Вы находитесь на странице: 1из 7

Solar Energy Vol. 50, No. 5, pp.

399-405, 1993

0038-092X/93 $6.00 + .00


Copyright 1993 Pergamon Press Ltd.

Printed in the U.S.A.

LEAD ACID B A T T E R Y STORAGE M O D E L


FOR H Y B R I D E N E R G Y SYSTEMS
JAMES F. MANWELL* and JON G, McGOWAN*
Renewable Energy Research Laboratory, University of Massachusetts, Amherst, MA 01003, U.S.A.
Abstract--This paper describes a new battery model developed for use in time series performance models
of hybrid energy systems. The model is intended to overcome some of the ditficulties associated with currently
used methods. It is based on the approach of chemical kinetics. This model, which can be used for charging
and discharging, is specifically concerned with the apparent change in capacity as a function of charge and
discharge rates. It assumes that the charge can be stored in two ways, either as immediately available or as
chemically bound. As described in this paper, it requires the determination of as few as three constants.
Examples of the deviation of the battery model constants and comparisons of the new model with those
used previously are given. Based on the success of the new model, it has been incorporated into the latest
versions of the University of Massachusetts's wind/diesel simulation codes.

1. I N T R O D U C T I O N / B A C K G R O U N D

Successful design, modeling, and operation of a hybrid


energy system using battery storage requires an understanding of the operation of the battery. For example, at the same time that a wind/diesel system
simulator [ 1] at the University of Massachusetts ( A m hearst) was being designed and constructed, an initial
lead/acid battery model was also developed [ 2,3 ]. Its
purpose was to aid in the conceptual design of the simulator and to provide a basis for the evaluation of similar energy storage-based systems.
The initial battery model was based on the battery
terminal voltage and its changes under charge, discharge, and recovery. The charging analysis was based
on the work of Facinelli[4], whereas the discharge
analysis was based on one developed by researchers at
the U.S. Department of Energy's Battery Energy Storage Test Facility [ 5,6 ], this model is sometimes referred
to as the BEST model. In both cases, the battery was
modeled as a voltage source, E, in series with an internal resistance, R0. Details of the components of the
model are summarized in [ 2 ], and its use for simulation
of the discharging process is discussed later.
Use of this battery model typically requires extensive measurements of voltage and current. These measurements are required for the determination of a large
set of constants (on the order of 10). In addition, the
analytical components of the model are essentially
phenomenological and were developed for either
charging or discharging, hut not both. Also, integrating
alternating charge/discharge cycles was a difficult and
cumbersome task and is of questionable validity.
In order to eliminate some of the problems of the
initial battery model, we have recently developed a
new battery model. The new model was developed
specifically for use in quasisteady time series simulations of power-producing systems that incorporate
battery storage. In such simulations it is desirable to
assume that many of the variables, including the power

from the battery, are constant over each time step. For
such simulations the primary concern is in describing
power flows rather than voltage characteristics. It was
also desired to be able to take advantage of the manufacturers' data for batteries to the m a x i m u m possible
degree in determining the parameters of the model.
The new model, as described in the next sections
of this paper, accounts for (a) decrease in capacity with
increasing charge or discharge rates, (b) recovery in
charge, (c) increase of voltage with charging current
and state of charge, and (d) decrease in voltage with
discharging current and state of charge. At present, the
model will not reflect (a) the sharp increase in voltage
near the end of charging, (b) the sharp drop in voltage
when the battery is nearly empty, or (c) temperature
effects.
2. A N A L Y T I C A L M O D E L D E V E L O P M E N T

2.1 Overall battery model


This model is similar to the initial ones in that the
battery is viewed as a voltage source in series with a
resistance, as illustrated in Fig. 1. As in the BEST
model, the internal resistance, Ro, is assumed to be
constant and the internal voltage, E, varies with state
of charge. As before, the terminal voltage, V, is given
by

V=E-IRo.
In addition, a current regulator is assumed to be
placed between the battery and the load. This is done
to ensure that the power draw, and hence current, is
constant over the time step. In practical terms, this
could be a pulse width modulated voltage convertor
such as that described in [ 7 ].
The voltage source is modeled as two tanks separated by a conductance (although a capacitor/resistor
analogy would serve as well). One tank holds charge
that is immediately available to be used by the load.
The other tank holds charge that is chemically bound.
The conductance corresponds to the rate constant of

* ISES member.
399

400

J.F. MANWELLand J. G. McGOWAN


R

~N/N/N/"

dq2 =

--

dt

+k(1 - c)qt - kcq2.

