利用人工智能的力量減少碳排放并降低成本

ReduceCarbonandCostswiththePowerofAI

January2021

ByCharlotteDegot,SylvainDuranton,MichelFrédeau,andRichHutchinson

Leadinginthenewreaty

climate&sustainability

BostonConsultingGrouppartnerswithleaders

inbusinessandsocietytotackletheirmost

importantchallengesandcapturetheirgreatest

opportunities.BCGwasthepioneerinbusiness

strategywhenitwasfoundedin1963.Today,

weworkcloselywithclientstoembracea

transformationalapproachaimedatbenefitingall

stakeholders—empoweringorganizationstogrow,

buildsustainablecompetitiveadvantage,and

drivepositivesocietalimpact.

Ourdiverse,globalteamsbringdeepindustryand

functionalexpertiseandarangeofperspectives

thatquestionthestatusquoandsparkchange.

BCGdeliverssolutionsthroughleading-edge

managementconsulting,technologyanddesign,

andcorporateanddigitalventures.Weworkina

uniquelycollaborativemodelacrossthefirmand

throughoutalllevelsoftheclientorganization,

fueledbythegoalofhelpingourclientsthriveand

enablingthemtomaketheworldabetterplace.

Foreword

Forbusinessesandgovernmentstoday,thereisperhapsnoth-ingmoreurgentthanfindingapathtonetzero.Thescienceisclear:inordertoavoidtheharmfuleffectsofclimatechange,greenhousegasemissionsmustbecutinhalfthisdecadeandwemustachievenetzeroby2050.That’sadifficulttask,butalsoahistoricopportunity.Countriesandcompaniesthat

proactivelytakeactiononclimatechangeandembracethe

net-zerotransitionareinthepositiontogainacompetitive

advantageandavoidbeinglockedoutofresources,technolo-gies,andtalentintheshortrunwhileplayingavitalrolein

savingtheplanet.Achievingthiswillrequirecreativity,innova-tion,andcollaboration.

BCGbelievesthatthiswillbeadefiningdecadeforclimateactionandwefeelitisourroletohelporganizationsfind

solutionsthatwillsparkthegreatestchangeandspeedupprogresstowardnetzero.Thefirmistheworldleaderin

workingwithgovernments,industries,sectors,andcompa-niestoimplementscaledtransformationsthatproduceapositiveimpact.Wehavethecapabilitiestosupportorgani-zationsinsectorswhereemissionabatementiseasierto

achieveorwhereemissionsaregreatestandthechallengeismoresubstantial.Withourexperience,foranytypeof

organization,wearewell-positionedtohelpdesignand

implementstrategiesthatbothfacilitatetheclimatetran-sitionandhelpbuildacompetitiveadvantageforclients.

Whilemanyorganizationsaremakingsignificantpledgestoreducetheiremissionsfootprint,decisivemovesto

quicklyreduceitarelesscommon.Thismayrepresenta

missedopportunity,ascrucialsignstodaypointtotherealvalueofanet-zerostrategy.Customers,investors,and

employeesareincreasinglydemandingthatcompanies

theydobusinesswithorworkformakerealprogresson

cuttingtheirnetcarbonemissions.Furthermore,pressurefrompolicymakersismounting.Asaresult,embedding

solutionstothosechallengesintotheheartofacompany’sbusinessmodelcanyieldoutstandinglong-termvalue

creation.Andwhensocialimpactisprofitable,it’sscalable.It’sacyclethatbenefitsusall.

Webelievethatcompaniesandgovernmentsneedtoaskthemselvesfourkeyquestionstodrivetheirnet-zero

strategies:

?Whatdoesmymarket/countrylooklikeinanet-zeroworld?

?Whatismymodelforsucceedinginthatmarket?

?WhatchangesshouldImakenowtoprepareforsuc-cess?

?Whataretheexternalmarketplace,regulator,andenvi-ronmentalconditionsthatIwillfaceinmovingtowardnetzero?

Withinthisdocument,youwillfindarelevantBCGarticlefocusingonthenet-zerotransitionviewedthroughthe

lensoftechnologies,operations,andstrategiesandthe

rolethatBCGcanplayinhelpingcompaniesandgovern-mentsimplementthesenewapproaches.TheParis

Agreementof2015establishedtheprocesstocollectivelyreduceglobalgreenhousegasemissionsintheatmo-

spheretolimitglobalwarmingtowellbelow2degrees

Celsius.TheCOP26conferenceinGlasgowthisNovemberwillprovideawindowintohowwellwearedoinggloballyinachievingthesegoals,makingthisaperfectmoment

foreachorganizationtorevisit(ordesignforthefirst

time)theirownnet-zerostrategies.Itistimetomove

frompledgestoaction.Wehavelessthanadecadelefttohalveouremissions—andembraceamassiveopportuni-tyfortheplanet,forourselves,andforthepeoplewho

dependuponourorganizationstodotherightthing.

