multi-physics integration framework mupif – design

Multi-Physics Integration Framework MuPIF – design, operation and optimization

of a photovoltaic cell

B. Patzák*, V. Šmilauer*, M. Appelª, R. Altenfeldª, L. Thielen†, A. Lankhorst†

*Czech Technical University in Prague, ªACCESS Materials&Processes, †CelSian Glass & Solar B.V.

www.mmp-project.eumech.fsv.cvut.cz/mupif

www.mmp-project.eu

MuPIF platform: design, operation and optimization of a photovoltaic cell 2nd ICMEg Workshop,Barcelona

Outline● Introducton to MuPIF platform

– History– Key features

● Case study: Selenization (CFD+thermodynamics)

– Macro-scale (furnace) models– RVE miscro-structure evolution

● Conclusions

www.mmp-project.eu


MuPIF platform● Initially developed under the national project funded by Czech Grant

Agency (Project No. P105/10/1402.), 2012-2014.● Currently developed in FP7 project - Multiscale Modelling Platform:

Smart design of nano-enabled products in green technologies, 2014-2016, 8 partners from 5 different countries, 2 industrial end-users.

● Industrial application for photovoltaic devices and LEDs ● Selenization (CFD+thermodynamics)● Light conversion (opto+thermal)

www.mmp-project.eu


MuPIF platform: Key features• Focus on combining existing tools to build a customized multi-physics and multi-scale simulation chains

• provides an underlying infrastructure enabling high-level data exchange and steering of individual applications.

• From discovery pf problem domain individual objects identified and grouped.

For each group, a base class introduced defining abstract interface (set of methods) • Interfaces allow to communicate with any object using generic interface, hiding the particular implementation details

• Interfaces allow for a plug-and-play architecture

Models

Data

Models

Properties

Models

Spatial Fields

Models

Discretizations

Time steps

Units

Microstructures

Courtesy: G. Schmitz

www.mmp-project.eu


Models (applications)

• Interfaces allow for a plug&play architecture

• Initial/boundary conditions as well as effective properties from lower scales imposed via set methods

• Internal state of the model updated via solveYourself, finishStep, etc. methods

• Effective properties for upper scales extracted using get methods

• Different ways how to implement interface – direct approach for apps with API or source code

– indirect approach (file exchange) for closed source apps

MuPIF Platform – API design pattern

time/process

scale

setboundaryconditions

provide effectivevalues

provideboundaryconditions

seteffectivevalues

Initialconds

output

www.mmp-project.eu


Data components

• Platform abstracts not only applications, but also high level data (Properties, Fields, Discretizations, Units, etc.)

• These entities represented by objects encapsulating raw data, metadata, and operations on the data

• Individual applications do not have to interpret raw data, they just use the interface

• Abstracts different data formats (hidden behind the interface)

• Redundancy avoided (operations on data are part of data itself)

• Advantage: the models receive (and provide) data in a self consistent package: raw data with operations on the data

MuPIF Platform – API design pattern

Model #1 Model #2

Model #3

Models/applications with generic API

High level data with API

Component-based simulation scenario

www.mmp-project.eu


MuPIF platform: Key features

• Based on interacting, distributed objects

• Transparent, distributed object system fully integrated using so-called proxies; SSL based encryption for security

• Integrates with existing grid middleware (Condor) or provides own resource manager

• Support new business models, such as software as a service

Internet

App1 App2

App1

App2

App2 Proxy

Name server

Pyro

Local scenario

Distributed scenario

www.mmp-project.eu


CIGS for solar cells: selenisation of a Cu-Ga-In precursor filmSEM picture of the CIGS microstructure (by Abengoa)furnace (~500°C) with Se containing

atmosphere

glas

Cu-In-Ga precursor filmMo film

Se gas

~ 10 cm

~ 2 µm

glas

crystalline CIGS (photoactive)Mo film

www.mmp-project.eu


Simulation model portfolio

multi-physics CFD code X-Stream (thermal fields & Se gas flow)

CALPHAD based thermodynamics (driving force & phase composition)

Phase–field software MICRESS (phase transformation &

microstructure)

T(x), [Se](x)

[Se](x), =f(CIGS)

www.mmp-project.eu


www.mmp-project.eu


t0 tend

texchange

CFD

RVE PF

T,

[Se]

Δ[S

e],

ε

T,

[Se]

Δ[S

e],

ε

CFD Simulation results

www.mmp-project.eu


Phase-field simulation results

T, [Se]

boundary conditions

for RVE

update everytexchange

[Se]

Se consumption

update everytexchange

Prototypic simulation of CIGS growth(no coupling to thermodynamics)

CIGS formation vs. time

Time [s]

CIG

S v

olu

me [

a.u

.]

RVE at location „mid“

www.mmp-project.eu


Functional for process optimisation (quality factor)

RVE1

RVE2

Quality factor map as a functional for process optimisationQF interpolation via Mupif

QF1

QF2

Quality factor defined as a product of individual QF (Cu/(Ga+In) ratio, rel. concentration of binary phases, thickness of MoSe layer, etc)

www.mmp-project.eu


Conclusions

• MuPIF enables to link different simulation models operating on different scales and distributed resources

• MuPIF provides services for data high level data exchange, interpolation / extrapolation, and I/O services

• The capabilities of MuPIF demonstrated on a simulation chain for CIGS thin film processing

• The developed simulation chain allows to asses the quality of the process (QF) and subsequent optimization of the process.

multi-physics integration framework mupif – design

Documents