Vehicle Supported Military Microgrids - Oakland...
Transcript of Vehicle Supported Military Microgrids - Oakland...
Automotive Research Center
ARC Conference Case Study 3 | 24 May 2011 Ann Arbor MI
Vehicle Supported Military MicrogridsDesign Scheduling and Regulation for a
ARC Conference Case Study 3 | 24 May 2011, Ann Arbor, MI
Design, Scheduling, and Regulation for a Forward Operating Base
Case Study Team
F l P P l b (PI) T l E l I Hi k H i P AFaculty: Panos Papalambros (PI), Tulga Ersal, Ian Hiskens, Huei Peng, Anna Stefanopoulou, Jeffrey L. Stein
Students/Post‐docs: Changsun Ahn, Abigail Mechtenberg, Diane Peters, John Whitefoot
TARDEC: Erik Kallio, Paul Makar, Jillian McDonaldIndustry: Peter Lilienthal (HOMER Energy)
Motivation 2
Fort Drum
Fort Devens
S lf id AFBSelfridge AFB
Wright‐Patterson AFBFort Hamilton
Reducing dependency on fossil fuel
Increasing use of renewable energy resources on grid
Increasing security, autonomy, and
robustness of military g yand national grids
Motivation 3
Fort Drum
Fort Devens
Selfridge AFB
Wright‐Patterson AFB Fort Hamilton
Motivation: Flexibility 4
Army energy requirements driven by mission requirements
67%Combat Vehicles 10%
37%
Peacetime Contingency Operations
3%
16%
5%3%
6%
67%Combat Aircraft
Tactical Vehicles
Generators
Non Tactical Vehicles
19%
11%
22%
3%
37%
Non‐Tactical Vehicles
Facilities
22%
Generators Generators
“U.S. Army Energy Security and Sustainability: Vital to National Defense,” Torchbearer National Security Report, April 2011
Vehicle Supported Military Microgrid Example 5
• Installation of microgrid at Schofield Barracks, tied to critical infrastructure
• Photovoltaic arrayPhotovoltaic array
• Dedicated electric vehicle charging
• Grid connected
• Conventional generator for extendedConventional generator for extended backup power
• Advanced stationary energy storage
• Load managementg
• AC architecture• Microgrid officially, fully operational as of March 2011of March 2011
• Data collection and operational support for one year
Contingency Basing Structure to Enable Wide Area Security and Combined Arms Maneuver
Any Base Variant
Super FOB
VariantNETWORK
Forward Operating Base
6000 Soldiers
Combat Outpost
Patrol Base
2000 Soldiers
Patrol Base
“At or near the tactical edge, over 300 Soldiers
g70% of fuel used by our military forces can be for power generation” [DARPA BAA‐11‐53]50 Soldiers
Case Study: Vehicle Supported FOB 7
Goal: Developing an integrated tool for designing vehicle supported FOBpp
Design & Operating points,d b
RegulationDesign & Scheduling
disturbance sizes
Battery & APU
Centralized controlLong time‐scale
Decentralized controlShort time‐scale
power constraints
V2G Mi idRegulationLiterature 8
V2G Microgrid
R d d
Adequacy ofexisting capacity
Regulation issues[1, 2]
Reduced Emissions[8, 10, 12, 16]
PEV controlfor regulation
[33, 34, 35]
g p y[3]
Ancillary services[4, 5, 9, 14, 15, 17]
Droop control[1, 27‐31]
Economics[4, 9, 15, 17,36]
Control [18‐23]
Inverter control[32]
TODAY’S CASE STUDY
Impact on Renewable Po er
Microgrid modeling, controls, and practical implementation
[38]
MicrogridDesign
PEV [40,41]V2G
Technological Power[5, 10, 12, 14, 15]
[ 0,4 ]
[42] Optimal dispatching of distributed energy resources
[39]
gDesignAspects[37]
References available upon requestDesign&Scheduling
FOB Simulations[43]
FOB with Solar + JLTVs with export power 9
APUBattery Pack
APU
~50 soldiers120 kW peak po er Joint Light Tactical Vehicle with on‐board120 kW peak power67 kW average power$6‐8/gal fully‐burdened
Joint Light Tactical Vehicle with on‐board auxiliary power unit and battery storage • 30 kW export power
k h bPower supplied by Solar Panels and Power‐Export JLTVs
cost of fuelRadial network
• 2 ‐ 20 kWh Li‐ion battery
Design & Operating points,disturbance sizes
Regulationg
Schedulingdisturbance sizes
Battery & APUpower constraints
Centralized controlLong time‐scale
Decentralized controlShort time‐scale
power constraints
Design + Scheduling Goal 11
Operational Constraints
Optimization Component Sizes and
Military Requirements
Minimizing Fuel and Capital Cost
SchedulingRequirements
Uncertainties
Design + Scheduling: FOB with Solar + JLTVs 12
1) Solar Panel Peak Power
2) Number of JLTVs2) Number of JLTVs3) Battery pack size(Export power fixed: 30 kW)
Solar Irradiance (W/m2)
Solar Radiance Inputs and Error 13
Solar SupplyHourly solar irradiance taken from NASA d f K b l
Solar Irradiance (W/m )24
NASA data for area near Kabul.
