Mark Mark ChristiniMark Christini Xiao iao Huiao Hu Combined Electromagnetic... · Mark Mark ChristiniMark Christini Xiao iao Huiao Hu ... System Design RMxprt Motor Design Q3D Parasitics

"A Combined "A Combined "A Combined "A Combined A Combined A Combined ElectromagneticElectromagnetic--CFD CFD Thermal Simulation Thermal Simulation

A Combined A Combined ElectromagneticElectromagnetic--CFD CFD Thermal Simulation Thermal Simulation for Power for Power Transformers"Transformers"for Power for Power Transformers"Transformers"

Mark Mark ChristiniChristiniMark Mark ChristiniChristiniXiao Xiao HuHuXiao Xiao HuHu

© 2010 ANSYS, Inc. All rights reserved. 1 ANSYS, Inc. Proprietary© 2010 ANSYS, Inc. All rights reserved. 1 ANSYS, Inc. Proprietary

Outline

Introduction

Magnetic Simulation using Maxwell 3D

Thermal Simulationusing Fluentusing Fluent

Summary

Future Work

© 2010 ANSYS, Inc. All rights reserved. 2 ANSYS, Inc. Proprietary

Introduction

• Power transformers are oil-filled to provide insulation as well as an effectiveprovide insulation as well as an effective cooling mechanism.

• Power transformer windings are very complicated and consist of hundreds ofcomplicated and consist of hundreds of copper conductors surrounded by pressboard insulating collars and flanges which direct the flow of oilwhich direct the flow of oil.

• Eddy losses in the copper conductors due to skin and proximity effects will vary based on their location in the windingbased on their location in the winding.

• This magnetic-thermal system creates an extremely difficult electromagnetic-CFD thermal problem


thermal problem.

Introduction Oil Filled Transformer Coolingg

OA rating - external FA/FOA rating – forced air/oil


radiators for natural cooling only

FA/FOA rating forced air/oil using fans/pumps to go through external radiators

Introduction Fluid Flow in Oil Filled Transformer

C t f t i l 3 h Oil-flow though winding


Cut-away of typical 3ph oil-filled power transformer

Oil-flow though windinghaving LV layers and HV disks

Electromechanical Design Flow

SimplorerSystem Design RMxprt

Motor Design

Q3DParasitics

PP := 6

ICA:

A

A

A

GAIN

A

A

A

GAIN

A

JPMSYNCIA

IB

IC

Torque JPMSYNCIA

IB

IC

TorqueD2D

Motor Design

PExprtANSYS CFD

Fluid Flowp t

Magnetics

Maxwell 2D/3DElectromagnetic Components

ANSYS MechanicalThermal/Stress

Model order Reduction


Co-simulation

Field Solution

Model Generation

Electromechanical Design Flow

Simulation Steps:1. Create and Solve the Eddy Current Transformer1. Create and Solve the Eddy Current Transformer

Model in Maxwell. 2. Export Geometry File from Maxwell.3. Read Geometry File and Create Mesh for Fluent.4. Map Ohmic Losses from Maxwell Using Built-in

capability in Fluent R13capability in Fluent R13. 5. Solve Fluent Model.


Transformer Example

• 1ph oil-filled power transformerMagnetic core in center• Magnetic core in center

• High voltage disk winding on outside

• LV layer winding on inside• Insulation tubes, plates and

spacers direct oil-flow• Cooling is OA (oil-to-air only)

with external radiators assumed

(Note: since radiators are not explicitly modeled, heat transfer coefficients on tank walls are


increased to represent the radiators)

Magnetic Simulation

• Maxwell 3D Eddy Current solver used

• Frequency = 60Hz sinusoidal

• Linear coreLinear core permeability = 300 with zero conductivity since laminated steel

• Skin and proximity effects considered in LV layer winding onlyy g y

• Model reduced to ¼ symmetry

• Insulation pieces


Insulation pieces included in order to export directly to Fluent

Transformer Geometry and Magnetic Meshg

• complicated insulation structure

winding tubes

• mesh = 435,000 tets

main HV-LV

ins lationinsulation


Core, clamp, and tie plate mesh

Winding and insulation mesh

disk key spacers layer vertical

spacers

Maxwell Statistics

• RAM Usage : 8GB

• Solve Time : 1.5hour using 8 processorsp

• Number of elements :elements : 435,000 tets


Magnetic Field Plot

• Indicates high ti fi ldmagnetic field

between HV and LV windings forwindings for loaded case

• Will cause significantsignificant proximity losses in LV layer windinglayer winding


