Improved charge algorithms for valve regulated lead ... - IEEE Xplore

2 downloads 62 Views 494KB Size Report
Elizabeth D. Sexton“, Robert F. Nelson2, and John B. Olson'. 'Recombination Technologies LLC, 909 Santa Fe Drive, Denver, CO 80204. 'Optima Batteries ...
Improved Charge Algorithms for Valve Regulated Lead Acid Batteries Elizabeth D. Sexton“, Robert F. Nelson2,and John B. Olson’ ‘Optima Batteries, Incorporated, 17500 E. 22ndAvenue, Aurora, CO, 80011 E-mail esextonO,optimabatteries.com ‘Recombination Technologies LLC, 909 Santa Fe Drive, Denver, CO 80204 ABSTRACT The cycle life obtained from valve-regulated lead-acid (VRLA) batteries is strongly influenced by the manner in which they have been charged over their lifetime. Although VKLA batteries initially behave similarly to their flooded counterparts, that behavior changes as the batteries age and the oxygen generatiodrecombination cycle begins to dominate at near 100% full charge. This means that an increasing portion of the applied charge is consumed in the recombination cycle and that more and more overcharge must be applied to maintain full capacity. The overall result is that the battery heats up because of increased overcharge and oxygen generation. Conventional charge approaches attempt to deal with rising temperatures by lowering the current during the overcharge phase. However, this approach does not ultimately prevent capacity loss, and a battery charged thusly typically will yield 200-300 cycles to SO% of initial capacity. The main failure mode appears to be undercharging of the negative plate, not positive-plate corrosion. These issues were studied under ALABC Project Nos. 8-007.1 and 13-007.2. Two approaches, caIled Partial State of Recharge (PSOR) and Current Interrupt (CI) were successful in extending battery life. PSOR uses nine limited recharge cycles followed by a tenth cycle using 120% charge return. The best PSOR cycle life to date is 1 160 cycles to 50% and 800 cycles to 80%. C1 uses a high current in the overcharge applied discontinuously to control battery temperature. CI effectively maintains negative-plate capacity, with an Optima group 34 deep-cycle battery yielding 415 cycles to 80% initial capacity and 760 cycles to 50%. Introduction

Both EV and HEV usages can be considered “cycling applications”, as both require alternating power utilization and recharge, although the details of each are very different. Irrespective of the application, using VRLA batteries in cycling requires consideration of the unique features o f VRLA batteries, especially in the design of appropriate charge atgorithms. Because of the limited amount of electrolyte, VRLA batteries operate in the “almost starvedalmost flooded” mode (3). Ideally, there is enough electrolyte to provide the desired capacity and also a vapor space to facilitate oxygen recombination.

Valve-regulated lead-acid (VRLA) batteries have become widely used in applications previously reserved for flooded lead-acid or other advanced batteries. VKLA batteries offer distinct advantages for some applications, such as automotive or motive power, in that they never need watering. Indeed, VRLA batteries have been used in electric vehicles, because of their high power, lower cost compared to other advanced batteries, and availability, as well as being maintenance Bee (I). The robust performance of VRLA batteries have also been of interest for hybrid electric vehicle (HEV) applications (2).

Table 1. Charging Reactions in a Lead-Acid Battery Optima Batteries, Inc., a leader in spiral cell VRLA batteries, developed deep-cycle batteries useful for electric vehicle (EV) and HEV applications, These are, 52 Ah, 12 V group 34 (group 34) and 16.5 Ah, 12 V (HEV) batteries. The group 34 is in current production while the HEV is a prototype. The HEV batteries have thinner plates than the group 34, providing improved charge acceptance and discharge power over conventional leadacid technology (3), as well as coulombic efficiencies over 90% over a range of states of charge and discharge currents (4). The group 34 has found acceptance among EV users because of its high power and good performance in pack applications.