(7)

Equations (6) and (7) can be solved with the help of


Laplace transforms to give

Rload
ql = qa.oe -kt +

(q0kc - I)(1 - e -kt)


k
Ic(kt-

1 + e -k')

(8)

q2 = q2,oe -kt + qo( 1 -- c ) ( 1 -- e -kt)

Fig. 1. Overall battery electrical schematic.

I(1 - c)(kt - 1 + e -kt)

(9)

k
a chemical reaction/diffusion process (assumed firstorder rate) by which the bound charge becomes available. The rate at which bound charge becomes available
is proportional to the difference in "head" of the two
tanks. Because a chemical kinetics process is used as
its basis, the model is referred to as the kinetic battery
model (KiBaM).
Figure 2 illustrates the model. Each tank has unit
depth, but different widths, corresponding to different
volumes. The width of tank 1 ("available") is c, and
that of tank 2 ( " b o u n d " ) is 1 - c. The combined width
of the two tanks is thus equal to 1, and gives a combined
tank area of unity. The combined volume of the tanks
is qm,x. Because the area is unity, when both tanks are
full the head, hmax is also equal to qm,x. The valve between the two tanks has a fixed conductance, k'. The
current regulator operates such that the current, I, is
constant during the time step of interest.
The equations describing the model are the following:

dql
dt

--

k'(hl

h2)

(1)
(2)

dq--'22 = k ' ( h l - h2),


dt

where q~ = available charge and qz = bound charge.


The head in each tank is given by the volume divided
by the area, that is,
hi = q l / C

(3)

h2 = q 2 / ( 1 - c).

(4)

For mathematical simplicity, a new rate constant k is


defined as
k'
k - c(1
c ' - " "- ~

"

art

= -I-

k(l - c)ql + kcq2

2.2 L o s s t e r m s
The battery model explicitly assumes that charge is
conserved under charging and discharge. Energy, however, is not conserved. When voltage is considered explicitly, energetic losses are determined by the difference in voltage between charging and discharging. This
will result in losses that are proportional to the current
squared times the internal resistance. When voltage is
not considered explicitly, it should suffice to assume a
nominal energetic efficiency for the battery.
2.3. T e m p e r a t u r e effects
As noted previously, the model does not presently
consider temperature effects. For many wind/diesel
systems, which were the original impetus for the model,
the batteries are in a conditioned building, where the
temperature is not expected to change drastically. Failure to reflect temperature effects should not be a serious
limitation in these cases. Where temperature is important, a family of constants could be obtained. It
may be noted, however, that there appears to be no
reason that an expanded model could not include temperature effects more explicitly. In particular, most
chemical processes are sped up at higher temperatures
(corresponding to a higher k in KiBaM ). Such behavior
is consistent with the higher capacities observed at
higher temperatures, suggesting that applying the Arrhenius rate equation to the model may be appropriate.
....................................................................... i

--~

1-c

"*-

-,- c "*- i

(5)

Substituting eqns (3), (4), and (5) into eqns (1)


and (2) gives
--

where q~,oand q2,0 the amount of available and bound


charge, respectively, at the beginning of the calculation,
and q0 = q~,o + q2,0.

(~)
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

(6)

I--~
I

k'

Fig. 2. Kinetic battery model.

401

Lead acid battery storage model


2.4 D e t e r m i n a t i o n o f c o n s t a n t s
The model can be used in two ways, depending on
whether or not voltage is to be considered explicitly.
When battery voltage variation with state of charge is
not of concern, three constants are needed for the
model: qmax, the m a x i m u m capacity of the battery; c,
the fraction of capacity that may hold available charge;
and k, the rate constant. They can be found from battery capacity data, such as those typically provided by
manufacturers. When terminal voltage is of interest or
when constants are to be determined by tests, it is also
necessary to consider internal resistance, Ro.
2.4.1 V o l t a g e not i m p o r t a n t . When voltage is not
of concern, the battery constants may be found from
discharge capacity data. For computational convenience it is useful to normalize capacities to that for
some slow discharge rate, corresponding to discharge
time T = t2. The data may then be expressed in terms
of a ratio, Ft,,t2, of capacities:

