etap - ecs presentation
TRANSCRIPT
A cloud-based modeling framework for complex advanced power systems
Dr. Mark Bryden
Simulation, Modeling, & Decision Science AMES LABORATORY
https://svs.gsfc.nasa.gov/vis/a000000/a004000/a004019/S_N_europe.0001.jpg
Energy and environmental challenges
Increasing energy use
Increasing impact on the environment
Increasing resource scarcity
Holistic solutions are needed
Interactions between
Engineered, human, and natural systems
are confounded by complexity
New modeling approaches are needed
Today we cannot model the richness, fullness, or complexity of engineered, human, or natural systems.
The Challenge
Many different models • No uniform, active storage space • Not readily accessible, citable, or maintained • Hard to locate and use existing code
Codes do not work together • Systems models use codes specifically built for them • Hard to use existing codes in a new systems model • Clunky
Systems modeling often lacks fidelity and granularity • Algebraic expression, ODEs, reduced-order models • Averaging and message passing
Modeling and the Design Process Today
Problem Definition
Conceptual Design
Preliminary Design
Detailed Design
Validation & Verification
Final Design Selection & Production
Engineering Design
Conceptual SystemModels
Preliminary System Models
Detailed “Spot” Models
S Suram and KM Bryden, Advances in Engineering Software, 90:169-182 (2015)
Proposed Design Process
Problem Definition
Conceptual Design
Preliminary Design
Detailed Design
Validation & Verification
Final Design Selection & Production
Engineering Design
Systems Modeling Analysis is preserved as a part of the product
Project Objectives
Demonstrate the framework and tools needed to create detailed systems models based on reusable, accessible models that enable policy, engineering, and operational decisions for advanced fossil energy systems.
This requires the development of • Web enabled information objects • Federated model sets • User interaction tools
Hyper
• Current - a hybrid fuel cell and gas turbine system • Goal - test the dynamic performance of any advanced power system that
includes a gas turbine cycle
Hyper: CSP - Fossil Hybrid
PDT 158
Air
Exhaust
Air Plenum V-301
Post Combustor
V-304
Load Bank E-105
F, T, P
Q!
Fuel Valve Model
Natural Gas
PT-151
C-100 T-101
TE-147
PT-116TE-112
FE-110
E-001FE-162
E-300
E-305
Generator G-102
TE-326
Fuel Cell Simulator
V-302
FE-380
ST-502
PT-305
FV-432A
FV-170
FV-162
FV-380
%
CSP Model
LHV Natural Gas
Pressure
Incident Sunlight
Thermal Energy Storage Model
TE-333
TE-350 PT-180
Hot Air Bypass
Cold Air Bypass
Bleed Air Bypass
F, T, P
Natural Gas
FV-432B
Control Algorithm
dtdP
TE-235
PT-236
Bottoming Cycle
F, T, P
Three questions
• Why can’t we integrate analysis into engineering decision making on-the-fly?
• Why isn’t engineering analysis like a game?
• Why do we continually make new models?
Federated Modeling
ontological and semantic independence
mod
el a
uton
omy
low high
low
high
federated model sets(autonomous models
with peer-to-peer controls)
composite models(one code with scripting)
unified models(frameworks with normalized
semantics)
centralized models(one code with unified schema)