BOSTONCONSULTINGGROUP1

T

hepressureonbusinessestorespondtothethreatofglobalwarmingisgrowing.Consumers,regulators,

andinvestorsalikeareincreasinglyscrutinizingtheclimateimpactofcompaniesineveryindustry.InhisJanu-ary2020lettertoCEOs,forexample,LarryFink,chairmanandCEOofBlackRock,theworld’slargestassetmanager,putcompaniesonnoticethatinvestors—amongother

stakeholders—nowexpectfulldisclosureofcompanies’performanceonarangeofenvironmental,social,and

governancefactors.

Butit’sonethingforcompaniestopaylipservicetothe

needtoreducetheirgreenhousegas(GHG)emissions.

Takingconcretemeasurestomakeadifference,especiallyintoday’spandemic-driveneconomicclimate,isanother

matterentirely.Thedifficultyandexpenseofmeasuring

thefullextentoftheircarbonemissions,andthenreduc-

ingoroffsettingthem,hasforcedmanycompaniestodelaytheeffort.

Inthiscontext,artificialintelligence(AI)canbeagamechanger.Itsabilitytodeliverdeepinsightsintomultiple

aspectsofacompany’scarbonfootprintandquick

cost-cuttingwinsoffersapromisingroutetoacceleratingsustainabletransformationandreducingexpensesina

timeofneed.Andbecausetheirsizegivesthemaccesstohugedatasets—akeysuccessfactorfordeployingAI—

largecompaniesareinanespeciallystrongpositiontobenefitfromitspower.

ThePromiseofAI

Thethreatofclimatechangeisgrowing,andtimeisrunningout.GlobalGHGemissionscurrentlytotalabout53gigatonsofcarbondioxideequivalent(CO2e),accordingtotheCarbonDisclosureProject.Ifwearetomeetthegoaloflimitingtheincreaseinaverageglobaltemperaturesto1.5°C,asspecifiedinthe2016ParisAgreement,wemustreducethoseemis-

sionsby50%bytheendofthisdecade,accordingtothe

Science-BasedTargetsInitiative.Inourexperiencewithanumberofclients,usingAIcanachieve5%to10%ofthatneededreduction—between2.6and5.3gigatonsofCO2e.

Meanwhile,BCGstudiesshowthatthepotentialoverallimpactofapplyingAItocorporatesustainabilityamountsto$1.3trillionto$2.6trillioninvaluegeneratedthroughadditionalrevenuesandcostsavingsby2030.

Thisadded-valuefigureforcompaniesdoesnottakeinto

accountchangesinthepriceofcarbonoffsets.Thatnum-ber,currentlysetataround$30intheEUEmissionsTrad-ingSystem,coulddoubleby2030.BCGexpectstospend

$80pertonby2030onhigh-quality,permanentGHG

removalaspartofits

NetZeropledge

.Atthisincreased

pricelevel,thevalueofreducingGHGemissionsthrough

theuseofAIwouldrepresentanadditionalsavingsof$208billionto$424billionforallcompaniesglobally.Ifcarbonoffsetpricesriseevenhigheroverthecomingyears,AI

opportunitieswillsurelyrepresentevengreatersavings.

In2030,usingAIforclimate

controlcouldhelpreduce

2.6to5.3

gigatons

ofGHGemissions,or

5%to10%ofthetotal

andcouldprovide

$1trillion

to

$3trillion

invalueaddedwhenappliedto

corporatesustainabilitygenerally

Source:BCGanalysis.

2REDUCECARBONANDCOSTSWITHTHEPOWEROFAI

HowItWorks

ThegreatstrengthofAIliesinitsabilitytolearnbyexperi-ence,collectingmassiveamountsofdatafromitsenviron-ment,intuitingconnectionsthathumansfailtonotice,andrecommendingappropriateactionsonthebasisofits

conclusions.Companieslookingtoreducetheircarbon

footprintshouldturntheAIspotlightonallthreecompo-nentsoftheeffort:

?MonitoringEmissions.CompaniescanuseAI-powereddataengineeringtoautomaticallytrackemissions

throughouttheircarbonfootprint.Theycanarrange

tocollectdatafromoperations,fromactivitiessuchascorporatetravelandITequipment,andfromeverypartofthevaluechain,includingmaterialsandcomponentssuppliers,transporters,andevendownstreamusersoftheirproducts.AIcanexploitdatafromnewsources

suchassatellites.Andbylayeringintelligenceontothedata,AIcangenerateapproximationsofmissingdataandestimatethelevelofcertaintyoftheresults.

