NREL+用于能源系統(tǒng)研究的大型語(yǔ)言模型-英_第1頁(yè)
NREL+用于能源系統(tǒng)研究的大型語(yǔ)言模型-英_第2頁(yè)
NREL+用于能源系統(tǒng)研究的大型語(yǔ)言模型-英_第3頁(yè)
NREL+用于能源系統(tǒng)研究的大型語(yǔ)言模型-英_第4頁(yè)
NREL+用于能源系統(tǒng)研究的大型語(yǔ)言模型-英_第5頁(yè)
已閱讀5頁(yè),還剩12頁(yè)未讀, 繼續(xù)免費(fèi)閱讀

下載本文檔

版權(quán)說(shuō)明:本文檔由用戶(hù)提供并上傳,收益歸屬內(nèi)容提供方,若內(nèi)容存在侵權(quán),請(qǐng)進(jìn)行舉報(bào)或認(rèn)領(lǐng)

文檔簡(jiǎn)介

Large

LanguageModels

(LLMs)for

EnergySystems

ResearchGrant

BusterWorkshop

onSolarApplicationsofArtificialIntelligence

andMachineLearningNovember

1st,

2023NRELProducesaLOT

ofText!Whatqualifies

aseffective

disseminationofresearch?NREL|

2LLMsfor

the

Dissemination

ofResearchQ:Tell

meabout

technical

progress

inthePuertoRico100%

study(PR100)?A:ThePR100study'sfirst-yearprogress

includes

initialmodeling

and

analysisresults,

as

wellas

thedevelopmentofhigh-resolution

wind

and

solar

resourcedatasetsforPuertoRico.Amajor

accomplishmentto

date

is

thedevelopmentoffeasiblescenarios

forPuertoRicotoreachitsgoalsof

100%

renewableenergyby

2050,

withinterimtargets

of40%

by

2025,

60%

by

2040,

thephaseout

ofcoal-fired

generation

by

2028,

anda30%

improvement

inenergyefficiencyby2040,

as

codified

inthePuertoRicoEnergyPublic

PolicyActof2019

(Act17).Themodel

wasprovided

withthefollowingdocuments

to

support

itsanswer:?

PR100One-Year

Progress

Summary

ReportChatGPT:

As

ofmy

lastupdate

inSeptember

2021,

therewas

nospecificstudyorprojecttitled"PuertoRico

100%

study(PR100)"

thatI'maware

of.NREL|

3LLMsfor

the

Dissemination

ofResearchQ:Whatwouldhappen

if

thepopulationof

PuertoRicowere

tosuddenlydouble?A:Idonotknowtheexact

consequences

of

thepopulation

of

PuertoRicosuddenly

doubling,

as

theprovidedarticles

do

not

discuss

such

ascenario..ChatGPT:

A

sudden

doublingofPuertoRico's

populationwouldleadto

awiderangeofsocial,

economic,environmental,andinfrastructural

challengesandchanges.Here'sabreakdown

ofsome

potentialimpacts…NREL|

4PotentialApplications…

Lots!NREL|

5Sohowdoesthis

work?Chunked

TextVectorEmbeddings{0.19,

0.17,

…,0.98}VectorDB{1.23,

4.11,

…,6.01}NREL|

6Sohowdoesthis

work?GPT-4UserQuery:Engineered

Query:Tell

meaboutlandavailability

inthePuertoRico100%study

(PR100)?Query

Vector:{1.65,

4.74,

…,6.23}Using

theinformationbelow,

answerthis

userquestion…SearchResponse:VectorDBBasedontheinitialmodeling

of

renewableenergypotential

andavailable

land

areainPuertoRico…Most

SimilarVector:{1.23,

4.11,

…,6.01}NREL|

7Windand

SolarSiting

Ordinances?

High-impact,

high-effort–

1,500

internhoursto

parselegaldocuments?

Siting

ordinancesfor

renewablesisarapidlychanging

legallandscape–

Nationalwind

ordinancesincreasedby

morethan540%from

2018

to

2022Lopez,A.,Cole,W.,Sergi,B.etal.”Impactofsitingordinances

onlandavailabilityforwind

andsolardevelopment”.

Nat

Energy8,1034–1043

(2023).

/10.1038/s41560-023-01319-3?

>3k

counties

intheU.S.–

Some

states

have

ordinancespertownshipand

municipality(>16k

intheU.S.)NREL|

8LLMsfor

Parsing

Siting

Ordinances?

21counties,

5windsetback

features,105totalordinances–

Structures–

Propertylines–

Roads–

Railroads–

Transmissionlines?

InitialResults:–

Ordinancedatabase:82%accurate*–

LLMwithsingleprompt:

86%accurate–

LLMwithdecision

tree:

90%accurate*someordinance

database

errors

may

bedueto

legal

amendments

made

afterApril

2022NREL|

9LLMsGuided

byDecision

Tree?

Decision

tree

structure

can

improveLLMLLMLLMAtheaccuracy

ofLLMreasoning–

Nodesandedges

have

promptsandfunctions

that

define

statesandtransitions0BC?

Takes

advantage

ofslow-logicsymbolic

reasoning

andfast-logictext

comprehension

fromGPT-412?

Rudimentary

implementation

ofaLLMneuro-symbolic

AIsystemDNREL|

10ComplexRules:Monroe

County,

WisconsinSetbackordinanceslastupdatedin

2019

(reference)NREL|

11ComplexRules:Monroe

County,

Wisconsin?

Setback

from

structures:–

Ordinancedatabase:1.1x

maxtipheight–

Single-query

LLM:?