F l i t 2 = qT=q
'
qT=t2 '

For example,
( 1 - e - k 2 ) ( 1 - c) + kc20
Fi,2o = 20((1 - e-k)(1 -- C) + kc)

Given any two Ft,.t2's, it is possible to calculate corresponding k's and c's. Equation 14 can be rewritten
as follows to give c as a function o f k (with Ft = Ft,,t2
for clarity). When the same c is obtained for a given
k at two different discharge rates, that k and c are the
ones to be used.
Ft( 1 - e - k " ) t 2 - ( 1 - e-kt2)tl
C = Ft(1 - e kl')t2 -- (1 -- e-kt~)tl

When multiple values of Ft,,t2 are known, a leastsquares fit can be used to find the best values of c
and k.
F i n d i n g q m ~ It is most convenient to find the maxi m u m battery capacity by starting from some reasonably slow discharge rate, such as the 20-h rate. The
m a x i m u m capacity is then found from eqn 11:

qmax =

-Ict.

(10)

The discharge current to empty the battery in time t,


IT-t, may be found in terms of c and k by setting q~
= 0 in eqn (10). Thus

It=,=(1

qma ck
--e-kt)(l--c)+kct"

kct

(17)

For example, if qT=2Ois the 20-h capacity, corresponding to a discharge current of


qT=20
I T = 2 0 --

2O

then

I( 1 -- e-k')( 1 -- C)
q l = qmaxC - -

qr=t{(1 - e-k')(1 -- C) + k c t }

T = t.

For example, the ratios Fi,2o and Fro,20 indicate the


relative a m o u n t of the charge that can be extracted at
faster rates (here 1 and 10 h) compared to the 20-h
rate.
Finding c and k. Assuming that the battery had been
full at the start of the tests, initially q j / q 2 = c / ( 1 - c).
U n d e r that condition eqn (8) becomes

(16)

-- k F f l l t 2 + kilt2

where
qT=t = discharge capacity at discharge time

(15)

qmax =

qT=20{(l -- e-k20)(1 - c) + k c 2 0 )
kc20

(18)

2.4.2 V o l t a g e effects i m p o r t a n t . When voltage


variations are important, the model may be expanded
with the following equation:
V= E-

IRo.

(19)

(11)

For the discharge case


The total charge out of the battery at a given rate is
simply the rate times the time. Therefore,

ql

E = Emi, + (E0.d -- E m l n ) - - ,

(20)

q l ,max

Ftl,t 2 -

F, .

lllT:q

tzlr:12

tjqmaxck
(1 - e-ktl)(1 -- C) + k c h
.
.
.
.
t 2qmax c k
( I -- e-kt2)( 1 -- C) + k c t 2

(12)

(13)
q|

E=

Eo,c + ( E m a x -

Eox) -

(21)

ql,max '

more simply expressed


t, [(i - e-k'2)(l - - C ) + k c t 2 ]
F'"'2=~2[(1
e-kt')(1-c)+kch]"

where Emi, = the m i n i m u m allowed internal discharge


voltage ( " e m p t y " ) and Eo,d = the m a x i m u m internal
discharge voltage ("full"). For charging

(14)

where E m a x = the m a x i m u m allowed charging voltage


and E0x = the m i n i m u m allowed charging voltage.
The internal resistance, Ro, is the slope ( d V / d I )
of V versus I when the battery is full. It may be found

J. F. MANWELLand J. G. McGOWAN

402

by performing a series of discharge tests at constant


current. The voltage corresponding to a full battery for
a given current is found by extrapolation. Plotting that
voltage versus the current and finding the slope gives
the resistance. A similar approach can be used for
charging.
Using the internal resistance, the voltage drop (for
discharging) across it is calculated for each discharge
current or interest. The battery is defined to be empty
at some internal voltage, Em~n, before the knee in the
discharge. This will correspond to a terminal voltage,
Vmin, of Vmi,, = Emin IRo.
--

3. APPLICATION OF THE ANALYTICALMODEL


The KiBaM can be used in a quasisteady hybrid
energy system model in charging and discharging situations as follows. For simplicity this procedure describes
the case in which voltage variations are not of concern.

