A new approach for exploration of tokamak power plant design space

Farrokh Najmabadi, Lane Carlson, and the ARIES TeamUC San Diego

4th IAEA Technical Meeting on First Generation of Fusion Power Plants: Design & Technology 8-9 June 2011IAEA, Vienna

Outline:

The New ARIES Systems Code. Developed Recently New Features/models.

The New System Optimization Process Parameter Space Survey as opposed to single-point

optimization. Data visualization Tools; Capability of “multi-dimensional” optimization and risk-benefit

analysis. Work in progress, preliminary results.

The Current ARIES Team Project

Systems code development

ARIES Systems Code is used for trade-off analysis and NOT as a Design Tool It is impossible at the present to combine all analysis tools

of various components into one package. Systems code, therefore, includes simple and reasonably accurate model of various subsystems.

We have always used the Systems code as an iterative tool during the design process: systems code identifies the area of interest in the parameter space, detailed design of a “preliminary” design point (“strawman”) helps to identify issues and refine systems code models, systems code identifies a new “strawman,” etc.

Typically, we have used COE as the object function for systems analysis while other objectives/constraint were included in the design process.

A New ARIES Systems Code was developed recently.

The original ARIES systems code was developed in late 80s Outdated programming. Some models were too crude and had to be refined

considerably during the design study.o A more complex model was deemed to be computationally

expensive at the time.o A single code was used to examine multiple confinement

concept (without using object-oriented programming and tool kits), i.e., models had to fit a large design window.

A new system code was developed recently. Currently only applies to tokamak concept. We are using this tool for the first time in our current

study.

ASC Flow Chart

(medium detail)

• Systems code consists of many modular building blocks

Major Features of the new ARIES Systems Code

Detailed 2-D axisymmteric geometry based on plasma shape and includes maintenance considerations.

Detailed power flow and blanket/divertor modules.

TF Magnet algorithm bench-marked against finite-element analysis of ARIES magnets.

Distributed PF algorithm which produces accurate description of stored energy, cost, and VS of a free-boundary equilibrium-based PF system.

New costing algorithm Based on Gen-IV costing rules.

Detailed power flow model is essential for blanket/divertor models

He pumping power in the divertor is a strong function of heat flux capability

Analysis of three options performed, with some amount of optimization vs. heat flux (more design optimization underway).

Results reduced to polynomial fits with a very high fidelity

4 6 8 10 12 14 160

5

10

15

20

25

30

"Plate"

Finger

T-tube

Surface Heat Flux (MW/m2)

Pum

ping

pow

er/T

herm

al p

ower

(%)

Thermal conversion efficiency models

The entire effect is caused by degradation of bulk outlet temperature due to internal heat transfer: very design-dependent

0 1 2 3 4 5 6 7 80.36

0.38

0.4

0.42

0.44

0.46

0.48q''(max)=0.25 MW/m^2q''(max)=0.5 MW/m^2q''(max)=0.75 MW/m^2q''(max)=1 MW/m^2

Maximum Neutron Wall Load (MW/m2)

Gro

ss C

ycle

Effi

cien

cy

1 2 3 4 5 6 70.45

0.47

0.49

0.51

0.53

0.55

0.57

0.59

q''(max)=0.25 MW/m^2q''(max)=0.375 MW/m^2q''(max)=0.5 MW/m^2q''(max)=0.625 MW/m^2

Maximum Neutron Wall Load (MW/m2)G

ross

Cyc

le E

ffici

ency

(DCLL) (SiC)

Major Features of the new ARIES Systems Code

Detailed 2-D axisymmteric geometry based on plasma shape and includes maintenance considerations.

Detailed power flow and blanket/divertor modules.

TF Magnet algorithm bench-marked against finite-element analysis of ARIES magnets.

Distributed PF algorithm which produces accurate description of stored energy, cost, and VS of a free-boundary equilibrium-based PF system.

New costing algorithm Based on Gen-IV costing rules.

Systems Optimization

6

7

8

9

10

0 1 2 3 4 5

Avg. wall load (MW/m2)C

OE

(c/k

Wh)

Issues with “Point-Optimization”

Experience indicates that the power plant parameter space includes many local minima and the optimum region is quite “shallow.”

The systems code indentifies the “mathematical” optimum. There is a large “optimum” region when the accuracy of the systems code is taken into account, requiring “human judgment” to choose the operating point.

Previously, we used the systems code in the optimizer mode and used “human judgment” to select a few major parameters (e.g., aspect ratio, wall loading).

“Mathematical” minimum

Systems code accuracy

Optimum region

Better Approach: Indentify “optimum region ” as opposed to the optimum point.

Identifying “optimum region ” as opposed to the optimum point

New “optimization” approach: Indentify “optimum” parameters space as opposed to the

optimum point. In addition, experience indicates that “optimum” design points

are usually driven by the constraints. In some cases, a large design window is available when the constraint is “slightly” relaxed, allowing a more “robust” and credible design.