P o s i t l v e Plate

+ 2 H2O = P b O z + tIzSO4 + 2 e‘ + 2 H 2 0 = O 2 + I H* + 4 e PbSO4 Pb

+

H2O = P b O + 2 e -

+

2 H’

Negative Plate

PbS04

+

2 Ht

+

2 e ’ = P b + H2S04

2 H’ + 2 e ‘ = H Z

02 + 4 H’ + 4 e. = 2 H 2 0

21 1

2

H’

The discussion of charging algorithms and cycle life must include a consideration of the charging reactions

summarized in Table 1. Although charging is dominated by the lead sulfate to lead dioxide (positive plate), and lead sulfate to elementai lead (negative plate) reactions, the gassing reactions are also important. Hydrogen evolution can happen at any charging voltage and occurs independently of other reactions (8). While the presence of a lead sulfate coating on the negative plate is thought to suppress hydrogen evolution, usage of low-antimony grid can also reduce hydrogen gassing (6). Oxygen evolved at the positive electrode at voltages reached during overcharge diffuses to the negative plate, where it is reduced to form water (5J Both gassing reactions are thermodynamically favored but kinetically hindered. Excessive oxygen reaching the negative plate will cause depolarization, resulting in undercharging and capacity loss (5 7). This depolarization phenomenon is believed to be the reason VRLA batteries must receive significant overcharge in order to achieve and maintain full capacity. In the overcharge, a portion of the current feeds oxygen recombination and part actually charges the negative plate. As the m o u n t of overcharge increases, oxygen recombination begins to dominate and the battery heats up. This contributes to accelerating capacity loss, and can lead to thermal runaway. Any charge algorithm that extends cycle life must provide a mechanism to charge the negative plate while controlling oxygen recombination

(5,8). Since the charge efficiency also changes as the battery ages, this complicates development of charge algorithms, as it is unlikely that a single routine would maximize efficiency over the life of the battery. Changes in eficiency are probably in response to changes in oxygen recombination. Indeed, it has been shown that a sulfate gradient on the negative plate develops after HEV cycling, with the greatest concentration of sulfate being at the bottom of the plate (9). This gradient undoubtedly develops as a result of oxygen recombination, but the exact mechanism is unclear. There is demonstrated inhomogeneity that develops on both positive and negative plates after even slight usage (9). Battery modeling has shown that there is a potentiat gradient drop down the cells (IO). This potential gradient may alter the oxygen production reaction on the positive plate, causing a gradual increase in sulfatian on the bottom of the negative plate and also contribute to locaiized plate inhomogeneity.

Optima Batteries Inc. has studied VRLA charge algorithms for the last two years as part of the Advanced Lead Acid Battery Consortium (ALABC) research programs B-007.1 and B-007.2. Two charging approaches have successfulIy extended cycle life. These are: Current Interrupt (CF)and Partial Slate of Recharge (PSOR). The strategy behind these approaches is minimizing overcharge and maximizing negative-plate polarization, resulting in improved efficiency. In PSOR, the battery is given nine cycles of essentially 100% charge return, and the tenth is a conditioning charge (120% nominal initial capacity), This allows the battery to be operated at an average 70% to 80% initial capacity, CI is used with any charge routine, and is applied to the overcharge phase. It involves interrupting the constant current overcharge with regular, short rest steps, on the 10-20 second time scale. This seems to allow charging of the negative plate while minimizing oxygen recombination. The best results have been obtained by combining PSOR with CI, yielding 800 cycles to 80% and 1160 cycles to 50% initial capaciry. Both techniques will be presented in detail, with a discussion for future work. Experimental Two sizes of Optima deep-cycle batteries, 52 Ah, 12 V (group 34), and 16.5 Ah, 12 V (HBV)were prepared for cycte life testing by receiving five conditioning cycles. These cycles consisted of a 25 A (group34) or 16.5 A (BEV) discharge to 10.5 V, followed by a constant current recharge to 120% nominal initial capacity. This was 4 A for 16 hours for the group 34 battery, and 1.7 A for I2 hours for the HEV battery.

