Hitachi

Energy基于TwinAI及optiSLang的干式變壓器溫升預測模型優(yōu)化JiamingZHANGSep.-2025Internal?Hitachi

EnergyLtd2025.

Allrightsreservedu

ResearchObjectsu

Softwaredescriptionu

Preliminary

results

HighVoltageWindings

LowVoltageWindingsu

Conclusions2

Internal?Hitachi

EnergyLtd2025.

AllrightsreservedContentsHITACHlu

ResearchObjectsu

Softwaredescriptionu

Preliminary

results

HighVoltageWindings

LowVoltageWindingsu

Conclusions3

Internal?Hitachi

EnergyLtd2025.

AllrightsreservedContentsHITACHlElectricaldesignElectromagnetic

loss

Safetyconcern

Lifeofthe

productTemperature

riseResearchObjects4

Internal?Hitachi

EnergyLtd2025.

AllrightsreservedDesign

margin

HITACHlHotair

outCoolair

inHeat

upEfficiency

Accuracyl

Higherefficiency

Costsavingofthe

manhoursl

More

accurate

results

Less

design

marginICostsaving

ofthe

materialElectromagnetic

lossTemperature

riseResearchObjects5

Internal?Hitachi

EnergyLtd2025.

AllrightsreservedThermalcalculationHITACHlBack

upHotair

outCoolair

inHeat

up

Objects:

HVWindings:

Temperature

rise

LVWindings:

Temperature

rise

BoundaryConditions:

Power

Rating:

200~3800

kVA

Electrical

Height:

600~1500

mm

Enclosure

Type/Cooling

method:

IP00/AN

WindingTechnology:

Cast(HV),

OW(LV)

Construction

Type:

Disk(HV),

Foil(LV)ResearchObjects6

Internal?Hitachi

EnergyLtd2025.

AllrightsreservedHITACHlu

ResearchObjectsu

Softwaredescriptionu

Preliminary

results

HighVoltageWindings

LowVoltageWindingsu

Conclusions7

Internal?Hitachi

EnergyLtd2025.

AllrightsreservedContentsHITACHl

Target:

Based

onthesimulation

results,the

metamodels

aregeneratedfocusingon

the

prediction

of

some

thermal

behaviors.

Inthis

project,thetemperature

rise

are

researched

at

HV

and

LVwindingsofthetransformers.

Tool:

optiSLang

is

usedtoanalyzethedatabaseandtrainthe

metamodel.

TwinAI

is

usedtooptimizethetool

in

higheraccuracy.

optiSLang

isanANSYStoolcontributingonthe

sensitivity

analysis

andthe

guidance

of

design

work.SensitivityAnalyzation

DesignOptimizationSoftwaredescription8

Internal?Hitachi

EnergyLtd2025.

AllrightsreservedSensitivityAnalyze

QualificationOptimizationDoERandom

Sampling

CertaintySamplingMetamodel

PredictionHITACHlSoftwaredescription_optiSLangHmACHDoERandomSampling

CertaintySamplingSensitivityAnalyze

QualificationMetamodelPrediction9

Internal?Hitachi

EnergyLtd2025.

Allrightsreserved10

Internal?Hitachi

EnergyLtd2025.

AllrightsreservedSoftwaredescription_TwinAIHITACHlu

ResearchObjectsu

Softwaredescriptionu

Preliminary

results

HighVoltageWindings

LowVoltageWindingsu

Conclusions11

Internal?Hitachi

EnergyLtd2025.

AllrightsreservedContentsHITACHlResults-

HVWindings_optiSLang

results

HmACHi12

Internal?Hitachi

EnergyLtd2025.

AllrightsreservedTwinAI

optimization_HVwinding

HACHi13

Internal?Hitachi

EnergyLtd2025.

AllrightsreservedOptimizedresultoptiSLangresultTest

resultTest

resultVsVsResults-

HVWindingsMetamodelvsTests2.83

K2.38

K0.25

KHITACHlOptimizedvsTestsMetamodelvsTestsOptimizedvsTestsInternal?Hitachi

EnergyLtd2025.

AllrightsreservedCFDvsTests14Results-

LVWindings_optiSLang

resultsHmACHi15

Internal?Hitachi

EnergyLtd2025.

AllrightsreservedTwinAI

optimization_LVwinding

H

ACHi16

Internal?Hitachi

EnergyLtd2025.

AllrightsreservedOptimizedresultoptiSLangresultTest

resultTest

resultVsVsResults-

LVWindingsMetamodelvsTestsHITACHlOptimizedvsTestsMetamodelvsTests0.25

K

0.68

KOptimizedvsTestsCFDvsTests?Hitachi

EnergyLtd2025.

Allrightsreserved0.15

KInternal117u

ResearchObjectsu

Softwaredescriptionu

Preliminary

results

HighVoltageWindings

LowVoltageWindingsu

Conclusions18

Internal?Hitachi

EnergyLtd2025.

AllrightsreservedContentsHITACHlItemsMetamodelvsCFDMetamodelvsTestCFDvstestHVWindingsTemperatureRiseAverage

Difference5×10-112.492.83LVWindingsTemperature

RiseAverage

Difference0.010.970.15ItemsOptimizedvsTestMetamodelvsTe