Professional Documents
Culture Documents
FEBRUARY 1993
Abstract-The primary difficulty in charging storage batteries charging process in a manner that minimizes stresses on
is in attaining process regulation that allows fast charging while
the battery. We begin with a general background discus-
avoiding destructive overcharging. This paper presents a two-
level approach’ to controlling the charging process. A general sion of batteries and problems often encountered in the
background discussion of batteries and charging problems is charging of secondary (rechargeable) batteries such as
followed by the presentation of a novel recirculating charge those used for autonomous or portable systems. The prob-
equalization technique that enhances the uniformity of batteries lem of providing sufficient charge for use while maximiz-
made up from long serial strings of cells. Next is a brief ing battery life is then recast as two nested control prob-
description of a straightforward means of embedding equaliza-
tion within an outer-loop supervisory control that maintains a lems. A n inner, servo-level charge equalization problem is
fast charging rate while providing overall protection against first addressed. Charge equalization enhances the uni-
overcharging. Simulation and experimental results confirm the formity of batteries made up from long serial strings of
applicability of the equalization control technique. cells. Maintenance of uniformity improves the life of the
battery as a whole. A novel approach to achieving equal-
ization is presented. An outer, charging supervision prob-
I. INTRODUCTION lem is then formulated with overall battery dynamics in
mind. The supervisor is to automatically source the cur-
I NCREASED concerns about the environment have
provided renewed impetus to investigations of alterna-
tive energy sources. Southern California’s clean air initia-
rent necessary for fast charging and simultaneously pro-
vide overall protection against overcharging. Successful
tive is clearly indicative of the level of adverse effects of equalization greatly simplifies the supervision task. Simu-
air emissions in that region due to excessive consumption lated and experimental results are presented from the
of fossil fuels. Strategic supply reliability and environmen- testing of the charge equalization methodology. The paper
tal impact, thus, have motivated renewed examinations of summarizes and comments on system implementation.
options for the conversion of equipment to use “clean”
forms of energy, such as electric power. 11. BATTTRIES:
CONCEPTUAL
BEHAVIOR
Transportation and industrial processes are two promi-
Batteries are electrochemical devices made up of one
nent areas that are receiving growing attention in terms of
or more building blocks, or cells. A battery cell, as dia-
conversion to electric power. The application of battery
grammed in Fig. l, contains a positive electrode, a nega-
technology, consequently, is a topic of increasing impor-
tive electrode of a different material, and a porous sepa-
tance in many proposed transportation and industrial
rator between the electrodes. These three elements rest
systems [l].Batteries have long been a prime option for
within a common electrolyte environment. During charg-
energy storage in autonomous electric propulsion systems
ing or discharging, the electrodes undergo an
such as those in electric road vehicles. They are com-
oxidation-reduction (redox) reaction, which effects the
monly used as the main energy source for portable equip-
transfer of ions to and/or from the electrodes. The flow
ment, such as communications gear and cordless tools.
of ions manifests the charge flow that appears as battery
End-user reliance on electrically powered equipment has
current. A higher rate of charge transfer implies a need
also prompted increased use of batteries in backup power
for greater exposure of electrode surface area to elec-
systems such as the uninterruptible power supplies (UPS’S)
trolyte. The increase in surface area is often accomplished
often found protecting critical computer systems.
through the use of sintered electrodes that are made from
Widespread application of battery technology has been
powdered electrode material. This material is stamped
limited due to cost per unit of stored energy and the
and heated to just below its melting point to yield an
penalties on battery life incurred by the rapid charging or
electrode that is quite porous. The porosity offers the
recharging required for meaningful levels of utility. This-
desired increase in effective electrode surface area in
paper describes a scheme for controlling the battery-
contact with the electrolyte.
A high rate of charge transfer also implies a need for
Manuscript received April 30, 1992. shorter electrode-to-electrode distance for ion travel. Bat-
The authors are with the Department of Electrical Engineering,
Auburn University, Auburn, AL 36849-5201. tery electrodes consequently are often fabricated as a set
IEEE Log Number 9204610. of closely interleaved plates. Closely spaced interleaving
’
ing the potential for overheating. Maintaining a high
charging current, however, can easily lead to overcharging
of the battery, which can result in irreversible degradation
Porous
L separator
Electrodes
of the battery due to electrolyte leakage and crystalline
dendrite growth [21.
