Power Optimization of Electro-Thermal Systems Dr. Andrew G. Alleyne

How can modern control techniques and preview be used to improve energy management in constrained systems?

Optimization & Control Methodology

Main Results

• Modern vehicles are a heterogeneous mix of complex interconnected systems of various energy domains

• Electrification of many systems is resulting in increased power loads and thermal waste heat • Enhanced optimization of power generation, distribution, storage, and utilization can be achieved

using dynamic model-based control to improve performance and efficiency while preventing thermal runaway

Vision: With intelligent decision making the power density of existing electro-thermal systems can be improved by a factor of 2

• Event-based control updates • Top-down information flow allows for

effective planning of system states • Bottom-up information flow allows for

effective disturbance estimation • Controllers use robust Model Predictive

Control (MPC) or Genetic Algorithms

• Higher-level controllers - Plan an efficient path using large prediction

horizons - Select mode of operation for lower-level

controllers • Lower-level controllers employ fast

optimization (i.e. Explicit MPC)

Candidate Architecture Aircraft electrical, thermal, and engine systems

12

13 18 21

22

19

17 20

23

161

15 24

2

9

8

7 11

10

53

6

4

14

e1 e2

e3e4

e5

e6

e7

e8e9

e10e11

e12

e13

e14

e15

e16

e17

e18e19

e20

e21

e22

e23

e24

e25

e26

e27e28

e29

Thermal PowerPneumatic PowerElectrical PowerChemical PowerMechanical Power

Source/SinkFast DynamicsMedium DynamicsSlow DynamicsAlgebraicSink

Qengine

Fuel Tank #1

BusBatt

MC 2

Generator

Fan Comp Turb

Sink

MC 1

FC 1

Fuel Tank #2

Qgenerator

Load

FuelAirLiquid Coolant #1Liquid Coolant #2ElectricalMechanical

HX1

HX2 ACM

ENGINE

Applying a graph based modeling approach

Electrical Control System

Thermal Control System

Information Flow

Vehicle Level

System Level

Subsystem Level

Component Level

Physical Level

Plant

60 sec

10 sec

1 sec

0.1 sec

0.001 sec

continuous

1C

2,1C 2,2C

3,1C 3,2C 3,3C

4,1C 4,2C 4,3C 4,4C 4,5C 4,6C

5,4C5,3C5,2C5,1C

Inputs/Outputs

• Overall vehicle and mission state

• Coordinate systems

• State of system, present and future loads and constraints

• Choose best mode

• Optimize operation of subsystems

• Identify constraints

• Determine setpoints for each control actuator

• Deliver servo-level inputs to change physical system

5-Level Control Hierarchy • Matches the natural hierarchy of many mobile systems • Handles temporal and functional separation of systems, subsystems, and components

Key Features

System Disturbances Electrical Loads & Thermal Sink Temperatures

Large electrical loads resulting in significant heat loads

Poor thermal sink availability means heat cannot be easily

removed from the system 0 20 40 60 80 100 120

40

60

80

100

Tem

pera

ture

[°C

]

Time [min.]

0 20 40 60 80 100 120

20

40

60

80

100

Tem

pera

ture

[°C

]

Time [min.]

Cent. Hier.MC2FCMC1

Gen. Tank 1 Eng. Tank 2Cent. Hier.

System States Electrical & Aircraft System Temperatures

Hierarchical control performs better than centralized during

stressful events

Hierarchical control can take anticipatory action well before

centralized control, while utilizing similar amounts of computational resources

System: Tracks vehicle level commands and determines state trajectories for medium time scale dynamics which are passed to subsystem level controllers.

Subsystem: Tracks system level commands and determines state trajectories for fast time scale dynamics. Also measures plant states and communicates that information up through the hierarchy.

Vehicle: Determines state trajectories for slow time scale dynamics and sends those commands to system level controllers. Has some knowledge of disturbances.

State MeasurementsDisturbancesInformation CommunicationActuator Inputs

Vehicle

Thermal Engine Electrical

ElectricalCoolant LoopACM

Plant

Control Hierarchy Vehicle, System, & Subsystem Levels

F/A-18F cutaway graphic by Giuseppe Picarella & Tim Brown

F-35C cutaway graphic by Lockheed Martin

Graph Model of Candidate Architecture Edges capture power flow, vertices represent system states

Controller Development • Select candidate architecture • Represent the architecture as a directed

graph by analyzing how power flows through the system

• Partition graph based upon time constants • Partition graph into various systems &

subsystems based upon functional purposes • Develop individual controllers and determine

information flow

12

2

15

e2

e3

e5

e16

e21

e4

e8

e15

e10

e6

e7

e14e13

e11

e12

e9

7

17

11

9

8

1314

106

3 5

4

12

13 18 21

19

17

10

8

7

15

e12

e13

e14e21

e18

e20

e19

e23

e2613

18

21 2217

20

23

164

e10

e19

e22

e24

e27

e28

e25

e26

12

2

1315

119

3

14

6

5,7,8,10

e2174

e10

e11

e4e3

e9

e15

e5

e16 e21

e13

e12

e144

1 2

24

16 17e1 e17 e22

e29

e2

12

19

17 20

23

16

10

8

7

15

4

21,22

13,18

e12

e18

e13

e14 e21

e19

e23

e26

e25

e10e22

e24

e28

12

13,18,21,22

19

17 20

23

1 2,16

4,5, 6,7,8, 10,15

11

3

14 9

e29e1

e2e11

e22

e10

e13

e14

e21

e12e18

e9e15

e3

e19

e24

e25

e28

e23

24 Vehicle

Engine Sys

Thermal Sys Electrical Sys

Thermal Sub 1

Thermal Sub 2

Electrical Sub

Graph partitioning

Developing controllers within the hierarchy

Preliminary Results