Title: NREL View of LongRange R
1The Role of Computational Science in Energy
Efficiency and Renewable Energy Research Steve
Hammond, DirectorComputational Sciences
CenterNational Renewable Energy
LaboratoryPACT02September 2002
2Outline
- Why NREL?
- Energy Scene
- NREL mission and programs
- Computational Sciences at NREL
- Architecture and Programming Model Observations
3Why NREL?
4The Energy Scene
- Viewed over decades, changes to energy markets
are sweeping and fundamental including impacts to
economy, environment, and national security. - Just after WW II, homes were heated with coal,
traveled on steamships, and leaded fuel. - Today, heat with natural gas, travel on jet
planes, use unleaded fuel. - In 1950, America imported 5 of its oil.
- Today, U.S. imports 62 of its fossil fuel.
5Dependence on Foreign OilU.S. Transportation use
of Petroleum
30
Domestic Petroleum Usage
Today
25
Imports
20
15
Millions of Barrels per Day
10
5
0
1970
1980
1990
2000
2010
2020
Actual Annual Energy Review 2000 Tbls 1.2, 5.1
and 5.12 Forecast Annual Energy Outlook 2002
Tbls 7 and 11 Split between Autos and Lt Truck
Transportation Energy Data Book Edition 21 Tbl
2.6 Updated April 2002
6Affluence Requires Energy Poverty Breeds
Global Insecurity
Affluence
Japan
United States
France
United Kingdom
South Korea
Mexico
Poland
El Salvador
Russia
China
Poverty
Bangladesh
Burkina Faso
Source Energy Information Administration,
International Energy Annual 2000 Tables E1, B1,
B2 Gross Domestic Product per capita is for
2000 in 1995 dollars. Updated May 2002
7Renewable Energy factoids
- North Dakota has enough wind to supply 35 of the
U.S. electrical demand. - The sunlight falling on the U.S. in one day
contains more than twice the energy we consume in
an entire year. - Renewable energy is plentiful.
- Continued RD is necessary to ensure that
renewables are efficient, reliable, and
affordable.
8National Renewable Energy Laboratory
- DoE Lab, located in Golden, CO
- Only national laboratory dedicated to renewable
energy and energy efficiency RD - Research spans spectrum of fundamental science to
technology solutions and assessment - Collaboration with industry and university
partners is a hallmark - About 1000 employees, and approx. 200M/year
9NREL Mission Statement
NREL develops renewable energy and energy
efficiency technologies and practices, advances
related science and engineering, and transfers
knowledge and innovations to address the nation's
energy and environmental goals.
10Major NREL Technology Thrusts
Supply Side Wind Solar - Photovoltaics -
Concentrating Solar Power - Solar Buildings
Biomass - Biopower - Biofuels -
Biorefinery Geothermal Hydrogen
Superconductivity Distributed Energy
Demand Side FreedomCAR and
VehicleTechnologies - Hybrid Vehicles
- Fuels Utilization Building Technologies- Zero
Energy Buildings Federal Energy
Management Advanced Industrial
Technologies Crosscutting Basic Energy Science
Analytical Studies Computational Science
11Solar Energy Research Facility
National Center for Photovoltaics Facilities
Outdoor Test Area
Outdoor Test Facility
12Solar Photovoltaics
- In the last 25 years, cost of electricity from PV
has dropped from 1/kWh to about 20/kWh today. - Strong efforts, government support in Japan and
Germany - U.S. market share 40
- RD fundamental science of materials, advanced
solar cells and processes, scale-up, lower cost
Sacramento Municipal Utility District (01026)
Source Technology Opportunities to Reduce U.S.
Greenhouse Gas Emissions, October 1997
13National Wind Technology Center
- Facility for NREL scientists and the U.S. wind
energy industry to conduct research, engineering
and development of advanced wind technology - Unique facilities
- 280 acres of ideal wind research land
- 16 outdoor turbine test pads
- Structural test facilities for turbine blades and
rotors - Test facilities for wind turbine components
Wind turbine structural test facilities (04364)
14Wind Energy
- In the last 25 years, cost of wind energy has
decreased from 40/kWh to about 4/kWh. - 2010 goal is 23/kWh
- Strong European research efforts.
- RD improvements in turbine designs, structural
dynamics, lower cost
Source Technology Opportunities to Reduce
U.S.Greenhouse Gas Emissions, October 1997
Green Mountain Power Wind Plant, Vermont (05592)
15GE WindEnergy 3.6 MW Prototype Turbine in Spain
16GE WindEnergy 3.6 MW Prototype Turbine in Spain
Boeing 747-200
17Buildings Research Activities
Investigation of boundary layer effects in
transpired collectors
Analyzing air leaks in HVAC equipment using
infrared cameras
Analyzing energy performance
18Solar Buildings Program
- Combine solar energy technology with
energy-efficient construction techniques. - Help create a new generation of cost-effective
buildings that have zero net annual need for
non-renewable energy.