1) Ocercharging-Induced Gas Generation, Pressurization,
and Leakage: During discharge, current flowing from the
Fig. 1. Conceptual diagram of an electrochemical battery cell.
battery is determined by the rate of charge transfer result-
ing from the redox reactions at the battery cell plates.
improves current production capability within the battery This rate is constrained by the amount of electrode sur-
but also increases the possibility of direct electrical con- face area exposed to electrolyte. Conversely, the level of
tact between the plates. The porous separator, often real- current flowing into the battery during charging is con-
ized as a membrane, prevents contact between the closely trolled by the electrical battery charging system and may
spaced plates. well exceed the natural redox reaction capacity of the
The electrolyte in the cell facilitates ion transfer. The battery cell electrodes. The imposition of excess current
electrolyte in some batteries, such as that in results not in charging but in electrolysis of the aqueous
nickel-cadmium (NiCd) cells, does not undergo a chemi- electrolyte away from the electrodes. This electrolysis
cal change. In other batteries, such as the ubiquitous frees gaseous hydrogen and oxygen instead of soluble
lead-acid battery and the highly energy-dense silver zinc hydrogen and/or oxygen ions. Gas bubbles adhering to
battery, ions in the electrolyte participate as part of the electrode surfaces block the access of electrolyte ions to
redox reaction. those surfaces. This reduces the effective electrode sur-
face area and diminishes the current-handling capability
A. The Conflict of Peflormance i‘ersus Cost of the charging process.
The major inhibitor of widespread use of batteries has Because gas displaces more volume than a liquid solu-
been the compromise encountered in trying to produce tion does, such an improper electrochemical reaction pro-
batteries of both useful energy/power density and eco- duces increases in pressure and temperature within sealed
nomical battery lifetime. Batteries of higher energy den- cells. Pressure buildup beyond the seal capability of the
sity usually use more expensive materials, such as the cell case results in leakage of hydrogen and oxygen gas
silver in silver-zinc (AgZn) batteries; or constrained reac- and possible leakage of aqueous electrolyte solution. Pres-
tion environments, such as the high temperatures re- sure-relief venting of battery gases mitigates case damage
quired for sodium-sulfur (NaS) battery operation. Com- and leakage, but any form of fluid loss lowers the volume
pounding the problem is the fact that many of the more of cell electrolyte available for electrode immersion. Low-
power/energy-dense batteries, such as those with zinc ering of electrolyte volume decreases the area of the
electrodes, display rather limited charging cycle lifetime. electrode/electrolyte interface. Fluid loss thus decreases
Higher energy density. thus, implies higher cost due to not the current-handling capability of the battery and is the
only constituent chemical elements or special packaging cause for a large proportion of battery failures. Addition
and safeguards but also more frequent replacement. Al- of replacement water restores the plate/electrolyte inter-
ternatively, a weight and/or volume penalty is incurred if face surface area but can dilute the electrolyte. Addition
one opts for use of cells, such as those in the common of fresh electrolyte obviates this problem, but fresh elec-
lead-acid battery, that have a low cost over their useful trolyte is often caustic; this option consequently may
lifetime. The cost-imposed limitation of available energy introduce an additional chemical-handling problem.
from the battery may be circumvented through more 2) Dendrite Growth: An additional, perhaps more seri-
frequent charging, as in the case of cordless telephones, ous problem that accompanies the overcharging of batter-
or through the use of alternate battery packs, such as ies is the formation of dendrites, or crystalline fingers.
those commonly available as accessories for cordless power These begin developing a) when charging begins after
tools. Such solutions are not applicable, however, in sce- only a partial discharge of the battery, b) during over-
narios requiring the use of high charge capacity with charging of certain types of cells (e.g., those with zinc
relatively infrequent charging. Electric vehicles are an electrodes), and c) during prolonged, low-current ‘trickle”
excellent example of such a situation. charging. Crystalline products of a charging redox reac-
98 IEEE TRANSACTIONS O N INDUSTRIAL ELECTRONICS. VOL. 40, NO. 1. FEBRUARY 1993
tion are supposed to completely dissolve from a cell the level of charge reaches that required by the battery
electrode during a full discharge. They then will ideally application, and for charging to then alertly stop at any
distribute themselves uniformly on the “clean” electrode onset of overcharging.
