24
Exploring Adaptive Reconfiguration to Optimize Energy Efficiency in Large - Scale Battery Systems Liang He 1 , Lipeng Gu 2 , Linghe Kong 1 , Yu Gu 1 , Cong Liu 3 , Tian He 4 1 Singapore University of Technology and Design, Singapore 2 Shanghai University of Engineering Science, China 3 University of Texas at Dallas, USA 4 University of Minnesota, USA

Exploring Adaptive Reconfiguration to Optimize …2013.ieee-rtss.org/wp-content/uploads/2014/01/session-5-CPS... · Exploring Adaptive Reconfiguration to Optimize Energy Efficiency

  • Upload
    buicong

  • View
    214

  • Download
    0

Embed Size (px)

Citation preview

Exploring Adaptive Reconfiguration to Optimize Energy Efficiency in Large-Scale Battery Systems

Liang He1, Lipeng Gu2, Linghe Kong1, Yu Gu1, Cong Liu3, Tian He4

1Singapore University of Technology and Design, Singapore2Shanghai University of Engineering Science, China

3University of Texas at Dallas, USA4University of Minnesota, USA

Battery-Powered Cyber-Physical Systems

single cell a few cells tens of cells hundreds of cells

system scale

Exploring Adaptive Reconfiguration to Optimize Energy Efficiency in Large-Scale Battery Systems, IEEE RTSS'13. 2

Large-Scale Battery Systems• Example: 6,800 battery cells for Tesla electric vehicles

• Challengeso cell unbalance issue

• Dynamic battery characteristics during operating cycleso non-linear cell properties: rate-capacity effects

• More energy can be delivered with smaller discharge currentso inevitable cell failure over time and backup cells

3Exploring Adaptive Reconfiguration to Optimize Energy Efficiency in Large-Scale Battery Systems, IEEE RTSS'13.

Reconfigurable Battery Packs

• Reconfigurable batter packs:adjustable connectivity among battery cells

o Controllable discharge pattern to take

advantages of rate-capacity effects

o Balancing cells

o By-pass failed cells

4

Reconfigurable Battery Pack

Exploring Adaptive Reconfiguration to Optimize Energy Efficiency in Large-Scale Battery Systems, IEEE RTSS'13.

Reconfigurable Battery Packs

• Existing efforts: achieving high configuration flexibility with low complexity

• Our investigation: utilizing the configuration flexibility to improve system energy efficiency

5Exploring Adaptive Reconfiguration to Optimize Energy Efficiency in Large-Scale Battery Systems, IEEE RTSS'13.

How could Reconfiguration Help

3.7 x 4 = 14.8 volt

Load required voltage7.4 volt

Voltage Regulator

3.7 x 2 = 7.4 volt

with reconfiguration

6

Design principle 1: Match the supplied and required voltage

levels

Exploring Adaptive Reconfiguration to Optimize Energy Efficiency in Large-Scale Battery Systems, IEEE RTSS'13.

• This also reduces discharge current for individual cells: rate-capacity effect

Design principle 2: Minimize the discharge

current of individual cells

How could Reconfiguration Help

7

Iload

0.5 I

0.5 I

load

Exploring Adaptive Reconfiguration to Optimize Energy Efficiency in Large-Scale Battery Systems, IEEE RTSS'13.

Adaptive Reconfiguration to Supply Load <V, P>

Identify cell strings matching load voltage V

Select as many as such disjoint cell strings and connect them in parallel

8

• To match the supplied and required voltages, we need to

• To minimize the discharge current of battery cells, we need to

𝑣𝑣1 + 𝑣𝑣2 + 𝑣𝑣3= V

Exploring Adaptive Reconfiguration to Optimize Energy Efficiency in Large-Scale Battery Systems, IEEE RTSS'13.

Cell-Graph Representation• A directed Cell-Graph G = <V, E, W>

9Exploring Adaptive Reconfiguration to Optimize Energy Efficiency in Large-Scale Battery Systems, IEEE RTSS'13.

• Bridges the gap between graph theory and battery system optimization

Identify All Feasible Voltage-V Cell Strings

10

Identify all simple paths of weight sum V in graph

• NP-hard in theory (O(NN)), but practically solvableo number of vertices c in the path

is constraint (O(Nc))• e.g., the number of involved cells

could only be c = 8~10 when supplying a 30 V load with Lithium-ion batteries, whose voltage is typically 3.2V-4.2V

o vertex degree d is constraint due to limited configuration flexibility (O(Ndc))

• e.g., 10s(1-4)p for commercial A123 power module

Depth-First-Search with Pruning

Exploring Adaptive Reconfiguration to Optimize Energy Efficiency in Large-Scale Battery Systems, IEEE RTSS'13.

Minimize the Discharge Current by Connecting as Many Feasible Cell String as Possible in Parallel

Double the discharge current!Formulated as a 0-1 integer programming problem with

one constraint

find the maximum disjoint subset of all these paths

11Exploring Adaptive Reconfiguration to Optimize Energy Efficiency in Large-Scale Battery Systems, IEEE RTSS'13.