The cycIing equipment used was Bitrode Corporation (Fenton, MO) LCN models 25 A, 18V and 200 A, 18 V. The temperature was measured in the HEV battery by inserting the thermocouple into a thermal well, located in the center of a cell. The temperature of the group 34 battery was measured by placing the thermocouple in the space between the two center cells, below the cast on strap level. A guide was fabricated to hoid the thermocouple against the cell. Both thermocouple placements yielded temperature data within a few degrees of the cell interior. Reference electrode and pressure transducer data were taken by drilling a hole in the battery cover and inserting a tube sealed with Teflon-silicone sealant. The reference electrode was n mercury/mercurous sulfate electrode (Optima design). The pressure transducer (Iomega) was nominal 0 to 15 PSI,and externally calibrated. Because it

was too difficult to make the sea1 gas tight, only reference electrode or pressure data were collected at a given time. Cycle life testing was carried out using discharge rates of 25 A for the group 34 battery or 16.5 A for the HEV battery. Cur off voltages were 10.5 V, corresponding to 100% depth of discharge (DOD) cycling. The algorithms were evaluated t o 80% and 50% nominal capacity, with SO% being used as the experiment termination point.

Of particular note is the amount of overcharge in the CI experiment. This is quite low, only 105% charge return initially and less than 110% at 300 cycles. This is quite remarkable given the need of VRLA batteries for substantjai overcharge in order to maintain full capacity. The low overcharge indicates improved efficiency of CI over conventional constant current charging techniques. This improved efficiency may be due to a disruption in the oxygen recombination cycle, thereby allowing a greater amount ofthe overcharge to charge the negative plate.

Results and Discussion Current Interrupt A group 34 battery was cycled using current interrupt (CI) overcharge, with results given in Figures 1. A standard cycle-iife curve is given for comparison, and the charge return is also plotted.

I

ID

111

160

111

111

111

ab0

111

4KQ

111

L6P

44#

LIQ

!$e

IQb

CI was also applied to an WEV battery. These data are shown in Figure 3.

figure 3.

4bQ

Cycle

Fig. I. Group 34 Battery Cycle Life Using C1

This figure requires some explanation.

,*,,

_

fl..

_

-

--

~

~

~

~

_

_

_

-

-

-

-

l

_

The HEV

battery was cycled up to cycle 319 using a fast in-rush,

The charge routine used 200 A to 60% previous charge return, 50 A to SO%, and 15 A to loo%, with C1 in the overcharge. The CI charge voltage profile is given in Figure 2. CI used a current pulse of 20 seconds followed by a 15 second rest. The resting voltage between current pulses was used to terminate the overcharge. CI used 5 A initially, but was gradually increased as the efficiency dropped. At the end of life, 25 A pulses were used. The CI charge routine yieided 415 cycles to SO%, and 760 cycles to 50%. This is compared to the standard CV cycte life of 170 cycles to 80% and 300 cycles to 50%. (CV algorithm is 14.7 V charge to 1 A, followed by a 2 A for 1 hour overcharge.) $..*

HEV Cycle LifelCl Charging

constant current routine (16.5 A charge to 80% charge return, followed by 3.3 A to end of charge, including overcharge). An attempt was made to use the point on the voltage-time overcharge curve where the slope was zero as the termination point. This resulted in the sharp decline in capacity seen at cycles 60 and 100. The capacity was recovered by giving the battery a conditioning charge. However, when the capacity began dropping at cycle 250, the algorithm was changed to include CI in the overcharge. This brought the battery back to 80% initid capacity briefly, with the capacity staying at 65% far roughly 250 more cycles, _

l

CI was successful in doubling the life of this battery using 50% initial capacity as the end OF life. Nonetheless,

_

_

l

_

_

-

-

-

I

-

usage of CI did not maintain the battery at the 80% capacity level. it is not understood why CI did not return the battery to full-capacity, and remains the subject of ongoing investigation.

--

..-_----

Mechnaisin of CI AIthough the focus of the ALABC program has been to find charge algorithms that would extend VRLA cycle life, some work has been done to

-

Figure 2 Cl Voltage Profile Cycle 40

213

+

understand how CI works, in order to optimize it over the life of the battery, One aspect of this is to collect reference electrode and pressure data, in order to study the effect of CT on the positive and negative plates. The next three figures show positive and negative reference voltage curves from the aforementioned HEV battery. The mercury/mercurous sujfate .reference electrode data are presented as absolute values (for the negative plate). Initial discharge curves from the HEV baitery are shown in Figure 4. The battery is clearly positive limited.