surface and pores during charging and should completely
dissolve again during the next full discharge. Crystal size 111. INNER-LEVEL
CHARGING
CONTROL:
CHARGE
is small under the conditions favoring uniform distribu- EQUALIZATION
tion. The need to automate the battery charging process is
A partial discharge leaves some crystals remaining on the motivation for developing a control scheme for opti-
the cell plate. Uniform distribution of newly formed crys- mization of battery charging. Specialized integrated cir-
tals on the cell plate implies the clumping of some of the cuits, in fact, have been designed for this very purpose
new crystals with some of those already on the plate. and are readily available. In most cases, however, these
Repetition of this clumping process results in the finger- circuits operate under the assumption that the cells in a
like growth of larger crystals known as dendrites. Dissolu- battery designed for high-power applications share uni-
tion of these larger crystals is less efficient during battery form charge/discharge characteristics. This assumption
discharge since they have smaller surface-to-volume ra- allows the user to treat the battery as just a two-terminal
tios. In some batteries, these crystals simply do not dis- energy storage element. Inherent limitations in the cell
solve. This problem thus results in a so-called “memory” manufacturing process, however, invariably lead to uncer-
effect: the cell’s full-current charge capacity will appear to tainties in cell characteristics. Two detrimental effects of
drop. This effect is exacerbated as the dendrites grow this nonuniformity are that the increased likelihood of
larger and cover more area. Depth of discharge prior to overcharging certain cells becomes more likely and that
charging consequently has an effect on the maintenance the useful charge capacity of the battery decreases.
of a battery’s effective charge capacity. The likelihood of overcharging increases due to the
Dendrite growth is particularly bad during overcharging combination of cell nonuniformity and the common use of
of battery cells with alkaline electrolyte and zinc elec- overall battery terminal voltage as an indicator of level of
trodes. A common means of inhibiting corrosion of the charge. Under normal charging conditions, the terminal
zinc electrode has been the use of electrolyte with a high voltage of a cell increases with the level of charge [2], [ 3 ] .
concentration of zincate ions (Zn0;- ). Battery charging The battery terminal voltage thus is a good level-of-charge
is a reduction reaction that transforms these soluble zin- indicator if the battery cells are truly uniform. In most
cate ions in the electrolyte into crystalline zinc. This realistic cases, cells are not uniform, and repeated cycles
process ideally occurs on the surface of the zinc electrode of charging and discharging lead to large nonuniformities
or in its pores. Overcharging implies a continuation of in cell charge levels and correspondingly different cell
reduction after all exposed electrode surface area has terminal voltage levels. During charging of a battery com-
already been covered with crystals. The continued reduc- posed of a series string of cells, some cells will conse-
tion of the extra zincate in the electrolyte results in quently reach full charge before others and before the
precipitation of zinc crystals within the electrolyte and not overall battery terminal voltage reaches its nominal limit.
necessarily at the electrode. This precipitation often re- Such a scenario leads to overcharging of a subset of cells.
sults in the growth of dendrites in the separator pores. Overcharging, as described in the discussion on battery
Trickle charging is a common means of improving ef- dynamics, is a major cause of such battery degradation as
ficiency of the charging process and also of avoiding loss of charge capacity. the simplest means of avoiding
overcharging. Imposition of too low a charging current, overcharging a nonuniform battery is to monitor the indi-
however, results in electric potential fields in the cell that vidual cell voltages and to stop the charging process as
are too weak to maintain the diffusion process that draws soon as any cell attains full-charge terminal voltage. This
the electrolyte ions to the innermost regions of the elec- strategy, however, limits the deliverable level of charge of
trode pores [2], [ 3 ] . Ions that are not diffusing through the entire battery proportionally to that of the cell with
collect on the electrode surface, where their concentrated the least charge at the stopping point of the charging
conversion results in dendrite growth that only further process.
inhibits diffusion into the pores. The need to minimize battery degradation by maintain-
Accumulated intervals of dendrite growth can result in ing cell uniformity is the motivation for using a dedicated
growth of the crystals through the insulating separator. control mechanism whose purpose is to equalize charge
Dendrite growth, in extreme cases, results in the bridging between cells. This charge equalization works as an inner
of the cell electrodes by the crystal. This effectively cre- servo-level control system that augments an outer-loop
ates a short circuit internal to the battery. Avoiding this state-of-charge process control, which serves as a strategic
problem over the long term implies storing of dendrite- supervisor. Fig. 2 shows the topology of this two-level
prone batteries in a discharged state (i.e., with a minimum control.
number of undissolved crystals) whenever possible. In
light of the effects of trickle charging and overcharging, it A. Charge Equalization Approaches
is best for a battery to charge at the maximum current Three techniques can be used to provide active charge
that does not result in excessive gassing or heating until equalization of serial cells. The first uses a controllable
~
Cell Vbus
~
SUPERVISORYCONTROL
Charge Level
Reference
,,
V R
-
- I
I
Ic
State-of-Charge-dependent
Charge Current Source
level
State-+Charge
I I
I t
Cell Charpe \
h i e l Reference
I
Charge
e
I
Equalized
I 1s Control
Battery
Charge
Y
+
I 1 Sensor
U3. Fig. Circuit model for charge equalization of a serial string of N
cells or N groups of cells using equalizing converters.
