,}\mathrm{C}\mathrm{O}_{\mathrm{2}}\mathrm{e}\mathrm{q} [ 2 ] (i.e., approximately 25% of global GreenHouse Gases (GHG) emissions worldwide), reducing the carbon footprint of data centers is important to meet GHG emissions targets set by international agreements (e.g., [ 3 ] ). Additionally, uncontrolled data centers tend to increase the stochasticity of the power system demand and to contribute to an environmentally and economically suboptimal operation of power grids [ 4 , 5 ] . For these reasons, carbon-aware solutions for data centers have been widely explored in the literature. An advanced search on IEEE Xplore combining the keywords “green” and “data center” for the period 2020–2024 results in over 2,500 publications [ 6 ] , covering strategic planning/design, modeling, energy efficiency, green workload schedulers, waste heat recovery, etc. For example, in [ 7 ] , a cooling control algorithm using reinforcement learning is presented and studied using a simulation model. The algorithm trains two neural networks predicting the best next control action and the system state (in this case Power Usage Efficiency PUE) and server room temperature. In [ 8 ] , the authors present an optimization problem to schedule tasks in Distributed Green Data Centers (DGDCs), considering local factors (s.a. electricity prices, wind speed, solar irradiance, etc). They maximize a biobjective function consisting of the aggregate revenue and Quality Of Service 1 1 1 The quality of service quantifies the performance of a data center to serve the workloads/tasks of users. (QoS) and use classified M/M/1 queuing models 2 2 2 An M/M/1 queue models the length of a queue in a single server system where job arrivals follow a Poisson process and execution times follow an exponential distribution. Such queue models are widely used to study workload scheduling performance for example. to model workload behaviour. As a final example, [ 9 ] showcases a framework to efficiently place data center locations to minimize the emissions of the power network under study, calculating the electricity dispatch of the system through a direct current optimal power flow. The paper shows that locating data centers close to renewable generation can be suboptimal and that grid operations and limits should be considered in the placement process. In practice, Data Center Operators (DCO) often purchase renewable energy or renewable energy credits to match their annual demand [ 10 , 11 ] . This approach has been shown to be inefficient and misleading as it cannot guarantee an actual reduction of the carbon footprint of the entity making the purchase [ 12 , 13 ] , thus highlighting the need for precise and granular carbon accounting [ 14 ] . 1.1 Background Recent literature highlights the relevance of integrating renewable energy systems and energy storage technologies to improve dispatchability and sustainability (e.g., [ 15 , 16 , 17 ] ). Various energy storage technologies are studied in [ 15 ] , with recommendations of technologies for diverse applications, such as energy management, voltage regulation, load leveling, etc. A recent and comprehensive review of hybrid energy storage systems integrated with renewables is provided in [ 17 ] . The study compares a wide range of storage technologies including batteries, thermal storage, compressed air, flywheels, and hydrogen storage. The use of these technologies is assessed in terms of cost, efficiency, maturity, environmental friendliness, etc. Moreover, it reviews optimization approaches for sizing and operation, including stochastic multi-objective optimization and techno-economic models. Notably, battery energy storage systems are highlighted as particularly well-suited for high-power, high-energy-demand applications, such as the ones associated to modern data center operations. Building on this, methods for power system optimization under uncertainty are reviewed in [ 18 ] and stochastic optimization is generally identified as crucial in capturing the uncertainty in integrated energy systems planning and operation [ 19 ] . These approaches are particularly relevant for data centers, which increasingly rely on the integration of renewable generation and require flexibility to achieve reliability objectives under uncertain conditions. Strategic planning of data centers has been studied with a general focus on methods to reduce their operational costs. For instance, [ 20 ] formulates a multi-objective optimization problem to determine the optimal capacity and location of a data center and associated energy storage systems, minimizing the costs and maximizing the QoS. [ 21 ] proposes a day-ahead scheduling of data centers based on the concept of a Virtual Power Plant (VPP), leveraging the flexibility of workloads (i.e., using workload migration and shifting 3 3 3 Workload migration refers to the process of moving workloads from one computing environment to another, while workload shifting refers to moving workloads in time. ), and the flexibility of back-up power devices (i.e., UPSs 4 4 4 Uninterruptible Power Supplies (UPS) are back-up power systems used to mitigate the risks on the data center supply unavailability due to temporary black-outs of the power grid. ). Some works on the planning of data centers consider their direct carbon footprint. For instance, [ 22 ] formulates a framework to select energy resources, balancing energy

Read the complete text