19NREL Bioenergy Facilities
Alternative Fuels User Facility
ThermochemicalProcess DevelopmentUnit
Field Test Laboratory Building
Bioethanol Process Development Unit
20Biomass Sources
Poplars
Switch grass
Wood chips
Sugar cane residue
Municipal solid waste
Alfalfa
21Biomass Electric
- Direct combustion 7500 MWe installed capacity
- Cofiring (wastes) demonstrations
- Biomass gasification combined cycle (energy
crops) in development - Regrowing biomass (energy crops) results in very
low or zero net CO2 emissions - RD ash chemistry and deposition, advanced gas
turbine technologies
Source Technology Opportunities to Reduce U.S.
Greenhouse Gas Emissions, October 1997
22Biomass Transportation Fuels
- Ethanol costs have plummeted
- 1980s 4/gal
- Current 1.22/gal estimate
- 2010 0.67/gal estimate
- Develop energy crops for bulk fuel
- Employ biochemical and thermochemical processing
- Displacing gasoline with ethanol in light-duty
vehicles gives 90 reduction in carbon emissions - RD low-cost production of enzymes, development
of microorganisms,energy crop advances
Source Technology Opportunities to Reduce U.S.
Greenhouse Gas Emissions, Oct 1997
23Distributed Energy Resources
- Goal
- Increase sources of energy production reduce
dependence on large critical targets. - Triple distributed power generation by 2010
- RD Needs
- Interconnection standards development and
controlled interconnection testing critical to
market development - Distribution systems research
- Distribution system behavior (modeling, testing)
- Distributed command and control systems
development
24Transportation Technologies Systems
- RD in advanced vehicle and fuel technologies.
- Many small efficiencies add up quickly
- Smart tires
- Reduced auxiliary loads
- Strong partnerships with public and private
organizations. - Goal reduce dependence on imported oil and
improve air quality.
25National A/C Fuel Use
- 7.1 billion gallons used for air conditioning
annually. - Equivalent to 10 of U.S. imported crude oil.
26Computational Sciences Challenge
- The application area is compelling and is of
increased importance since Sept. 11 - Lessened dependence on foreign energy sources
improves homeland security. - Distributed energy reduces risk.
- Global affluence and stability requires Energy.
- NREL has a tremendous, dedicated staff.
- Very strong theory and experimental focus.
- People ask, Why fix it if isnt broken?
27Computational Science
Theory is difficult to validate and building
prototypes slow and expensive. Numerical
simulation enables the study of complex systems
and natural phenomena that would be too expensive
or dangerous, or even impossible, to study by
direct experimentation. Computation is an equal
and indispensable partner, along with theory and
experiment, in the advance of science and
engineering.
28Vision
- Create a world-class computational science center
that enable the best research in energy
efficiency and renewable energy research. - Augment and substantially enhance core
laboratory-wide HPC research, capability, and
competency in support of NRELs mission.
29Goal for 2005
- Establish 1st class research group
- Computational scientists for simulation, data
mgmt, analysis, visualization, and collaborative
problem solving environments. - Integrate CS into NREL mission and
national/international renewable energy research
community. - Establish multidisciplinary Expedition Teams to
solve most pressing energy efficiency and
renewable energy problems. - Establish links with other DoE Labs, C.S.
programs. - Establish summer student, post doc, and visitor
program.
30Three CS focus areas at NREL
- Simulation Numerical methods algorithms
- Computational fluid dynamics
- Computational chemistry
- Structural mechanics
- Data Currency of the scientific process
- Analysis and visualization
- Data mining
- Integration Holistic view
- Tools, Systems, and Enterprises
- Collaborative Problem Solving Environments
31Wind Turbines
- 70m blade challenges
- No wind tunnels large enough
- Shifting winds cause rotation and blades cut back
into own wake - Structural analysis
- Aero-acoustics
- Simulate turbine ensemble
- Incorporate accurate PBL information into
simulations
32Materials Research
- Internationally recognized Solid State Theory
Group - Study materials and alloys for more efficient
photovoltaic systems and solid state lighting. - Nanostructures given desired properties, what
is the best material?
33Molecular Mechanical Modeling of Cellulases
H2O
H2O
H2O
enzyme
H2O
Reaction Box
H2O
cellulose
- Model chemical systems using empirical force
fields to describe how molecular potential
energies change with atomic coordinates. - Forces in a molecule can be computed from
potential energies. With the atomic forces, the
motions of the molecule can be computed by
solving Newton's Equations of Motion for all of
the atoms in the molecule.
34CFD - HVAC Systems
- Simulation predicts the distribution of air
velocity and temperature using building geometry,
vent locations, and fan performance curves. - Avoid over-engineered system.