-

Figure 6. HEV Cell Voltage Cycle 502 Examination of the pressure and voltage data from the last fourth of the HEV battery life shows an interesting relationship between the pressure and voltage (Figure 7). The battery pressure rises and falls as the current is applied, with the pressure curve having the same overall shape as the voltage curve. This suggests that CI is directly modulating oxygen generation. Because the pressure does not keep increasing, it is likely hydrogen generation is not significant at this point in battery life, although this remains to be shown. 11.0

,

( LIB

This situation begins to change after 150 cycles (Figure 5). Here, the battery shows the positive and negative plate as being more equal. The negative plate also shows the characteristic voltage "hump" associated with the onset of oxygen recombination .

I

$0

m

, I

U

IL

(P

nme (mlnutesl

Figure 7. Cl Voltage and Pressure - Cycte 426

0

IO

21

II

24

LO

61

IO

no

BO

Partial State of Recharge Knowing that too much overcharge too early in the battery life causes the battery to fail prematurely, a technique of giving the battery a limited charge with only a periodic full charge was developed. Called "partial state of recharge" (PSOR), this technique is different from other partial-state-of-charge regimes in that the battery always receives a full discharge to 10.5 V . However, the battery is charged to only -100% chatge retum of the previous discharge by limiting the charge step to a voltage cut off, initially at 15.1 V. The battery receives a full conditioning charge to 120% nominal capacity every ten cycles. This has the effect of limiting the average overcharge on each cycle to 106%. A

110

Tlme Imlnutss)

Figure 5. HEV Cell Voltage Cycle 105 I

At the end of life, the battery is clearly negative limited (Figure 6), and the effect of CI is evident. The positive plate shows little voltage increase from using CI charging. However, the negative plate shows voltage behavior that tracks with the applied current, showing that CI is successful in polarizing the negative plate.

214

group 34 cycle life curve using this approach is given in Figure 8.

almost 50% more Ah than the simple constant current overcharge, and 44% more Wh.

,I

l

l

_

-

-

-

_

_

_

Yhrughpui I

,

I

LO

.

111

I

110

r

i

-

_

-

_

_

l

_

I

-

-

-

-

~

-

-

-

-

-

-

-

-

1 1 8 9 2 A n . 248345 Wh 241

110

IS0

1-1

too

6IP

,BO

LI1

,IO

49% 44%

248345

445622

The variability between the Ah and ?Vh percent improvement in Table 2 is due to the collapse in voltage behavior evident after 600 cycles (Figure 10). The end of charge voltage on the overcharge cycles decreases and the temperature increases. This corresponds to an overall decrease in end-of-charge voltage for all cycles, meaning that the voltage difference between initial discharge and cut-off voltage is less, lowering the Wh. The PSOR (without CI conditioning charge) battery was taken off test before the voltage collapsed completely, as the capacity limit was reached. Indeed, the voltage behavior shown in Figure 10 is correlated with a drop in end of charge voltage for the nine PSOR cycles at cycles 300 and'600. At cycle 300, the battery would not meet the 15.1 V cut off, and the voltage limit was reduces to 14.7 V . At 600 cycles, the battery failed to meet this limit, and the voltage cut off was adjusted downward to 14.1 V, another indicator of the continuing collapse of the charge voltage

.

IPD

%Improvement

PSQR w/CI PSOR WOCK 38580 19632

Ah Wh

m*

Cycler

Figure 8. PSOR with Conditioning Charging The charge return is illustrated by the dashed line and the discharge capacity is shown by the solid line. The charge is 25 A to 80% charge return, followed by 10 A to 15.1 V, The full conditioning charge every I O cycles is 5.2 A for 12 hours. The battery went to SO% of initial capacity ai 320 cycles, and feil to 50% at 660 cycles, in spite oFmore than 100% overcharge at the end of life.

Although the group 34 cycle life shown in Figure 7 is promising, it was hoped at the project onset that it would be possible to reach 1000 cycles with an Optima group 34 battery. It was decided to combine PSOR with CI; that is, to use the PSOR partial recharge algorithm for nine cycles, then use a tenth cycle recharge that combines a constant current charge to 100% followed by a CI overcharge to 120%. This result is presented in Figure 9.

