classical algorithms fail - Electrical & Computer Engineering

download report

Transcript classical algorithms fail - Electrical & Computer Engineering

Field Sustainment Power Conditioning
#TA3-040-5
Brad Lehman, Northeastern University
Khalil Shujaee, Clark Atlanta University
Wes Tipton, Army Research Laboratories
Presented by: Florent Boico
Dept. Elect. & Comp. Engin.
Northeastern University
May 31 2005
Motivation/Background
• Background
– Dept. Army recently mandated that all training exercises must use
rechargeable batteries;
– Estimated to save $70M annually (versus non-rechargeable);
– Soldiers like rechargeable batteries so much that they are bringing them into
combat also;
– About 75% of Army rechargeable batteries are BB390 NiMH (4lbs). (BB390
has 2 x 12V legs and can be used as either 24V or 12V battery.)
•
Motivation for Solar Chargers
– Soldiers carry four BB390 batteries
(= 16 lbs) for portable electronic
equipment;
– Forward field observers, scouts,
special ops, are constrained to stay
within 10 miles of TOC (Tactical
Operation Center where there is a
charging facility shelter);
– Portable solar arrays carried by
soldier (~1lb) reduce number of
batteries carried and eliminate the
need to stay near TOC.
Operation area
10 mi.
Soldiers using batteries
TOC
Spare batteries, generator ,
Chargers, shelter, etc.
Motivation/Background
• Portable Solar Chargers
– Being field tested by CERDEC C2D
Army Power Division;
– Best solution: soldier directly
connects solar array to battery and
lets charge all day while on a
mission;
– Companies have attempted and
failed to build power electronic
charge regulators to control the
charging.
• Issues when Solar Charging
– Experiments show reduced battery
capacity;
– Batteries sometimes overheat and
vent: can no longer be used;
– Batteries are constantly being
recharged by soldier, even if
battery lightly discharged
• Leads to reduced life cycle of
battery.
Three different portable solar
arrays charging three NiMH
batteries.
(Picture courtesy of Dennis Lane, CERDEC,
C2D, Army Power Division)
NiMH Overcharge Detection
Full State Of Charge
Conventional (known) NiMH charge control algorithms stop charging (or
switch to trickle charge) when battery voltage begins to decrease or when rate of
cell temperature begins to substantially rise.
Solar Charging BB390
I
–
–
–
–
V
Companies have attempted to work with CERDEC to build solar chargers:
• Chargers failed: They falsely terminate charging before completion;
• CERDEC refuses to use any of these chargers.
Known charging algorithms are applicable to constant power source:
• Termination for “dumb” NiMH batteries (BB390) occurs based on battery V, dV/dt, time,
and sometimes temperature T or dT/dt.
Solar arrays produce varying current sources depending on clouds
• Fast charge
Slow charge
Fast charge …
How to correctly predict charge termination for DUMB batteries like BB390 (basic research)?
Typical sunny day
measurement
The negative voltage
slope shows that the
battery is overcharging
Typical cloudy day
(scattered clouds)
Changes in the solar
array current cause
changes in the battery
voltage.
Conventional charge control algorithm falsely terminate charging !
Variation of Temperature Throughout the Day
Temperature inside the battery pack
42
40
38
36
34
32
30
28
26
24
0
50
100
150
200
250
300
350
400
450
Time (min)
Large variations of the temperature can falsely trigger charge termination
on temperature based algorithms.
Phase 1 Charger Prototype (2004)
Prototype
BB390
leg1
voltage
leg 1&2
Battery
thermistors
leg1
leg2
ADC
PIC µC
charging current
Serial connection for
data logging (optional)
leg2
A prototype of the charger has been
built.
Its characteristics include :
• charge monitoring by
sensing :
-voltage & current across
each leg
- cell temperature
• clamps on the BB390 battery.
• optional RS-232 connection for
data logging and evaluation by a
computer.
• algorithm fully upgradable.
Voltage Charge Control Algorithm (2004)
Initialisation :
Vmax=0
dpos=0
After reset the
algorithm waits
for a positive
voltage slope to
allow overcharge
detection
Charge, n=n+1
sense Vbatt(n)
& Ibatt(n)
If the voltage
suddenly drops or
Vmax=Vbatt(n) yes
if strong
discrepancies in the
current is detected,
yes
the algorithm is
reset to prevent
Reset :
false overcharge Vmax=0
detection
dpos=0
yes
Minimum and maximum
current over a period of 5
minutes
Vbatt(n)>Vmax
dpos=1
no
|dV/dt|>thsld1
no
no
Imax- Imin>thsld2
dV/dt>thsld3
no
Vmax- Vbatt>0.1V
& dpos=1
yes
Trickle charge
yes
no
2005 Results :
•Refined Maximum Power Point Tracker
•Differential temperature based charge control
algorithm developed
Phase II: Maximum Power Point Tracking (MPPT)
Bypass
switch
Iout
Up-Down
converter
PV array
d
DPWM ADC
filter
µ controller
Current
Sensing
resistor
We have built preliminary Phase II chargers
that include MPPT:
• Adjusting the duty ratio of the Up-Down
converter forces the solar array to operate at its
maximum producing power point;
• MPPT adaptively optimize charging to
different NiMH batteries (12V, 24V, 9.6V, etc.)
• Bypass switch improves power efficiency
when MPPT not needed.
Higher Charging Current is Achieved with MPPT
Battery
Voltage
Current when solar array is directly Current when the proposed charger
with MPPT is used.
connected to the battery
4.8V
660 mA
720 mA
12V
310 mA
310 mA
24V
0 mA
160 mA
New Charge Charge Control Algorithm
Developed in 2005: Differential Temperature Method
•
•
•
The algorithm is based on measuring the difference in the temperature
of each of the two legs of the BB390.
When overcharge occurs in one leg, it can be detected by comparison
with the other leg.
Method gives improved robustness
 Not sensitive to changing illumination conditions
 Not sensitive to changing ambient temperatures
 100% success rate after dozens of experiments!
2005 - Differential Temperature Algorithm
•
•
The algorithm is based on measuring the difference in the temperature of each
of the two legs of the BB390.
When overcharge occurs in one leg, it can be detected by comparison with the
other leg.

