Optimization of PHEV/EV Battery Charging

1

Optimization of PHEV/EV Battery Charging

Lawrence WangCURENT YSP PresentationsRM 52511:00-11:25

2

Outline•Background and Introduction•Problem Statement and Technical

Approach•Battery charging system model and

simulation•Minimum loss algorithm formulation and

test•Conclusion and future work

3

Background

• EV and PHEV will help solve energy dependency and environmental issues & are becoming popular

• All EV/PHEV rely on batteries

• One of challenges is battery charging – the success of the EV will depend on availability of charging infrastructure

4

Battery Charger Systems

On-Board Charging System

Off-Board Charging SystemZhang et al “Research on energy efficiency of the vehicle's battery pack arch on energy efficiency of the vehicle's battery pack,” 2011 ICEICE

5

Review of Previous Charging Schemes•Previous studies showed optimal charging

(complex function) was very close to the CC-CV method.

E. Inoa, Jin Wang, “PHEV Charging Strategies for Maximized Energy Saving,” IEEE Transactions on Vehicular Technology, 2011

6

Problem Statement & Approach• Problem Statement

▫Given a certain maximum time to charge a battery from a starting SOC to a final SOC, find the optimal current and voltage to minimize energy loss

• Approach▫Model for both battery and charger▫Determine total loss and charging time relationship

with charging current, voltage and SOC▫Design an method for selection of current and

voltage.

7

Battery Model & Verification

Min Chen, G.A. Rincon-Mora, “Accurate electrical battery model capable of predicting runtime and I-V performance,” IEEE Transactions on Energy Conversion

Battery model is modified to be consistent with the battery in the Nissan Leaf Electric Vehicle. Components are scaled by a factor of 96 cell units and a 66.2 Ah capacity. Verified to be consistent (16kWh)

8

Charger & Loss Modeldcv

CAv ABvBCv

Laps bps cps

ans bns cns

C Battery

p

n

avbv

cv

0.92

0.93

0.94

0.95

0.96

0.97

0.98

0 50 100 150IDC (A)

Charging efficiency variation with current.

9

For a given SOC

Vary maximum current and

voltage

Calculate battery and charger power

loss

Integrate for energy loss

Matlab Model

End

10

Loss as Function of Current and SOC

11

Time as Function of Current and SOC

12

Voltage Relationship Loss vs. Current for 3 Maximum Voltages Time vs. Current for 3 Maximum Voltages

The loss and time approaches an asymptote for higher maximum voltages. (E.g. setting a maximum voltage of 450 will yield the same result as 412.8V, as it cannot reach that voltage within the allotted time.)

13

Observation•Raising the maximum voltage allows charging

to be achieved faster, but also creates more loss at the same current. Ultimately, the decrease in time corresponds to a lower loss.

•There is a current that will lead to minimum loss. This current is consistent throughout all SOC.

•This means that for all currents below this minimum, it is better to use the threshold current.

14

Minimum Loss Algorithm

Determine Imin from

Input treq

treq < tmin

YesNo

treq = tmin

Determine SOCinit

Imin < Icmin

I = Icmin

Yes

I = Imin

No

15

iz{1,1}

w

pz

dotprod9

w

pz

dotprod8

w

pz

dotprod7

w

pz

dotprod6

w

pz

dotprod5

w

pz

dotprod4

w

pz

dotprod3

w

pz

dotprod2

w

pz

dotprod10

w

pz

dotprod1

Mux

Mux

Weights

IW{1,1}(9,:)'

Weights

IW{1,1}(8,:)'

Weights

IW{1,1}(7,:)'

Weights

IW{1,1}(6,:)'

Weights

IW{1,1}(5,:)'

Weights

IW{1,1}(4,:)'

Weights

IW{1,1}(3,:)'

Weights

IW{1,1}(2,:)'

Weights

IW{1,1}(10,:)'

Weights

IW{1,1}(1,:)'

pd{1,1}

Output

a{1}

Process Output 1

Process Input 1

Layer 2

Layer 1

a{1}

Input

Implementation•A neural network was developed as a

function 𝑡 = 𝑓ሺ𝑆𝑂𝐶𝑖𝑛𝑖𝑡 ,𝐼𝑐,𝑉maxሻ

16

Test in Simulation

# SOCinit treq (hrs) Loss (MJ) DLoss

1 20% 0.5 5.10 15.7%

2 20% 1.0 3.53 41.7%

3 20% 8.0 2.98 50.7%

4 50% 0.5 2.13 35.8%

5 50% 1.0 1.74 47.6%

6 50% 8.0 1.72 48.2%

17

Conclusion and Future Work•A simple voltage and current selection

criterion is developed in CC-CV charging mode that will result in minimal loss for a given charging time.

•Both battery and charger loss are considered. The algorithm tested in simulation benefits both slow and fast charging.

•Experimental verification will be performed in the future.

18

Acknowledgement

•CURENT faculty, staff, and graduate students

19

Questions