“The

setback

for

large

windenergysystems

fromoccupiedcommunity

buildings

andnonparticipating

residences

is

thelesser

of1,250feet

or3.1

times

themaximum

blade

tip

height.For

participating

residences,

nonparticipating

propertylines,public

road

right-of-way,

andoverhead

communication

andelectric

transmission

ordistribution

lines,

thesetback

is

1.1timesthemaximum

blade

tipheight.”Setbackordinanceslastupdatedin

2019

(reference)NREL|

12ComplexRules:Monroe

County,

WisconsinLLM

with

decision

tree:?USER:Isthere

text

inthe

following

legal

document

that

describes

howclose

icansite

alarge

windenergy

system

tobuildings

andstructures?

Ignoretext

onsmall,

mini,

orprivate

wind

energy

systems.Ignore

text

regarding

setbacks

fromparcels,

rights-of-way,

andutility

easements.

Please

only

say

"Yes"

or"No".??ASSISTANT:

YesUSER:Based

onyour

last

message,

choosethe

option

that

best

describes

the

required

setback

frombuildings

and

structures:#1.1Thesetback

frombuildings

andstructures

isasingle

distance

value#7.1Thesetback

frombuildings

andstructures

hasmultiple

conditions

suchasafixed

distance

and

amultiple

ofthetotal

system

height#0Noneofthe

above

options

are

descriptive

of

this

ordinance

related

to

setbacks

frombuildings

andstructures?ASSISTANT:

#7.1Thesetback

frombuildings

and

structures

hasmultiple

conditions

suchasafixeddistance

andamultiple

ofthe

total

system

heightNREL|

13ComplexRules:Monroe

County,

WisconsinLLM

with

decision

tree:?USER:Letsassume

the

supporting

tower

is377feet

with

blades

279

feet

long.

So

the

rotordiameterwould

be558

feet

andthe

total

system

height

wouldbe

656feet.

What

wouldthe

multiple

setbackdistances

be?Please

showyour

work.?ASSISTANT:

Based

onthe

information

provided,

wecancalculate

the

setback

distances

for

the

differenttypes

ofbuildings

andstructures:1.Occupied

community

buildings:

Thelesser

of1,250

feet

or3.1times

the

maximum

blade

tipheight

3.1*656feet

=2,033.6

feet.

The

lesser

value

is1,250

feet.2.Participating

residences:

1.1times

the

maximumblade

tip

height

1.1*656

feet

=721.6feet3.Nonparticipating

residences:

Thelesser

of

1,250

feet

or3.1times

the

maximum

blade

tipheight

3.1*

656feet

=2,033.6

feet.

Thelesser

value

is1,250

feet.So,the

multiple

setback

distances

are:1,250

feet

for

occupiedcommunity

buildings721.6feet

for

participating

residences1,250

feet

for

nonparticipating

residencesNREL|

14ComplexRules:Monroe

County,

WisconsinLLM

with

decision

tree:??USER:Whatisthefinal

setbackvalue?Iftheordinancestates

whichofthemultipledistancesshouldbechosen,usethatguidance.

Otherwise,

choosethelargestsetbackvalue.ASSISTANT:Theordinancedoesnotprovide

specific

guidance

onchoosingbetweenthemultiple

distances.Therefore,

we

willchoose

thelargestsetbackvalue.Thefinalsetbackvalueis1,250

feet.??USER:State

thefinal

setbacklike

this:"ThesetbackisXXX(units)"ASSISTANT:Thesetbackis1,250feet.NREL|

15Powered

byOpen-Source

Software?

ELM:Energy

LanguageModel–

Utilities

to

workwith

theOpenAIAPI–

Text

embeddingsearch–

Decisiontree

framework–

PDF-to-textutilities–

Recursive

summarizationandtextprocessing–

AsynchronousAPI

callsNREL|

溫馨提示

  • 1. 本站所有資源如無(wú)特殊說(shuō)明,都需要本地電腦安裝OFFICE2007和PDF閱讀器。圖紙軟件為CAD,CAXA,PROE,UG,SolidWorks等.壓縮文件請(qǐng)下載最新的WinRAR軟件解壓。
  • 2. 本站的文檔不包含任何第三方提供的附件圖紙等,如果需要附件,請(qǐng)聯(lián)系上傳者。文件的所有權(quán)益歸上傳用戶(hù)所有。
  • 3. 本站RAR壓縮包中若帶圖紙,網(wǎng)頁(yè)內(nèi)容里面會(huì)有圖紙預(yù)覽,若沒(méi)有圖紙預(yù)覽就沒(méi)有圖紙。
  • 4. 未經(jīng)權(quán)益所有人同意不得將文件中的內(nèi)容挪作商業(yè)或盈利用途。
  • 5. 人人文庫(kù)網(wǎng)僅提供信息存儲(chǔ)空間,僅對(duì)用戶(hù)上傳內(nèi)容的表現(xiàn)方式做保護(hù)處理,對(duì)用戶(hù)上傳分享的文檔內(nèi)容本身不做任何修改或編輯,并不能對(duì)任何下載內(nèi)容負(fù)責(zé)。
  • 6. 下載文件中如有侵權(quán)或不適當(dāng)內(nèi)容,請(qǐng)與我們聯(lián)系,我們立即糾正。
  • 7. 本站不保證下載資源的準(zhǔn)確性、安全性和完整性, 同時(shí)也不承擔(dān)用戶(hù)因使用這些下載資源對(duì)自己和他人造成任何形式的傷害或損失。

評(píng)論

0/150

提交評(píng)論