Developed a new approach to Systems Analysis based on surveying the design space instead of finding only an optimum design point (e.g., lowest COE).

This approach requires generation of a very large data base of self-consistent physics/engineering points.

Modern visualization and data mining techniques should be used to explore design space.

Now to visualize those points…

It is a challenge to visualize large data sets:

“Lots of numbers don’t make sense to ‘low-bandwidth’ humans, but visualization can decode large amounts of data to gain insight.” - San Diego Super Computer Center

How do others visualize large data sets?

Visualizing large datasets is a difficult task, almost an art.

Too much information can be overwhelming/deceiving.

106 points exceed computer monitor real estate.

NOAA (National Oceanic and Atmospheric Admin.) data-logging buoys

SDSC Earthquake Simulation

We have developed a visualization tool to utilize the scanning capability of the new systems code

VASST - Visual ARIES Systems Scanning Tool Desire to visualize the broad parameter space

and to extract meaningful relationships Graphical user interface (GUI) permits 2D plots of

any parameter

Purpose: give the user visual interaction and explorative power extract meaningful relationships understand design tradeoffs

18

Pull-down menus for common parameters

Blanket usedNumber of points in database

Constraint parameter can restrict/filter

database on-the-fly

Auto-labeling

Correlation coefficient

Save plot

Color bar scale

Edit plot properties

Populate table with click

Turn on ARIES-AT point design for reference

VASST - Visual ARIES Systems Scanning Tool

All self-consistent design points

Snap-shot of COE vs electric output for all data points (SiC blanket)

SiC blanket

Example: COE vs Inboard Divertor Heat Flux

20

SiC blanket, 1 GWe

constraint: BT < 8.5 T

Example: COE vs Inboard Divertor Heat Flux

2121

SiC blanket , 1 GWe

constraint: BT < 7.5 T

Example: COE vs Inboard Divertor Heat Flux

SiC blanket , 1 GWe

constraint: BT < 6 T

Example: COE vs Inboard Divertor Heat Flux

SiC blanket , 1 GWe

constraint: BT < 5 T

Example: Impact of field strength

For a given maximum bN: Increasing Bt typically increase the size of parameter space (e.g.,

relax many parameters: bN , nG/n, H98, …) but does not reduce the minimum cost.

Bt can be viewed as a compensation/insurance knob against R&D failure to achieve certain plasma parameters

Bt < 5 T

CO

E (m

ills/

kWh)

Heat flux on in board divertor (MW/m2)

Bt < 7 T

CO

E (m

ills/

kWh)

Heat flux on in board divertor (MW/m2)

SiC Blanket

Summary

Exploring the “optimum region” as opposed to the “optimum point” has many desirable features: Prevents the design point to be pushed into a “constraint

corner” and allows for a more robust design point. Clearly identifies the trade-offs and the “weak” and

“strong” constraints, helping the R&D prioritization. Identifies compensation/insurance knobs against failures

of R&D to achieve certain parameters. The extensive data base can also be used for

risk/benefits analyzes by defining different optimization parameters.

Project goals and plans

Goals of Current ARIES Study

Revisit ARIES Designs with a focus on detailed analysis edge plasma physics and plasma-material interaction, high heat flux components and off-normal events in a fusion power plant. What would ARIES designs look like if we use current

predictions on heat/particles fluxes? What would be the maximum fluxes that can be

handled by in-vessel components in a power plant? What level of off-normal events are acceptable in a

commercial power plant? Can the current physics predictions (ITER rules, others)

be accommodated and/or new solutions have to be found?

Frame the “parameter space for attractive power plants” by considering the “four corners” of parameter space

Reversed-shear(βN=0.04-0.06)DCLL blanket

Reversed-shear(βN=0.04-0.06)SiC blanket

1st Stability(no-wall limit)DCLL blanket

1st Stability(no-wall limit)SiC blanket

ARIES-RS/ATSSTR-2EU Model-D

ARIES-1SSTR

Lower thermal efficiencyHigher Fusion/plasma powerHigher P/RMetallic first wall/blanket

Higher thermal efficiencyLower fusion/plasma powerLower P/RComposite first wall/blanket

Higher power densityHigher densityLower current-drive power

Lower power densityLower densityHigher CD power

Decreasing P/R

Phy

sics

E

xtra

pola

tion

Study is divided into Two Phases:

Phase 1: Examination of power-plant parameter space with the

systems code to understand trade-offs. Examination of capabilities of in-vessel components

(heat and particle fluxes) in both steady-state and off-normal events, e.g., ELM, disruption (thermal).

Phase 2: Detailed design

o To verify systems code predicationso To develop an integrated designo To document trade-offs and R&D needs.

Project began in Jan. 2010.