- Plumes of hot gas and smoke from fires can be
predicted in order to assist in safety analysis - Additional applications to biohazard or chemical
attacks.
35Transportation Efficiency
- Passenger comfort is the focus of vehicle climate
control. A computational model of the interior of
the vehicle yields comfort indicators that allow
the design to be improved. - Study thermal loading.
- Optimized climate control leads to smaller air
conditioners and heating systems greater fuel
economy.
36Digital Functional Vehicle Wheel
CAD Tools
Mechanica Ansys Abaqus
ADVISOR The Math Works Saber
Packaging/Visualization
StructuralAnalysis
Vehicle System Level Analysis
partner info
(current,concepts)
ICEM-CFD Fluent, STAR/CD
New, Better Components
Component Data
CFD
Data Integration
1 - Energy Focus
Digital Functional Vehicle
energy interactions data
Users
Vehicle Dynamics
ADAMS ADAMS/CAR
integrated systems data
2 - Emissions
Cleaner, More Efficient Vehicles
Data Integration
VehicleData
Materials
MVision Matls
partner info
Lab Partners
Market Data
(current,concepts)
Optimization
Motor Wiz Crash Analysis/Dyna
Customers Research
Genesis Visual Doc BMX, DOE
37Strategy
- Visit the diverse research groups to assess
short-term and long-term needs. Set priorities. - Cannot go it alone
- Look for collaboration opportunities with
university community. - Develop close ties to other DoE CS programs.
- Seek appropriate funding opportunities.
- Initiate student, post doc, visitor program.
- Integrate NRELs research agenda and CS
Initiative into the national CS infrastructure.
38Initial Assessment
- Tremendous opportunity to make a difference.
- Most extant modeling and analysis occurs with
greatly simplified models. - Most work currently done on PCs and workstations
- Limited use of machines at other DoE labs.
- Solid State Theory, Wind, Molecular Mechanical
Modeling, and Transportation Systems biggest
initial HPC users.
39First steps
- Started March 4th.
- IBM SP starter system arrived March 28th.
- Hired three staff members.
- Planning office space, fall occupancy, 9 staff.
- Identify key early adopters and set milestones
for early success stories. - Work with other Research Center staff to develop
CS strategic plan and implementation. - Long term - planning new building for FY09
occupancy.
40NRELs Applications
- Complex geometries lead to unstructured grids.
- Heterogeneous computations.
- Often grids are adaptively refined (dynamic)
leading to load balancing issues. - Global communications required.
41Scientific Computing Observations
- The high performance computing market is a niche
market at best. - With the exception of a few Grand Challenge
applications and publicity stunts, vast
majority of scientific computing now done on 128
PEs or less. - Limited scaling in current application codes.
- Better turn-around times - queue wait times
excessive for large PE counts. - Linear charging schemes.
42The Case for Shared Memory
- Relatively easy to build shared memory systems
with 128 PEs - Whats so good about shared memory systems?
- Much simpler programming model.
- Easy to start parallelizing.
- Natural load balancing threads.
- Dynamic load balancing nearly impossible with
MPI. - Tools, tools, tools,
43Shared Memory Systems
- Need improved hardware and software support for
shared memory programming. - Low thread overhead (spawn, synchronization, )
- User control of memory placement (critical in
distributed shared memory, NUMA, systems) - Application development environment in shared
memory systems far superior to distributed memory
systems. - Sun and SGI continued viability???
- Multiple processors on a chip could help.
44Large Systems
- Suppose that we have efficient shared memory
systems. - Cant ignore the grand challenge problems.
- Chunky clusters - build large systems from
clusters of shared memory systems. - While at NCAR we studied use of large systems for
climate modeling.
45Message Passing Analysis
- Suppose that you could sustain 16 Gflops on a
single shared memory system. - A sustained Teraflop system could be assembled
from a cluster of 128 of these nodes. - Used 2D decomposition of NCARs CCM3.2 (MPI).
- For simplicity, we assumed computation and
communication are both perfectly load balanced. - Counted number of bytes and number of messages
per day of simulated climate (2 day - 1 day) per
node.
46High Resolution Climate Model Message Passing
Statistics
NCARs CCM3 T170L18 4x32 decomposition, 128
node system.
47Interconnect Needs Thin and Fat Pipes
- Low Latency Thin Pipes
- Control, Status,
- Global Reductions
- High Bandwidth Fat Pipes
- Bulk data transfer
- Data transpositions
- Real world applications need both.
- CM5 had the fat tree and the control network.
48Summary
- Energy is crucial for prosperity and security.
- NREL is Americas premier laboratory for energy
efficiency and renewable energy research. - NREL is establishing a world class Computational
Science capability in support of and commensurate
with its other research efforts. - Improvements in shared memory systems would be
beneficial to scientific computing community.
49(No Transcript)