3.1 Discharging
At each time interval, find the amount of power
required, P , ~ . Using a nominal system voltage, V,
find the desired discharge current, Id = PneejV.
Next, find the m a x i m u m discharge current, Id,. . . .
that can be obtained by using eqn (8), assuming that
at the end of the interval ql = 0. Here t = the length
of the system model time step. Note that q0 and qL0
are the total charge and the available charge, respectively, at the beginning of the time step:
k q l , o e -kt + q o k c ( 1 -

Id . . . .

~-

--

e -kt

e -kt)

c(kt - 1 + e-kt) '

(22)

The m a x i m u m discharge that can be obtained is


then Id.maxt. If the required charge is greater than that,
then the rest must be made up by an auxilliary, e.g.,
diesels. If the required charge is less than the maximum
available, then Id is found above. There is also no deficit

in this case. The amount of charge left at the end of


the interval in both forms, q~ and q2, is found from
eqns (8) and (9), using Id.
3.2 Charging
As above, find the desired charging current, I~
= P, eeo/V. Note that in charging Ic will be a negative
number.
Find the m a x i m u m allowable charge rate, lc.m~x,
from how much may be put into the batteries. It is
found from eqn (8), with q~ = cq . . . . In this case the
m a x i m u m allowed charging current is

I,. . . . .

-kcqmax + kql,oe TM + q 0 k c ( 1 - e - k l )
1 - - e - k ' + c ( k t - 1 + e TM)
. (23)

If/~ is less negative than Ic . . . . . then I = Pried~V, as


above. Otherwise, Ic = I,, . . . . . The final values of q~
and q2 are determined from eqns (8) and (9) as they
are in the discharge case.
4. SAMPLE ANALYTICALRESULTS
The following figures illustrate the use of KiBaM.
Figure 3 summarizes the manufacturer's battery capacity versus discharge current data for a G N X 6000C
battery. That battery has a rated capacity of 74 Ah at
the 8-h rate. Superimposed on the data is a curve predicted by the KiBaM, whose constants were determined
from the data. As can be seen, the predictions closely
match the data. Figures 4, 5, and 6 illustrate predictions
from the BEST and KiBaM models for a hypothetical
six-cell lead acid battery similar to the Exide batteries
used in the University of Massachusetts's wind/diesel
system simulator[ l].
In the BEST discharge model[5,6], the battery is
modeled as a voltage source, E, in series with an internal resistance, R0. E varies with the effective depth
of discharge, X, which in turn is a function of both the

8O
"o

70
o~-

ca

,,1
t,,}

GNX data

60

a
5O

-I-

o,,

4O

I=

30
20

20

40

60

80

100

120

Discharge C u r r e n t , A m p s
Fig. 3. Comparison of model and manufacturer's data.

Lead acid battery storage model

403

200

180

P
Q.

160

E
>,
4-J
~

140

0
m

120

Ki6aM
100
0

10

20

30

40

50

60

Discharge Current, Amps


Fig. 4. Comparison of capacities from the Battery Energy Storage Test (BEST) and kinetic battery model
( KiBaM) models.

discharge current and the battery history. The model


is based on six observable quantities: current, I; terminal voltage, V; discharged ampere hours, q; time,
t; and q', ampere hours discharged at a time, t', that is
earlier than the model calculation. The BEST model
equations are
V= E-

IRo

E = Eo - A X / Q

(24)
- MX/(Q

- X)

(25)

X = q + ( D l q + (1 - D ) ( I Q ) ) / I o
(IO)

( q - q ' ) / ( t - t')dq'.

KibaM
1.1

.1KiBaM

=
Q.
w

100

,.0

>
0
OC

(27)

In addition to ( I Q ) (the forward averaged current),


the other constants are: E0 (voltage for zero current
limit), Qo (Ah capacity limit at zero current), I0 (factor
affecting decrease of capacity with current), A (initial
slope of voltage-Ah curve, normalized), M (constant

200

.=

(26)

10

20

Recovery Period, Hrs


Fig. 5. Comparison of recovery from the Battery Energy Storage Test (BEST) and kinetic battery model
(KiBaM) models.