Fig. 2. Topology for a two-level battery-charging control system.
remain stable [SI. The converter corresponding to the last Dynamics of Cell/ converter 1
cell is not allowed to divert current; the current through
this last cell, therefore, is i,, which must be directly
controlled through i s to regulate the voltage of the cell.
The charging of N - 1 cells is thus controlled indepen-
dently of the system, whereas charging of the Nth cell
?,bl
(i.e., the last cell to reach V,) is controlled directly by i , .
Fig. 4 shows a power flow diagram for a single cell and
converter where H is the current gain of the converter,
z,, is the cell characteristic, and C is the control algo-
rithm for charging. Whether a converter is on and to what
extent it diverts current depends on the amount that ub is
greater than V,. A further refinement can be made to the
diagram that decouples power flow and control. If the
characteristics of the converters, as well as the cells, are
considered equivalent, then when n converters are oper- Fig. 4. Power flow diagram where H is the current gain of the con-
ating with U,,! > V,, the resulting power flow diagram for verter, 2, is the cell characteristic. and C is the control algorithm for
the system simplifies to that in Fig. S. The system transfer charging.
function from source current i, to charging current i,
becomes
i, 1+ cz,
-
- -
(4)
i, 1 + Czb - nHCZ,.
(5)
C. Open-Cell Operation of Equalization Concerters Fig. 5. Simplified power flow diagram with n of the N converters
operating.
When using certain batteries the possibility exists that a
cell, or group of cells, may open circuit upon failure. This
would leave the power system defunct. However, in this
situation the active shunting system could continue charg- Constant-current charging is easy to implement since
ing and discharging the remaining operational cells and source current regulation is not particularly complicated.
The cell voltage rises freely during charging and can be
provide redundancy in the power system. If a cell fails, the
shunt converter will automatically regulate the voltage used as an indicator of level-of-charge for some cell types.
This approach has two drawbacks: charging times are
across the failed cell during charging and allow current to
flow in the entire string. During discharge all N convert-large if the constant current is small, and pronounced
gassing and possible overcharging may result from high
ers will be activated. The converters across the functional
cells will draw energy from the cells and supply energy tocharging currents as cells near full charge.
the bus. The converter across the open cell will recirculate In constant-voltage charging, the charging current is a
energy and maintain the terminal voltage across the open free variable: currents are high at low charge levels and
cell. The converter algorithm will need to allow standard taper off as charging progresses. This tapering of current
parallel operation of the converters for this mode of is due to the dependency of the charging reaction rate
operation. upon the immediate reactant concentration. In terms of
the electrode as a reactant element, the reactant concen-
IV. OUTER-LEVEL CHARGING CONTROL: tration is approximately proportional to the cumulative
SUPERVISORY CONTROL area A of the electrode surface and pores that has not yet
The equalization strategy just described requires two been covered by crystalline reaction products. Since for-
inputs: a source current i, and a charge level reference ward progress of charging implies a decrease in A , a
V,. The role of the supervisory control system is to simplistic, locally linear model of the nongassing process
provide these inputs. The means for determining appro- would be
priate values for these inputs borrows from the two basic
dA
approaches to battery charging: constant current and con- _--
stant voltage. dt
HUNG e t a / EXTENSION OF BATTERY LIFE VIA CHARGE EQUALIZATION CONTROL 101
The maximum allowable charging current to avoid gassing v. SIMULATION AND EXPERIMENTATION
is proportional to the reaction rate and, thus, to the rate A. Simulation of Equalization
of change of area d A / d t . Any portion of source current The system model used for the simulation of equaliza-
above this level contributes to electrolytic generation of tion is of the form shown in Fig. 4. Control block C of Fig.
hydrogen and oxygen. The trajectory described by (4) in 4 represents the controlled current-to-voltage characteris-
response to a constant ub is an exponential decay in A to tic (or conductance) of the dc-to-dc converter as it oper-
a terminal cumulative unreacted area as t + =, ates above a certain input control threshold voltage V,.