?PredictingEmissions.PredictiveAIcanforecastfutureemissionsacrossacompany’scarbonfootprint,in

relationtocurrentreductionefforts,newcarbonreduc-tionmethodologies,andfuturedemand.Asaresult,thecompanycanset,adjust,andachievereductiontargetsmoreaccurately.

?ReducingEmissions.Byprovidingdetailedinsight

intoeveryaspectofthevaluechain,prescriptiveAI

andoptimizationcanimproveefficiencyinproduction,transportation,andelsewhere,therebyreducingcarbonemissionsandcuttingcosts.

Inshort,AIcanhelplargecompaniesreducetheirenviron-mentalimpactwhilealsoalleviatingthefinancialpressuretheyfaceastheyemergefromtheCOVID-19crisis.

Industriesthatcanbenefitfromthisapproachinclude

industrialgoods

(seethesidebar“ASteelmakerCuts

EmissionsandCostswithAI”)

,transportation,pharmaceu-tical,consumerpackagedgoods,energyandutilities

(see

thesidebar“AI-PoweredSustainabilityataLargeOiland

GasCompany”)

,andothers.

ReapingtheBenefits

Togainthesebenefits,companyleadersmustmakeita

topprioritytotargetareaswithhighcarbonemissionsandsignificantcosts—especiallythosewithapotentialpay-

backperiodoflessthan24months.EventhepracticeofAIuseslargeamountsofenergy,andcompaniesshould

subjectitsemissions,too,toanalysis.

(Seethesidebar

“MitigatingAI’sCarbonFootprint.”)

Werecommendthatcompaniesadoptathree-prongedapproach:

?Aimhigh.UseAItoimprovethevisibilityofcarbon

emissionsacrossthevaluechain.Thendeterminewheretoapplythetechnologytoreducethefootprint,startingwiththelargestsourcesofcarbonemissionsandcosts.

?Startsmall.IndesigningyourAIapproach,useproto-typesandpilots,tocreateastrongbasisforfurther

learninganddevelopment.Usetheminimumviable

product(MVP)concepttodesignaworkableAIsystem,andtheniterateonit,integratingfeedbacktomakeitbetter.

?Scalefast.ScaleuptheMVPsolutionandtransform

theorganizationaroundittoincreaseitsimpact.InvestinbuildingcorecapabilitiesandenablersinparallelwithscalinguptheMVP.Theseshouldfocusondeveloping

enablingtechplatformsatscale,definingnewwaysofworking,andimplementingtheorganizationandgover-nancemodelsneededtoalignAIandoverallstrategy.

BOSTONCONSULTINGGROUP3

ASteelmakerCutsEmissionsandCostswithAI

Recently,aglobalsteelproducerwantedtooptimizeits

productionprocessestoreducecarbonemissionsand

lowercosts.Inthecourseofjustsixmonths,weimple-

mentedAI-basedprocesscontrolsdesignedtoeliminate

wasteandreduceenergyintensity.Thousandsofsensorscollectbillionsofdatapoints,whicharethenfedintothecontrolsystem’salgorithms.Thisenablesthecompanytopreciselycalculateandpredictenergyneedsandtrackandreducesourcesofwaste.

Sinceimplementingthesecontrols,thecompanyhas

pursuedasubsetofinitiativesthathavealreadyachievedcarbonemissionsdecreasesof3%,representingapproxi-

mately230,000tonsofCO2peryear,alongwithcostreduc-tionsof$40million—asignificantbenefitforacompany

with$8billioninrevenue.

Asimpressiveasthesegainsare,thecompanyhadalreadymadesignificantprogressinoptimizingitsoperations

beforeadoptingthesechanges.WecalculatethatAIwouldhaveamuchgreaterimpactfortheindustryasawhole,

helpingsteelcompaniesreducetheiremissionsby5%to

10%andtheircostsby1%.Ifeverysteelcompanyachievedsuchreductions,theindustrywouldemit200millionto

400millionfewertonsofCO2eachyear.

4REDUCECARBONANDCOSTSWITHTHEPOWEROFAI

AI-PoweredSustainabilityataLargeOilandGasCompany

AmajorEuropeanoilandgascompanywasfacingproduc-tionlossesduetounexpectedproblemswithmachinery

andtodependenceoncontrolsystemsthatfolloweda

report-and-reactapproach.Tocompensateforthelosses,thecompanyhadtoincreaseproduction,leadingtohigheremissionsandincreasedcosts.

