Carbon Neutral Data Centers

I. INTRODUCTION

  • Data centers (DCs) are experiencing vast growth, being the pillars of the 4th industrial revolution and the engine of the digitalized world.

  • DCs are carbon-intensive due to massive energy consumption.

  • It's estimated that DCs will account for 8% of global carbon emissions by 2030.

  • Technological and policy instruments for neutralizing carbon emissions haven't been thoroughly investigated.

  • Several cloud providers (Google, Facebook) have pledged to achieve carbon neutrality in their DCs.

  • Survey Paper Proposal: Proposes a roadmap towards carbon-neutral DCs including policy instruments and technological methodologies.

  • Survey first presents the carbon footprint of DCs and insights into major emission sources.

  • Discusses carbon neutrality plans for major global cloud providers.

  • Introduces the carbon market as a policy instrument for cost-efficient carbon emission offsetting.

  • Proposes achieving carbon-neutral DCs by:

    • increasing renewable energy penetration

    • improving energy efficiency

    • boosting energy circulation

  • Presents a comprehensive review of technologies on the topics above.

  • Envisions a multi-pronged approach towards carbon neutrality using a digital twin powered industrial AI framework.

  • Discusses three key scientific challenges for putting such a framework in place.

  • Presents several applications for the AI framework.

II. DATA CENTER CARBON FOOTPRINT

  • DCs attract global concerns from governments and cloud service providers because of their energy demands to serve as infrastructure of the global digital economy.

  • DCs consume 1.8% of electricity in the United States [5]

  • The share is expected to reach 20% globally in 2025 [6].

  • Electricity generation is driven by carbon-intensive fossil fuels (natural gas, coal), therefore electricity consumption will result into high carbon emissions.

  • The data center industry contributes 0.3% of global carbon emissions [7]

  • Climate change caused by excessive carbon emissions poses a major threat, and data centers pose a major threat to climate change mitigation efforts.

  • Fig. 1 shows the estimated global carbon emissions from data centers from 2018 to 2030, with escalation, indicating urgent need to reduce emissions.

  • Climate Neutral Data Center Pact [8]:

    • Formed by leading cloud infrastructure providers and data center operators in Europe, one year after the European Green Deal adoption

    • 25 companies and 17 associations agreed to a self-regulatory initiative to make European data centers climate neutral by 2030

    • Includes ambitious measurable targets such as purchasing carbon-free energy, water conservation, reuse and repair of servers, and heat recycling

  • China:

    • Government announced carbon peak and carbon neutrality targets [9].

    • Tencent is accelerating its carbon neutrality plan, achieving zero carbon emissions through AI.

    • Chindata Group aims for carbon neutrality for next-generation hyperscale data centers in China by 2030, using a 100% renewable energy solution.

  • United States [10]:

    • Google pledged to run all global data centers and corporate campuses on 100% carbon-free power by 2030 [11].

    • Microsoft declared to be carbon-negative by 2030 and remove all historical carbon footprint by 2050 [12].

  • Two ways to realize carbon-neutral data centers:

    • Reduce emissions by improving energy efficiency or increasing renewable energy penetration.

    • Balance emissions through policy instruments such as the carbon offset mechanism.

  • Surveys have been summarized on energy efficiency optimization [1], [2], renewable energy integrated data center management [3] and data center low-grade waste heat recovery [4].

    • [1]: Tutorials on how to achieve energy-proportional data centers by managing IT devices at multiple scales.

    • [2]: Comprehensive survey on joint optimization of the IT and cooling system of a data center, where enabling techniques, modeling issues, related optimization problems and testbeds for green data centers were discussed.

    • [3]: Insightful survey on how to improve renewable energy utilization and introduced several techniques as well as basic carbon market mechanisms

    • [4]: Reviewed low-quality waste heat recovery systems for data centers where several potential data center waste heat applications were discussed in terms of technical feasibility, economic applicability, and thermal efficiency..

  • None of the previous surveys presented a comprehensive framework for reducing emissions or considered the carbon credit mechanism.

  • Table I summarizes the differences between this paper and other survey papers on data center management.

  • This survey paper summarizes existing literature on carbon markets, energy efficiency improvement, renewable energy management, and waste heat recycling.

  • First presents the carbon footprint of a data center from the lifecycle and daily operations perspectives to shed light on various sources of carbon emissions.

  • Second, gives a definition of carbon-neutral data centers and metrics for evaluating carbon efficiency.

  • Presents industrial efforts from global cloud providers towards carbon-neutral data centers.

  • Introduces the carbon market as a policy instrument for achieving carbon neutrality in a cost-efficient manner.

  • Summarizes works on the optimization of renewable energy integrated data centers based on various optimization goals, and discusses common renewable energy options.

  • Presents the works on improving data center energy efficiency from both the IT and cooling system perspectives.

  • Provides a brief overview of existing applications for data center low-grade waste heat recycling.

  • Envisions a digital twin-assisted industrial AI framework for carbon-neutral data centers.

III. CARBON NEUTRALITY OF DATA CENTERS

  • Carbon footprint can be coarsely categorized into three classes:

    • carbon footprint from manufacture of the data center and its purchased equipment

    • carbon footprint from daily operations

    • carbon footprint of end-of-life disposal of equipment and decommissioning of the data center.

  • Lifecycle carbon flow of a data center depicted in Fig. 2.