Advantages of the constant-voltage approach depends The block is separated, in Fig. 5 , into conductance and
on judicious determination of a constant-charging voltage dead band blocks. For proof of feasibility, the system was
that avoids gassing and overheating of the cells. A con- physically realized using a flyback converter and given a
stant charging voltage, however, implies that the charging conductance characteristic with a pole at approximately 1
voltage is unavailable as a level-of-charge indicator, and kHz and a 3-dB gain of approximately 50. Input and
too low a charging voltage results in excessively long output filters were also used but not included in the
charging times. Finally, unconstrained charging currents model.
at low levels of charge can result in the overheating of The system block Z,, which represents the battery cell
cells at the beginning of the charging process due to the inputpoutput characteristic, is modeled as an open-circuit
high charging currents initially necessary to maintain the voltage source, a current-to-voltage integrator, and a non-
specified constant charging voltage [ 2 ] ,161. linear resistance, or
The current-diverting capabilities of the equalization
circuitry can be used to marry the benefits of both of the
basic charging approaches. An example of such a combi-
uh = Z,(i,)= 1.85 V + i,, . R , + k [ i , dr (7)
nation is the use of the two-level charging control to The characteristic of the resistance R,, depends on the
effect constant current charging while simultaneously state of charge and cell condition and cannot be general-
avoiding overcharging. In such a scheme, the supervisory ized. However, the value used for this development is
controller acts as an adjustable current source and V, is sufficient to show feasibility of the system. For worst-case
fixed at the cell terminal voltage associated with the onset analysis, a characteristic depicting aging is commonly used
of gassing when charging the cell at a reference current as is done in this simulation. Any change in the character-
i, R e f . The source current i , would initially be set to i , R e t , istic would also indicate variations between different bat-
which should be chosen somewhat lower than the thermo- teries and battery types. For this development, the resis-
dynamically imposed charging current limit. The equaliza- tance is mathematically approximated by a third-order
tion converter for each cell commences conduction as the polynomial fitted to empirically measured data, as shown
terminal voltage for that cell reaches V,. The supervisor in Fig. 6, and takes the form
monitors the duty cycles of the active converters, which
reflect the current gains of the n active converters and R, = 0.0974 - O.O227i, + 1.51
which also serve as the state-of-charge feedback in Fig. 2. x 10p3ii+ 6.33 x 10-3i: (8)
These duty cycles are used to identify the charging cur-
rent i, being sent to the N - TI cells with inactive con- The power flow through the dc-to-dc converter is mod-
verters. The supervisor then regulates i , to a constant eled by the H block. The converter has a fixed output-to-
i h R e fby adjusting i,. This adjustment yields a gradual input voltage ratio set by the taps on the battery voltage
decrease in is, since the increasing number of active string. If all voltage taps are set for the same value, as is
converters will divert an increasing amount of energy to assumed here, then for the tapping of N cells (or N
be recombined as a component of i,. The automatic groups of cells), the ratio of output-to-input voltage is N.
decrease in i, reduces the total power dissipation convert- For a near-constant voltage ratio, the current is propor-
ers as heat. Lower heat dissipation is important when the tional to power. For a converter with an efficiency q , the
battery and charging system will be in confined quarters. transfer function for H is
The charging process ends when the last of the convert-
ers becomes active; i.e., n = N. Throughout the charging (9)
process, those cells with inactive converters charge from a
102 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 40, NO. 1. FEBRUARY 1993
_
1,
--~
N c 0.10
is N - nq' ( 10)
initially charged to 90% capacity, reaches the threshold I +ib3 (40% initial charge)
voltage V, as shown in Fig. 7(b), the top converter diverts c
1 ',\
1,
(initially charged to 80%) reaches threshold, the middle
converter diverts and reduces the cell current i,, and
further increases the current through the third cell. The 0.0
third current increases each time a converter is activated 0 20 40 60 EO 100
and, in the limit, as i,, and i,, approach zero, i, = ni, = Time (min)
3 A as predicted by (5). The converter for the third cell (a)
(40% charge) is not needed because V h 3 does not exceed
V, during the 100 min of charging.
B. Equalization Test Circuit Implementation threshold, V ,
For developing the power topology as described in this
paper, a bank of 30 AgZn cells was used and grouped into
three groups of 10 cells each. Since only n - 1 converters
at most can be on to maintain a stable system, only two of
the three groups of cells have associated converters as
shown in Fig. Na). In a complete system, each group of 1.1 -
cells would have an associated converter, since it is un- 1
known which group would reach peak charge last. The 1.650 20 40 - 60 EO 100
i
HUNG et al.: EXTENSION OF BATTERY LIFE VIA CHARGE EUIJALIZATION CONTROL 103