C
Battery or leg
When battery overcharges, it heats up.
2005 - Differential Temperature Algorithm

C
Battery can also heat up due to external causes
(e.g. : sun) but is not fully charged yet.
2005 - Differential Temperature Algorithm

C

C
How can one differentiate between overcharging and
external heating ?
2005 - Differential Temperature Algorithm
Two Independent Legs
thermistors
Supposing one leg is charged and the other is left open :
• In case of overcharge the temperature will rise in one leg only
• If the pack is heated from outside, the temperature will rise in both legs
2005 Differential Temperature Algorithm

C

C
thermistors
T1-T2> threshold
Overcharge
T1-T2< threshold
No Overcharge
2005 - Differential Temperature Algorithm
•The algorithm functions as
follows :
Charge
leg 1&2
no
dT1/dt > thsld1
yes
yes
Timer1>10min
no
no
no
dT2/dt > thsld1
yes
Charge leg 1
Stop leg 2
timer1=0
Charge leg 2
Stop leg 1
timer1=0
increment timer
increment timer
d(T1-T2)/dt
>thsld2
d(T2-T1)/dt
>thsld2
yes
yes
leg1 fully charged
leg2 fully charged
Switch to trickle
charge in each leg
yes
no
no
Timer1>10min
-Detects a rise in temperature in any of
the legs.
-Keep charging the leg that has been
detected as potentially overcharging
-Measure the slope of differential
temperature measurement to detect
overcharge
-After a certain time if no overcharge is
detected, charge is resumed in both
legs.
•The differential sensing of the
temperature reduces the effect of
external heating and ambient
temperature on the measurement.
• Overcharging detection of one
leg via temperature using this
method is more robust.
•Thresholds are current dependant.
2005 - Differential Temperature Algorithm
Experimental Results
External heating
V
I
Thsld reached,
Entering potential overcharge mode.
Normal charge resumed,
Internal balancing takes place.
Thsld reached,
Entering potential overcharge.
Overcharge detected on leg1,
Charging process over.
T1
dT1
dt
d (T1  T2 )
dt
Algorithm based
on simple
derivative fails
New algorithm delivers
accurate full SOC
detection
External heating does not fool new algorithm
Conclusion
•More robust voltage charge control algorithm (2004)
•Differential temperature charge control algorithm (2005)
•Maximum Power Point Tracker (2004-2005)
•Algorithms are being implemented inside the
prototype charger.
• The phase one prototype is stand alone and mechanicaly compatible with BB390
batteries