404

J.F. MANWELLand J. G. McGOWAN


14
KiBaM
12

10
l/)

>

100

200

300

Amp Hrs
Fig. 6. Comparison of voltage as a function of discharged ampere hours.
related to end-of-discharge voltage drop, typically
0.0116), D (dynamic response parameter, typically
0.31 ), and (IQ) (the forward averaged current). The
BEST constants were derived from a combination of
manufacturer's data, experiments undertaken at the
University of Massachusetts, and suggestions made by
Hyman et al.[5]. The numerical values for the other
constants (on a per-cell basis) were
R0 = 0.0013 ~2
Eo = 2.0762 V
Qo = 225 Ah
I0 = 6 0 A
A =0.1.
Battery capacity was estimated by finding the amp
hours that had been discharged by the time the internal
voltage had reached to 1.9167 V per cell ( 11.5 V for
the six cells), assuming varying constant current discharges. The KiBaM constants were found from the
20- and 50-A discharges to be k = 0.5821, c = 0.40168,
and qmax = 196.6 Ah. The internal resistance was assumed to be the same as that used in the BEST model
(0.0078 ~2 for the entire battery). For the purpose of
voltage comparisons, E0 was taken to be 12.457 V (6
2.0762 V) and Emi, was assumed to be 11.5 V.
Figure 4 is analogous to Fig. 3, in that it depicts
capacity versus discharge current. As can be seen, the
predictions are quite close. Figure 5 illustrates the effect
of a varying recovery period for two representative discharges. In each case the battery was modeled as discharging at a constant rate to the specified internal
voltage. Then the discharge rate was held to zero for
the recovery period, followed by another constant current discharge, until the minimum voltage was again

reached. As can be seen, both models predict a recovery


within a few hours. KiBaM predicts a somewhat greater
total recovery than does the BEST. At this point, it
cannot be determined which is closer to reality. Figure
6 illustrates discharge voltage comparisons for the two
models. The general trends for both models are similar
until the knee is reached, where the KiBaM is no longer
applicable. The major apparent difference is in the
curvature of the voltage drop curve from the two models: BEST is always concave downward, whereas the
opposite is true for KiBaM. Again, the significance of
the difference is not yet known.
5. CONCLUSIONS
A relatively simple battery model has been developed that reflects many of the observed properties of
deep-cycle lead acid batteries. These properties include
(a) apparent decrease in capacity at high charge or
discharge rates, (b) conservation of charge, (c) nonconservation energy, (d) capacity recovery, and (e)
terminal voltage variation with capacity and current.
The model involves the determination of relatively few
constants. For many situations of interest, only commonly available manufacturer's data may be needed.
The model is particularly well suited for application
in quasisteady simulation codes where charge or discharge rates may be assumed to be constant over the
time step. Future work to be carried out includes (a)
detailed evaluation of the range of applicability of the
model, (b) inclusion of temperature effects, (c) application to other than lead acid batteries, and (d) incorporation of this component model into other hybrid
renewable energy systems models (such as PV hybrid
systems).

Acknowledgment--This work was supported by U.S. Department of EnergyContract No. XK-7-07106-1,administered


by the National Renewable Energy Laboratory (NREL).

Lead acid battery storage model


REFERENCES

1. W. Stein, J. F. Manwell, J. G. McGowan and W. Q.


Jeffries, Further developments with the University of
Massachusetts wind/diesel system simulator, Proceedings
of A WEA Windpower 90 (1990).
2. J. F. Manwell and J. G. McGowan, Developments in
battery storage for wind/diesel systems, Proceedings of

Sixth Annual Battery Conference on Applications and


Advances, January 1991, Long Beach, CA, 1991.
3. J. F. Manwell, J. G. McGowan, W. Stein, A. Rogers, and
M. E. Jones, Development of quasi-steady analytical
models for wind/diesel systems, Proceedings of Ninth
ASME Wind Energy Symposium, ASME Publication
SED-Vol. 9, 223-230 (1990).

405

4. W. A. Facinelli, Modeling and simulation of lead-acid


batteries for photovoltaic systems, Proceedings of 18th
IECEC, 1983.
5. E. Hyman, W. C. Spindler, and J. F. Fatula, Phenomenological discharge voltage model for lead acid batteries,
Proceedings of AIChE Meeting, November (1986).
6. E. Hyman, Modeling and computerized characterization
of lead-acid battery discharges," BEST Facility Topical
Report RD 83-1, NTIS Report DOE/ET/29368-TI3
(1986).
7. J. F. Manwell, W. Stein, J. G. McGowan, and W. Q.
Jeffries, Pulse width modulated dump load control for
wind/diesel systems, Proceedings of EWEC '90 554
(1990).

Вам также может понравиться