Torectifytheproblem,weredesignedthecontrolsystemtoimplementamachine-learning-basedpredict-and-act

approachandcreatedanintegratedoperatingcenterto

unifytheviewsonallplantequipment.Wealsoencour-

agedthecompanytoadoptachangemanagementstrate-gytofostertheadoptionofthenewtools.

Thenewend-to-endsystemusesanumberofmachine-

learningmodels,includingtoolsthatpredictmaintenanceproblemsandCO2emissionsforeachproductionunit.Thiscapabilityenablesplantengineerstopredicttheenergy

consumptionandemissionsofalloftheirunitsforthenextthreetofivehours—andtoisolate,analyze,andfixanyunitresponsibleforexcessemissions.

Asaresult,thecompanylowereditscarbonemissionsby1%to1.5%,representing3,500to5,500tonsofGHGsperyear,andreduceditscostsbyapproximately$5millionto$10million.Thesystemalsoachieved87%accuracyin

predictingequipmentfailuresandhadan80%successrateinforecastingemissionsanomalies.

ByscalingupthisAI-poweredtoolandusingmachine

learningtoreplicateitacrossmachinesandplants,oilandgascompaniescangainacomprehensive,real-timeviewacrossalloftheirproductionoperations.

BOSTONCONSULTINGGROUP5

MitigatingAI’sCarbonFootprint

AsvaluableatoolasAIcanbeinhelpingcompaniesac-

countforandreducetheircarbonemissions,runningAI

programscontributestotheproblem.In2019,researchersattheUniversityofMassachusettsestimatedthatthe

energyrequiredtotrainastate-of-the-artnatural-languageprocessingalgorithmtoproducehuman-liketextwould

generatemorethan270,000kilogramsofCO2emissions.

Sincethen,thepowerofstate-of-the-arttechnologyhas

becomemuchgreater:OpenAI’slatestversionofitsAI

model,calledGPT-3,ishundredsoftimesmorepowerfulthanlastyear’smodelandproducesthesameamountofGHGemissionsasacarbeingdrivenmorethan700,000kilometers.

Thelesson:exponentialincreasesinAI’salgorithmiccom-plexityandthegrowingadoptionofAIsolutionsthroughouttheeconomyandsocietyrequireAIpractitionerstofind

waystomitigatethetechnology’senvironmentalfootprint.

That’sthegoalof

CodeCarbon

,acombinedeffortofBCGGamma;Mila,aworld-leadingAIresearchinstitutein

MontrealfoundedbyTuringAwardrecipientYoshuaBen-gio;Pennsylvania’sHaverfordCollege;andComet.ml,acollaborationplatformformachinelearning.

CodeCarbonautomaticallycapturestheinformationneed-edtoestimateCO2emissionsduetocomputing,enablinguserstotrack,record,andvisualizetheemissionsreleasedinthecourseofeveryexperimentandacrossmultiple

projects.

Italsoprovidesvisibilityintotheleversthatdeveloperscanpulltoreducetheirfootprint.Thisinturnallowsthemto

designmorefrugalandmoreefficientalgorithms,positiontheirserversinlocationswhereproducingpowergeneratesthelowestlevelofemissions,andchoosethemosteffi-

cienthardwarefortheirsystems.Bypullingtheselevers,

datascientistscanreducetheemissionsreleasedintrain-ingalgorithmsbyuptotentimes.AndbecauseCode-

Carbonisanopen-sourceprogram,userscancontributetoitsfurtherdevelopment.

6REDUCECARBONANDCOSTSWITHTHEPOWEROFAI

ActNow

AIhasalreadydemonstrateditsnear-termvalueinhelping

companiesreducetheirGHGemissionsandcutcosts.By

generatingapositiveROI,oftenwithinayear,itshould

quicklybecomeafinancialbenefittocompanies,rather

thanyetanothercost.WebelievethatAIcanbeespecially

valuablenow,ascompaniesrecoverfromtheCOVID-19

crisis,inloweringcostsandbeginningthetransitiontoa

low-carbonfuture.

Inthelongerterm,asthepriceofcarbonemissionsrises

andasadvancesenableAItotacklemorecomplexclimate

issues,thetechnologywillbecomeincreasinglyimportant

inmitigatingtheeffectsofglobalwarming.

Nowisthetimeforleadingcompaniestobeginreaping

thebenefitsofAI.Aimhigh,startsmall,andscalefast.

BOSTONCONSULTINGGROUP