Current Density Plot

• Indicates high current density atcurrent density at ends of LV foil windings where fringing flux cuts g gthe layers

• HV disks solved asHV disks solved as stranded so uniform loss densityy


Loss Density Plot in Windings

• Spatial loss density (W/m3) is directly(W/m ) is directly imported into Fluent


Summary of Losses

• Spatial loss imported directly from Maxwell– HV winding = 6,252 W– LV winding = 26,438 W

Top clamp

Right tieplate

• Additional loss assigned in Fluent

Right tieplate

Core

HV indingassigned in Fluent– Core loss = 10,000 W– Top clamp = 1,000 W

Bottom clamp = 1 000 W

HV winding

LV winding

– Bottom clamp = 1,000 W– Left tieplate = 500 W– Right tieplate = 1,000 W

Left tieplate

Bottom Clamp


Export Winding Loss from Maxwell

• Spatial loss exported to link file using menu inside of Maxwell 3D


Import Winding Loss into Fluent

Ohmic loss distribution in Maxwell Ohmic loss distribution in Fluent

Total amount of loss


Total amount of loss matches exactly that from Maxwell (32690 w)

Transformer Geometry and CFD Mesh

Copper mesh Insulator mesh

• Cell count : 12 M• Element : Hex in

most regions


most regions• Conformal fluid solid

mesh Fluid meshSteel core mesh

Fluent Statistics

• RAM Usage : A rule of thumb is that 1G memory is needed for 1M computational fluid cells running single precision. The case has 5 5M computational fluid cells and 6 5M computational solid cells5.5M computational fluid cells and 6.5M computational solid cells. And it needs approximately 8G memory running single precision. Running double precision, which is typical, would need 16G memory.

• Solve Time : on 16 CPUs, it takes approximately 24 hours to finish 3000 iterations. Depending on the setup, such a case takes a few th d t t th d it ti tthousand to ten thousand iterations to converge.

• Number of Unknowns : It solves seven equations for each computational fluid cell and it solves one equation for eachcomputational fluid cell, and it solves one equation for each computational solid cell. Thus, the total number of unknowns is approximately 45M.


LV Layer WindingVelocity Vector DistributionyFluid flow at top of layer windings directed by insulation layers and top winding plate


Temperature Distribution

Temperature distribution shows absolute temperature (in Kelvin)


Summary

• Using Maxwell 3D and Fluent, an oil-filled power transformer was simulated to determine flow pattern and temperature distribution inside the transformer.

• This information can help designers visualize the flow and• This information can help designers visualize the flow and temperature inside the transformer and thus help design better cooling system for the transformer.


Future Work

• Increase complexity of winding conductors/stranding so that AC losses in HV disk winding can be considered

• Increase complexity of insulation structure by adding baffles and an outer winding tube on HV disk winding tobaffles and an outer winding tube on HV disk winding to observe zig-zag oil flow through windings


Appendix : Fluent Setup

• Solver : steady state• Viscous model : laminar (can use turbulence model based on Ra

d P b )and Pr number)• Natural convection model : Boussinesq model• Boundary conditions : symmetry and wall with heat transfer

coefficientcoefficient• Pressure velocity coupling scheme : SIMPLE• Pressure spatial discretization : PRESTO!• URFs : 0 1 for momentum and 0 2 for pressure• URFs : 0.1 for momentum and 0.2 for pressure

Fluent Tutorial 5, Modeling Radiation and Natural Convection, gives all the details on how to set up a natural convection problem in Fluent.

The test case is also available upon request.


The test case is also available upon request.

Documents

Mark Mark ChristiniMark Christini Xiao iao Huiao Hu Combined Electromagnetic... · Mark Mark ChristiniMark Christini Xiao iao Huiao Hu ... System Design RMxprt Motor Design Q3D Parasitics