Building an Extensible Framework for Testing New Engineering … · 2015. 4. 27. · % Recycle...
Transcript of Building an Extensible Framework for Testing New Engineering … · 2015. 4. 27. · % Recycle...
Simulation, Modeling, & Decision Science AMES LABORATORY
Dr. Paolo Pezzini Mr. Peter FinzellDr. Mark Bryden
Building an Extensible Framework for
Testing New Engineering Concepts
Current challenge
Advanced concept power systems are oftentightly coupled and difficult to control
• Protect equipment
• Enable coupling of disparate thermal cycle and components
• Larger optimum control envelope
• Support transient operation
Cyber‐Physical System
ModelsCyber-
PhysicalSystem
Hardware
Increasing Accuracy of Exploration
Increasing Flexibility of Exploration
Goal – develop advanced concept power plants
• Development and validation of novel sensor control strategies
• Testing of sensor hardware
• Low cost investigation for new engineering concepts
• Exploring other advanced computational algorithms
Today’s discussion
• Development and validation of novel sensor control strategies
• Testing of sensor hardware
• Low cost investigation for new engineering concepts
• Exploring other advanced computational algorithms
Multi‐agents control solutions Multivariable control schema
Motivation
Air
Exhaust
Air PlenumV-301
MixingVolumeV-304
Load BankE-105
F, T, P
Q
FuelValveModel
NaturalGas
C-100 T-101GeneratorG-102
TE-326
Fuel CellSimulator
V-302
FE-380
PT-305
FV-432
FV-170
FV-162
FV-380
%
Real-TimeSOFC
SystemModels
LHVNatural GasPressure
Load
% Recycle
Real-TimeGasifierSystemModel
Biomass
Oxidant
Flow
Composition
Temperature
SyngasOutput
Hot AirBypas
s
Cold Air Bypass
Bleed AirBypass
Motivation
Air
Exhaust
MixingVolume
Load Bank
Generator
Cyber-physical Devices
Hot AirBypass
Cold Air Bypass
Bleed AirBypass
Anode
CathodeCathode
Electrolyte
Fuel
Fuel Cell Simulator
Traditional Single‐input Single‐output PIDs failed
PID HyperFuel Valve20 % ‐ 80%
Turbine SpeedSet Point
Turbine Speed [rpm]
PID Hyper0‐200%
FV‐170 = 0‐100%
FV‐380 = 100‐200%
Fuel FlowCathode Mass Flow
Equivalence Ratio Set Point
Cathode Airflow [kg/s]
Multi‐agent Control Strategy
Sensors
Control Action
Local Decision
Actuator
Local Agent N
Sensors
Control Action
Local Decision
Actuator
Local Agent 1
…………… .
Multi‐agent Control Strategy
Cathode Airflow
Control Action
Local Decision
Cold‐air Bypass
Cold‐air Bypass Agent
Turbine Speed
Control Action
Local Decision
Electric Load
Electric Load Agent
Multi‐agent Control Schema
Air
Exhaust
MixingVolume
Load Bank
Generator
Cyber-physical System
Hot AirBypass
Cold Air Bypass
Bleed AirBypass
Fuel
Fuel Cell Simulator
AnodeAnode
Cathode
Electrolyte
Block Repository
Cathode Airflow
Control Action
Local Decision
Cold‐air Bypass
Cold‐air Bypass AgentElectric Load Agent
Global Goal
Turbine Speed
Control Action
Local Decision
Electric Load
Less Complexity• No System Modeling• Easy Tuning• Easy Adaptability
Multi‐agent Steady State Results
Stable control at steady state
Multi‐agent Transient Results
Slow Response during tracking
Multi‐agent PID and Transients
Multi-agent transient results PIDs transient results
Developing a Multivariable Control Schema
State Space System Modelling
uInput
yOutput
2
1
4241
3231
2221
1211
4
3
2
1
44434241
34333231
24232221
14131211
4
3
2
1
uu
bbbbbbbb
xxxx
aaaaaaaaaaaaaaaa
xxxx
4
3
2
1
2
1
10000100
xxxx
yy
High Complexity in• System Modeling• Control Law• Tuning• Adaptability
Mitigation of Centralized and Decentralized Interactions
Air
Exhaust
MixingVolume
Load Bank
Generator
Cyber-physical System
Hot AirBypass
Cold Air Bypass
Bleed AirBypass
Fuel
Fuel Cell Simulator
AnodeAnode
Cathode
Electrolyte
Multivariable Steady State Results
Stable control at steady state
Multivariable Transient Results
Multivariable PID and Transients
Multivariable transient results PIDs transient results
Summary
Multi-Agent Control Schema
• Stable control at nominal condition during the first implementation on a hardware system
• Slow response
Multivariable Control Strategy
• Good performance under disturbance rejection and tracking
• Simultaneous control of actuators causes oscillations in advanced power systems
Future Work
• Optimization for the multi‐agent schema
• Reproducibility in the results
• Increase of the size for the multivariable controller
Acknowledgements
MESA Team in Ames
Dan Bell (PhD grad, Iowa State U)
Zach Reinhard (PhD grad, Iowa State U)
Hyper Team in NETL
Dr. David Tucker
Farida Nor Harun (PhD grad, McMaster U)
Valentina Zaccaria (PhD grad, U of Genoa)