12

min max {discharge current of cells}

• Identify all simple paths of weight sum Vi for each load, similar to single-load scenario

• Heterogeneous load requirements

• Batteries supporting different loads may have different discharge currents

• A greedy algorithm• Select the load with the largest

discharge current

• Select the string with the least conflict

Supply a Set of Loads <Vi, Pi>

Exploring Adaptive Reconfiguration to Optimize Energy Efficiency in Large-Scale Battery Systems, IEEE RTSS'13.

Select the String with the Least Conflict

“String1 conflicts with String 2”

String 1

String 2

String 3String 4

select the path with the least confliction

String 1 String 2 String 3 String 4String 1 1 1 1String 2 1 0 0String 3 1 0 1String 4 1 0 1

13Exploring Adaptive Reconfiguration to Optimize Energy Efficiency in Large-Scale Battery Systems, IEEE RTSS'13.

Practical Issues

• Energy Consumption for Relay Operationo 25 mA driven current and 3 ms delay per relay operation

o for a typical 2900 mAh battery cell, the ratio between the relay energy cost and the battery capacity is

• 25 mA x 3 ms / 2900 mAh < 7.2 x 10 -6

• Power supply during Configuration Switchingo super-capacitor is a straightforward approach

14Exploring Adaptive Reconfiguration to Optimize Energy Efficiency in Large-Scale Battery Systems, IEEE RTSS'13.

Experiment

15

Battery Pack16 2450mAH AA batteries organized into 8 two-cell modulesLoads1-4 bulb clusters consisting of three 0.5A 3.6W bulbsBaseline (4S2P)Non-reconfigurable conf.: two parallel connected 4-module strings

Exploring Adaptive Reconfiguration to Optimize Energy Efficiency in Large-Scale Battery Systems, IEEE RTSS'13.

Experiment: System Operation Time

16Exploring Adaptive Reconfiguration to Optimize Energy Efficiency in Large-Scale Battery Systems, IEEE RTSS'13.

the lighter the load, the larger the gap between supplied and required voltage, and thus the larger the improvement

1-2 bulb clusters 1-4 bulb clusters 2-4 bulb clusters

4.3 hour 2.9 hour 1.8 hour

Experiment: Temperature

17

*Environment Temperature: 28oC

Methods Cell Temperature after ExperimentsNon-Reconf. 4S2P 43oC

Adaptive Reconf. 31oC

High cell temperature degrades performance in terms of both energy efficiency and reliability.

Burned battery pack in Boeing 787, 2012[1]

[1] Boeing 787 Dreamliner battery problems, http://en.wikipedia.org/wiki/Boeing_787_Dreamliner_battery_problems, 2012.

Exploring Adaptive Reconfiguration to Optimize Energy Efficiency in Large-Scale Battery Systems, IEEE RTSS'13.

EV Driving Trace based Emulation

Emulated LoadsDrive the Mitsubishi Mi Electric Vehicle in downtown Singapore for 45 minutes, and record the discharge/charge voltage & power during this process

18Exploring Adaptive Reconfiguration to Optimize Energy Efficiency in Large-Scale Battery Systems, IEEE RTSS'13.

EV Driving Trace based Emulation

Emulated BatteriesDischarge profile of Panasonic NCR18650 battery cellsA battery pack of 64 x 64 cells is emulated

Exploring Adaptive Reconfiguration to Optimize Energy Efficiency in Large-Scale Battery Systems, IEEE RTSS'13. 19

EV Driving Trace based Emulation

the more unbalanced the cells, the larger the improvementExploring Adaptive Reconfiguration to Optimize Energy Efficiency in Large-Scale Battery Systems, IEEE RTSS'13.

• A control parameter α captures the unbalance degree of battery cells: a smaller α indicates a more unbalance degree

20

Conclusion• Proposed adaptive reconfiguration schemes to

optimize energy efficiency in large-scale battery systems

• Transformed the problem of system reconfigurations to the topology control of the battery cell graph

• Designed practical algorithms to identify desirable reconfigurations for discharging

• Experimental results demonstrate the effectiveness of our proposed adaptive reconfiguration schemes

21Exploring Adaptive Reconfiguration to Optimize Energy Efficiency in Large-Scale Battery Systems, IEEE RTSS'13.

Thanks!

Questions?

22Exploring Adaptive Reconfiguration to Optimize Energy Efficiency in Large-Scale Battery Systems, IEEE RTSS'13.

AMP20 Energy Modules by A123 Systems

Exploring Adaptive Reconfiguration to Optimize Energy Efficiency in Large-Scale Battery Systems, IEEE RTSS'13. 23

6S – 28S (Nominal voltage: 19.2 – 89.6V)1P – 13P (Nominal capacity: 20 – 260Ahs

24

Load #2Load #1 Load #3Dis

char

ge c

urre

nt select a new string for load #1

Load #2Load #1 Load #3Dis

char

ge c

urre

nt

min max {discharge current of cells}

• Identify all simple path of weight sum Vi in graph for each load, similar to single-load scenario

• Heterogeneous load requirements

• Batteries supporting different loads may have different discharge currents

Provide a Set of Loads <Vi, Pi>

Exploring Adaptive Reconfiguration to Optimize Energy Efficiency in Large-Scale Battery Systems, IEEE RTSS'13.