CUWO.

The temperature rise shown in Figure 10 is attributed to the increase in oxygen recombination and the increasing overcharge needed to maintain battery capacity. The drop shown at cycle 850 is due to the introduction of n rest step designed to allow the battery to cool during overcharge. Remarkably, the battery went another 500 cycles after the conditioning charge temperature increased dramatically. IO

I4 L

,*,

-

-

I I

--

_

-

- - --

- - dv - - - I

-

.

~~---------------~-----------------

1

111

$00

111

111

1 0

L11

111

IO&

100

4111

CYClt

Figure I O . PSOR End of Conditioning Charge Voltage and Temperature vs. Cycle

215

The decline in end of charge voitage corresponds also to an increase in charge retum. This is shown in Figure 11. Here, the percent charge retum is averaged over 100 cycle increments, and stays at less than 110% until cycles 600700, where it increases. This is reflective of unrecoverable efficiency loss, and is likely due to the battery becoming negative plate limited. Nonetheless, the battery yields another 500 cycles before it drops below 50%. 111

indicating that the mechanism of facilitated oxygen recombination may not be directly related to water loss.

Future work in the ALABC program will focus on these mechanistic issues, and optimizing the improved algorithms. An understanding of general charging issues will also be used to design the next generation of advanced VRLA batteries.

I

Acknowledgment

This work was funded jointly by the Advanced Lead Acid Battery Consortium and Optima Batteries Inc. through the member cost share program. Their support is acknowledged.

REFERENCES I . Olson, J.B. and Puester, N.H. Procscdings of Eleventh Annual Battery Conference Long Beach, CA January, 1996 2. Olson, J.B. Proceedings of IECEC ’96,Washington, DC August, 1996 3 . Olson, I.B. and Sexton, ED. Proceedings of TwelBh Annual Battery Conference Long Beach, CA January, 1997 4. Sexton, E.D. and Olson, J.B. Proceedings of Thirteenth Annuul Batteq Conference Long Beach, CA January, 1998 5. Nelson, R.F. 7’”Asian Battery Conference, 1997 6. Lam, L.T.,Douglas, J.D., Pillig, R. and Rand, D.A.J. J. Power Sources 1994 48, 2 19-232 7 . D. Berndt and U. Teutsch J. Electrochem. Soc. 1996 143(3), 790 8. S. Altung and B. Zachau-Christiansen J. Power Sources 1994 52,201 9. Sexton, E.D. and Olson, J.B. Proceedings of Fourteenth Annual Boidely Conference Long Beach, CA January, 1999 IO. Harb, I.N. and LaFollette, R.M. Proceedings ofthe 33*”Intersociety Energy Conversion Eplgineering Conference Colorado Springs, CO August, 1998

Cyda htarval

Figure 11. O h PSOR Charge Return vs. Cycle Conclusions

The two best algorithms, Current Interrupt and Partial State of Recharge, have been shown to extend cycle life of Optima deep-cycle batteries. Indeed, the combination of the two techniques has given a group 34 cycle life of 800 cycles to SO% and 1 I60 cycles to 50% initial capacity, and 50% more Ah throughput than PSOR alone. Clearly, CI offers improved efficiency over conventional charge methods. Further work will focus on optimizing the technique, especially as it relates to changes in the charge efficiency of the battery as it ages. There remains, however, the vexing question of how CI works, and the general question of VRLA charging. The HEV battery cycle life was improved relative to the cycle life without CI. Yet, CI was not successful in returning full capacity to this battery. There is probably a secondary mechanism related to battery overheating, from receiving too much overcharge too early in its life. This appears to be a different phenomenon from the end-of-life behavior shown by the “PSOR with/CI“ battery, where the end of charge voltage drops as the charge eficiency drops. Here, the drop in end-of-charge voltage is caused by oxygen recombination, in that the negative plate voltage drops in response to the reaction of oxygen, There is also the issue of oxygen transport being facilitated as the battery ages, due to separator dry out, the development of gas channels and increased “void space”. Yet, the Optima batteries discussed herein showed a weight loss of 4% etectrolyte,

216