Design and Scheduling optimization used variable but ou
r of Day
optimization used variable, but deterministic solar input.
Ho
Solar Error ModelingDay of Year
1
gAn error variable created for non‐linear model predictive control algorithm of the solar panel output power.
APU Model 14
Non‐Linear APU ModelSurrogate model constructed for generator from Cummins dieselgenerator from Cummins diesel generator data (50 generators in data set)
Efficiency
Model Inputs and OutputsAPU efficiency calculated as a function of APU rated power and ppercent load
APU
Model Inputs: Power Load w/ Error 15
Known
2kW/soldier2kW/soldier
Power Load Profile
(Part 1) Diurnal power load modeled with radial basis functions based on 5 key parameters.
(Part 2) Seasonal power load modeled with average daily temperature sine function to model change in HVAC needs based on published ECUs power load based on temperature.
Diurnal Power Load 16
Parameters
Adding Seasonal Variations 17
80) Summer
High
60
70
tion
(oF Air Conditioning
HeatSummer
Comfort Level (65/75)
40
50
re V
aria
t SummerLowHeat Only
Winter
20
30
mpe
ratu
r
High
WinterLow
0 5 10 15 200
10Tem
50‐FOB = 4*ECUs .
0 5 10 15 20
Time of Day (hr)Used our Temperature model and HVAC model from N. C. McCaskey, “Renewable Energy Systems for Forward Operating Bases: A Simulations‐Based Optimization Approach ” Colorado State University, 2010
Total Power Load (kW)
20100
110
Operational Power Load (kW)
20 60
65
70
Diurnal Yearly Load without Error
s du
ring
the
Day
10
15
70
80
90
s du
ring
the
Day
10
15
40
45
50
55
Hou
rs
5
10
50
60
70
Hou
rs
5
10
25
30
35
Total Power Load w/ Error 1 (kW)
130
365 Days of Year
50 100 150 200 250 300 350
40
4-ECU: HVACs (kW)
365 Days of Year
50 100 150 200 250 300 350
20
the
Day
15
20
100
110
120
he D
ay
15
20
40
45
Seasonal Variations from HVACfor Yearly Power Load without ErrorOperational + Seasonal Variationsfor Yearly Power Load without Error
Hou
rs d
urin
g t
10
60
70
80
90
Hou
rs d
urin
g th
1030
35
Power Load used in this Case Study.
365 Days of Year
50 100 150 200 250 300 350
5
40
50
60
365 Days of Year
50 100 150 200 250 300 350
5
20
25
Sample FOB Case to Compare 19
Example of CONUS FOB Demand Data with UM Interpretation 20
• Data collected in support of NetZero+ JCTD at Ft. Irwin
Mathematical Framework – MG Hubs 21
Hub‐based energy network [Geidl]
diesel gas
electricity electricity
+ ‐
H2H1
diesel gas
solar H3
Model Inputs L: load at hubModel InputsIndividual hub layoutsComponent sub‐modelsPower loads vs. time
L: load at hubc: energy conversion factorP: energy input to hubs: energy storage efficiency
Solar supply : change in energy storageE
Geidl, M. and G. Andersson, “A modeling and optimization approach for multiple energy carrier power flow,” in Proc. IEEE Power Eng. Soc. PowerTech, St. Petersburg, 2005, Russian Federation.