  • The combustion of fossil fuels for the production of electricity (CO2 emissions) is the driving force for the carbon flow.

  • The electricity powers IT devices and infrastructure, extracts useful materials, manufactures equipment, and disposes obsolete equipment.

  • Many data centers have diesel generators for emergency power.

  • Carbon emissions will also occur in transportation of equipment.

  • Carbon emissions from material extraction and manufacturing are referred to as embodied emissions or embedded emissions, while emissions from daily operations are referred to as operational emissions.

  • If devices can be recycled and reused, the corresponding carbon emissions should be deducted from the lifecycle carbon footprint.

  • Fig. 3 depicts the operational carbon flow of a data center.

  • A data center has four subsystems: IT system, cooling system, power distribution system, and waste heat recovery system.

  • The waste heat recovery system can utilize the waste heat dissipated by servers.

  • A traditional data center power distribution system consists of the substation connected to the electrical grid, the Uninterruptible Power Supply (UPS) and the diesel generator to ensure 24 × 7 operation.

  • The on-site renewable energy generator emerged as a viable option for powering the data center.

  • An increasing number of data centers are integrating with energy storage systems to store excess energy and discharge it when an energy shortage occurs.

  • The IT, cooling, and waste heat recovery subsystems are powered by electricity, a mixture of the renewable and the nonrenewable sources, or green and brown electricity.

  • The waste heat recovery system is able to reuse the waste heat for domestic heating, driving absorption chillers, desalinating salty seawater and providing heating and drying for a biomass generator.

  • The use of waste heat allows data center operators to reduce either the carbon footprint of the data center or their customers, and to bring back carbon credits that can be used to offset their carbon emissions.

  • With the on-site renewable energy generator, the data center operator can sell excess renewable energy back to the electrical grid via the net metering technology, resulting in a reduction on the net electricity consumption from the electrical grid, as well as its carbon emissions.

  • According to the United States Environmental Protection Agency (EPA) [21], three scopes of carbon emissions of a business entity should be considered:

    • SCOPE1 Emission: Direct emission from the combustion of fossil fuels, company vehicles and any other fugitive activities such as the use of diesel generators.

    • SCOPE2 Emission: Indirect emission from the purchase of electricity, heat, or steam from local utilities.

      • Location-based method: based on the carbon emission and electricity production data and averaged within a geographical boundary during a predefined time period.

      • Market-based approach: refers to the carbon emission associated with the electricity or heat supplier.

    • SCOPE3 Emission: Covers a variety of other indirect emissions which are not covered by SCOPE2 Emission.

IV. CARBON NEUTRALITY OF DATA CENTERS

  • Definition: Carbon neutrality means that carbon emissions from an entity throughout its lifecycle are completely offset by the carbon emissions it removes from the atmosphere.

    • Carbon removal technologies: aim to remove existing carbon dioxide from the atmosphere. Typical carbon removal technologies include BioEnergy with Carbon Capture and Storage (BECCS) [26], direct air capture [27].

    • Carbon-neutral technologies: focus on zero carbon emissions by replacing carbon-intensive fossil fuels with carbon-free renewable energy, improving energy efficiency, etc.

V. CARBON NEUTRALITY METRICS

  • Carbon Usage Effectiveness (CUE):
    CUE=β⋅E<em>totalE</em>IT=β⋅PUE,CUE = β \cdot \frac{E<em>{total}}{E</em>{IT}} = β \cdot PUE,
    where:

    • ββ = carbon emission factor (kgCO2eq/kWh) of the electrical grid.

      • E<em>totalE<em>{total} and E</em>ITE</em>{IT} are the energy consumption of a data center and the IT devices within it respectively

      • PUEPUE is Power Usage Effectiveness

  • Carbon Free Energy (CFE) Score:
    CFEScore(CFE Score (%) = \frac{CFE<em>{contracted} + CFE</em>{grid}}{E_{total}},
    where:

    • CFEcontractedCFE_{contracted} is the CFE from the contracted supplier via PPA or REC

    • CFEgridCFE_{grid} is the CFE from the electrical grid

    • EtotalE_{total} is the total load of a data center.

    • CFE<em>contracted=minE</em>total,CFEtotalCFE<em>{contracted} = min{E</em>{total}, CFE_{total}}

    • CFE<em>grid=(E</em>total−CFEcontracted)⋅γ,CFE<em>{grid} = (E</em>{total} − CFE_{contracted}) \cdot γ,

    • γγ = renewable energy penetration rate of the local electrical grid.

  • Avoided Emission:
    e<em>avoided=E</em>total⋅β−(E<em>total−CFE</em>contracted)⋅β=CFEcontracted⋅β,e<em>{avoided} = E</em>{total} \cdot β − (E<em>{total} − CFE</em>{contracted}) \cdot β = CFE_{contracted} \cdot β,

where:
* ββ represents the carbon intensity of the electrical grid.

VI. CARBON MARKET

  • A Cap-and-Trade

  • Carbon Taxes

  • Carbon Credits

VII.RENEWABLE ENERGY OPTIONS FOR DATA CENTERS

  • On-Site Renewable Energy

  • Off-Site Renewable Energy

  • Power Purchase Agreement (PPA)

  • Renewable Energy Certificate (REC)

VIII. OPTIMIZATION GOALS

  • Minimize Grid Electricity Procurement

  • Minimize Operational Costs

  • Maximize Operational Profits

  • Maximize Utilization of Renewable Energy