Approach
Federationschema
inputs outputs
Model A … Model C … Model E … Duplicate Model F … Model G … End duplicate
Cloud
Model A
Model B
Model C
Model D
Model E
Model F
…
Library
Federation management system
Department of Mechanical Engineering, Iowa State University 10
Modeling strategy
High fidelity input data
Empirical and mechanistic simulation
Userinterface
K. M. Bryden, Proceedings of the 7th International Congress on Environmental Modelling and Software (2014).
What’s needed
Cloud-based information objectsModels
Federated modelsCloud-based information objects
Engineering decisionsFederated models
Work flow
Library
Systems Modelers
Federationschema
inputs outputs
Model A … Model C … Model E … Duplicate Model F … Model G … End duplicate
Federation management system
Users
Department of Mechanical Engineering, Iowa State University 10
Modeling strategy
High fidelity input data
Empirical and mechanistic simulation
Userinterface
Analysts
Cloud
Model A
Model B
Model C
Model D
Model E
Model F
…
Federationschema
inputs outputs
Model A … Model C … Model E … Duplicate Model F … Model G … End duplicate
Library
Department of Mechanical Engineering, Iowa State University 10
Modeling strategy
High fidelity input data
Empirical and mechanistic simulation
Userinterface
Components and information flow
Cloud
Model A
Model B
Model C
Model D
Model E
Model F
…
… Federation management system
Federationschema
inputs outputs
Model A … Model C … Model E … Duplicate Model F … Model G … End duplicate
Library
Department of Mechanical Engineering, Iowa State University 10
Modeling strategy
High fidelity input data
Empirical and mechanistic simulation
Userinterface
Components and information flow
Cloud
Model A
Model B
Model C
Model D
Model E
Model F
…
… Federation management system
User experience
Model C1Model A Model E
Model C2
Systems Model
Query
Modify
Explore
Stateless modeling
Model A
Each request is treated as an independent request • communication is an independent set of a request and
a response• no session information is retained• the code has no knowledge of the actions of other
codes within the federation• each member of the federation performs a specific
task• models are reusable for other analysis• models can be strung together like beads on a
complex weaving
Global variables
in
Global variables
out
App implementation (microservices architecture)
Web application
Traffic data from predictive model
Weather data from weather model
Public event data service
User
Back end (server based) • extract relevant data from microservice responses• use in system model• provide needed info to front end
Front end
Micro services
User
Properties of a microservices architecture
• independently deployable
• easy assembly of various models and information sources
• models can be implemented using different programming languages, databases, hardware, and software environments
• direct replacement of a federation member can be performed without disruption to other members or the federation
• microservices perform (provide) one task only
• microservices are reusable
FMS - model interaction
Initialization
Closeout
Wait
Initialization
Closeout
Initial State
Computation
Final State
Persistent memory
Persistent memory
Cache memory
ModelFMS
Model space architecture
Model CModel A Model E
Federation Management SystemSystems Model
Model Space
Data Access Layer
Cloud Platform
Small cookstove model
Secondary Air
Primary Air
ConvectiveHeat Transfer Zone
Flame ZoneGas Phase Combustion,Pollutant Formation
Fuel bed ZoneDrying,Pyrolysis, Char Combustion
Air Flow =Buoyancy
-Friction
Air + Combustion Products
ConductionConvectionRadiation
Air Flow
Mesh Packed bed Flame Heat
transfer Air flow L2 norm Check
Update and increase the detail as needed
Mesh Packed bed Flame Heat
transfer Air flow Check
CFDModel
2D bed model
Design questions
• What is the tradeoff between cost, acceptance, and impact for a particular geometry change to an energy technology?
• How do I maximize the impact of the energy system?
• If I change materials what will be impact of the on the environment and the people?
Systems level design
Local End Use Data
Local Device Data
Local Fuel Data
Local demo- &
geo-graphic
Data
Capital Cost
Energy Consumed
& Delivered
Climate
Health
Fuel and Time cost
LCA Software
Implem-entation
Oppor-tunity Cost
Operating Cost
Incre-mental Cost
Program Data
Rebound
Displacement
User Selection of Device/Use
Scenarios
Social
Secondary Air
Primary Air
ConvectiveHeat Transfer Zone
Flame ZoneGas Phase Combustion,Pollutant Formation
Fuel bed ZoneDrying,Pyrolysis, Char Combustion
Air Flow =Buoyancy
-Friction
Air + Combustion Products
ConductionConvectionRadiation
Air Flow
Components • Cookstoves• Solar hot
water• Lights
Village Energy Model • Rebound• Climate impacts• User acceptance
Agronomic Model • Erosion• Fertility• Crop yield
+ +
Systems modeling
Current Practice Proposed
Software Libraries Web-enabled microservices
Language Specific (overcome with cross-compilers) Agnostic
Integration Cost High Low
Data re-use Not implicit Implicit
Latency Designed to be low Can be higher (needs to be addressed)
Work Plan
Models to cloud-based information objects
Federated modeling tools
User interaction tools
Progress
Models to cloud-based information objects • Complete establishing a separable model set for the NETL Hyper
facility • Developed and implemented the needed linking schema for a simple
energy problem • Outlined the schema needed to link models together using a peer-to-
peer ontology User interaction tools
• Chose UnityTM as the user interaction environment • Initiated development of a simple example based on “hand crafted”
connection points
New modeling paradigm
Snap
Build
Do
Goal
Decision making environments that integrate all the information, models, and other artifacts related to a product or process.
Acknowledgements
U.S. Department of Energy, Office of Fossil Energy U.S. Department of Energy, Office of Energy Efficiency
and Renewable Energy U.S. Army, Armament Research, Development and
Engineering Center Deere & Co
Prof. Dan Ashlock Dr. Aaron Bryden Prof. Kris Bryden Prof. Arne Hallam Prof. Richard LeSar Prof. Tom Shih
Dan Bell Steve Corns Peter Finzell Steven Gent Gengxun Huang Nate Johnson Peter Johnson
Balu Karthikeyan Nordica MacCarty Doug McCorkle David Muth Zach Reinhart Sunil Suram Aditya Velivelli
Current and former PhD students Collaborators Major Funders
And many undergraduate and masters student researchers and our colleagues and friends in the ISU Virtual Reality Applications Center
Contact info
Mark Bryden 515-294-3891 [email protected]