Design Optimization Framework 22
Optimal design of a microgrid requires balancing capital costs of equipment with expected operation costs
requires solving a nested optimal scheduling problem for each designrequires solving a nested optimal scheduling problem for each designcould also consider other costs (e.g., transportation, maintenance)
Optimal Design ProblemOptimal Design Problem
Optimal Scheduling Problem
minimize: capital + operation costs
variables: component sizes Optimal Scheduling Problemvariables: component sizes
constraints: ‐ operational limitsb k ’
minimize: operation costs (e.g., over 1 year)
‐ backup req’ts‐ reliability‐ volume/footprint limits‐ silent watch req’ts
y )
variables: energy generation / storage scheduling
q‐ other military req’ts constraints: operational limits
A $7/ ll f ll b d d t f f l
Initial Results: Design + Scheduling 23
89 kW Solar Panel(~4,100 ft2 rigid PV or ~14,400 ft2 of flexible PV)
Assumes $7/gallon fully‐burdened cost of fuel
( 4,100 ft rigid PV or 14,400 ft of flexible PV)
Plug‐in JLTVs: 4
Battery Capacity: 2.1 kWh
Added Weight: 35 kg
PV panel + JLTV w/ export power( li d) $9 00
“Business as Usual” – Two generators( li d) $ 00 Cap Cost (annualized): $91,500
Fuel Use (annual): 38,000 gallonsChange in Fuel Use: ‐29%
Cap Cost (annualized): $4,400Fuel Use (annual): 53,600 gallons
Example Optimal Scheduling Result 24
Optimal scheduling on a time horizon results in strategic energy storage use
GEN
BATT
LOAD
Optimal Scheduling Result for One Day Power Flow
Design & Operating points,disturbance sizesg
Schedulingdisturbance sizes
Centralized controlLong time‐scale
T i di b l d i d
Vulnerable Operating State w/ Disturbance 26
Transient power disturbances can cause power loss and equipment damage
Regulation problem can show design limitations and suggest improvements
50100Total Power LoadSolar
Drop in Solar Power of 50 kW
30
40
60
80
er (kW)
kW)
SolarGeneratorBattery
10
20
20
40
tery Pow
e
Power (k
10
0
20
0 Batt
‐10‐20
1 3 5 7 9 11 13 15 17 19 21 23
Hour of Day
Design & Operating points,disturbance sizes
Regulationg
Schedulingdisturbance sizes
Centralized controlLong time‐scale
Decentralized controlShort time‐scale
What is the regulation problem? 28
Voltage regulationKeep length constant
Frequency regulationKeep rotational speed constant
Significance of regulation 29
60 Hz60 Hz60 Hz
Frequency Variation Voltage Variation
LOW HIGH
Speed
LOW HIGH
Speed
LOW HIGH
Speed
110V
LOW HIGH
110V
LOW HIGH
110V
LOW HIGH
110V
LOW HIGH
110V
LOW HIGH
110V
LOW HIGH
Speed
Temperature
Efficiency
Power
Speed
Temperature
Efficiency
Power
Speed
Temperature
Efficiency
Power
Voltage increases.Light is getting bright.Voltage exceeds the limitation.Light bulb burns out.Devices may get harmed.
Voltage decreases.Light is getting dark (brown out).Current to the devices increases.
Voltage exceeds the limitation.Current to the devices increases too much.
Power
May burn out earlier than designed.
Power
Impedance goes down.Current goes up and cooling goes down.May burn out earlier than designed.
Power
Frequency and power balance 30
Power Generation = Power DemandPower Generation > Power DemandPower Generation < Power Demand
60Hz
60.5Hz
60Hz
60.5Hz
60Hz
60.5Hz
59.5Hz59.5Hz59.5Hz
PowerGeneration
PowerDemand
PowerGeneration
PowerDemand
PowerGeneration
PowerDemand
Frequency regulation framework 31
Demand fluctuation
Disturbance
PV power drop
Disturbance
Demand fluctuation
Disturbance
PV power drop
Disturbance
Demand fluctuation
Disturbance
t
Pdemandd
t
PPV
t
Pdemandd d
t
PPV
t
Pdemandd d
Performance Goal
Battery PackAPU Battery PackAPU
• Frequency: 60Hz ± 0.5Hz• Voltage: 110V± 5%
Fast ActuatorLimited Energy Capacity
p2
Battery PackSlow Actuator
Unlimited Energy Capacity
p1
APUFast Actuator
Limited Energy Capacity
p2
Battery PackSlow Actuator
Unlimited Energy Capacity
p1
APU
tt
p1
tt
p1
Simple vs. full model 32
Simple Modelfor Control Design
Full Modelfor Validationg
LinearFocus on frequency
NonlinearFocus on frequency and voltageFocus on frequency Focus on frequency and voltage
Includes battery model
Battery model 33
T.K. Lee and Zoran Filipi
D. Di Domenico, A. Stefanopoulou, and G. Fiengo, "Lithium‐ion battery state of charge and critical surface charge estimation using an electrochemical model‐based extended Kalman filter," Journal of Dynamic Systems, Measurement and Control, Transactions of the ASME, vol. 132, no. 6.
based on
Power Controller Design 34
60Hz
60.5Hz
Housing59.5Hz
Office
Housing
HousingC1
C2
ControllersControllers
Requirements
Assuming 2.3C max rate
Frequency Control Result 35
60Hz
60.5Hz
Housing60Hz
60.5Hz
Housing60Hz
60.5Hz
Housing60Hz
60.5Hz
Housing60Hz
60.5Hz
Housing60Hz
60.5Hz
Housing60Hz
60.5Hz
Housing59.5Hz
Office
Housing
HousingC159.5Hz
Office
Housing
HousingC159.5Hz
Office
Housing
HousingC159.5Hz
Office
Housing
HousingC159.5Hz
Office
Housing
HousingC159.5Hz
Office
Housing
HousingC159.5Hz
Office
Housing
HousingC1
C2C2C2 .C2C2 .C2 .C2 .
RequiredBattery Power
Simple vs. full model 36
Simple Modelfor Control Design
Full Modelfor Validation
Simple Modelfor Control Design
Full Modelfor Validationg
LinearFocus on frequency
NonlinearFocus on frequency and voltage
g
Focus on frequency Focus on frequency and voltageIncludes battery model
Simulation with full model 37
Battery assumed constant voltage source
Simulation with full model 38
19kWh Battery with diffusion dynamics
No solution!
Physical explanation 39
Current
VoltageVoltage
Voltage drop limits AC power
One solution 40
……
……
…
Add more cells in parallel
Simulation with full model 41
38kWh Battery with diffusion dynamics
Simulation with full model 42
53.8kWh Battery with diffusion dynamicsy y
Evolution of battery sizing 43
53.8 kWh
19 kWh
8.4 kWh
After design and After regulation After consideringAfter design and scheduling
considerations
After regulationconsiderations
After considering battery dynamics
Design & Operating points,disturbance sizes
Regulationg
Schedulingdisturbance sizes
Battery & APUpower constraints
Centralized controlLong time‐scale
Decentralized controlShort time‐scale
power constraints
Additi l t i t f R l ti bl t l t 53 8 kWh b tt it
Final Results: Des. + Sched. + Reg. 45
87 kW Solar Panel(~4 000 ft2 i id PV (1 9 t i t ) ~14 000 ft2 f fl ibl PV (6 6 t i t ))
Additional constraint from Regulation problem: at least 53.8 kWh battery capacity
(~4,000 ft2 rigid PV (1.9x tennis courts) or ~14,000 ft2 of flexible PV (6.6x tennis courts))
Plug‐in JLTVs: 4
Battery Capacity: 14.3 kWhAdded Weight: 240 kgAdded Weight: 240 kg
28% saving in fuel
New DesignOriginal Design
28% saving in fuelcompared to “business as usual”
New DesignCap Cost (annualized): $94,100Fuel Use (annual): 38,600 gallons
Original DesignCap Cost (annualized): $91,500Fuel Use (annual): 38,000 gallons
Summary of findings 46
• Critical coupling exists between short and long time scale system behaviortime scale system behavior
• Proposed integrated approach can take this coupling into accountcoupling into account
• Smart energy management strategies can d f lreduce fuel use
• Battery voltage fluctuations may impose additional design considerations
Next steps 47
• Develop and incorporate models for additional components (e.g., hybridized vehicles)components (e.g., hybridized vehicles)
• Develop framework for designing and analyzing reconfigurable topologyanalyzing reconfigurable topology
• Analyze interplay between prediction errors d h iand horizon
• Develop flexible and scalable controllers
• Validation of proposed methods and tools
Conclusion 48
O i d h d d l h lOur integrated methods and tools can help achieve this power and energy vision
Fort Drum
Fort Devens
Selfridge AFB
Wright‐Patterson AFB Fort Hamilton