Sie sind auf Seite 1von 6

ENERGYCON 2014 May 13-16, 2014 Dubrovnik, Croatia

A New Model of Lead-Acid Batteries Lifetime


in Smart Grid Scenario
Roberto Langella, Alfredo Testa, Chiara Ventre
Department of Industrial and Information Engineering
Second University of Naples
via Roma n29 81031, Aversa, Italy
roberto.langella@unina2.it
alfredo.testa@unina2.it
chiara.ventre@studenti.unina2.it
Abstract A new model of lifetime of LA batteries in smart grid
scenario is proposed. The model falls in the category of "fatigue
models" and is able to take into account the detrimental effect of
cycling the battery at different values of final SOC. Its main
characteristic is that it is easy to implement and it only requires the
well known CF vs. DOD graphs that are usually given by the
manufacturers. Two case studies proposed by an EU project
demonstrate the accuracy of the proposed model despite its simplicity.
Keywords: Lifetime, Lead-Acid, Smart Grids.

NOMENCLATURE

Ny

Lead-Acid.
Cycles to Failure.
Depth Of Discharge.
Nominal Capacity.
State Of Charge.
Lifetime of the battery.
Effective Energy of the battery.
Number of Equivalent Cycles.
Battery Current and Voltage.
Total Fractional Damage.
Current referred to a discharge of 10 hours.
Remaining Capacity.
Number of cycles per year.
Number of discharge cycles per year for DODi.

I. INTRODUCTION
Modeling the lifetime of batteries is important because it is a
fundamental parameter which influence the total cost of the
system in which they are included during all its life cycle [1, 2].
In particular, this aspect is commonly considered one of the key
weak factors in the long-term operation of renewable based
Distributed Generation systems in a Smart Grid scenario.
Reference is made to the wear out of the battery that assumes
the end of lifetime reached once the capacity of the battery, under
standard test conditions, is below 80% of the nominal capacity.
Usually, the wear out of a battery is assumed once a
predetermined value of some factors, which can easily be
determined (Ah-throughput, number of cycles and manufacturing
age, ...), has been exceeded. A precise lifetime prediction is
difficult, because typically not all factors can be taken into
account. Nevertheless, quantitative lifetime models are a
necessary tool for the design of systems taking into account
lifetime cost and reliability issues.

978-1-4799-2449-3/14/$31.00 2014 IEEE

Ageing estimation methods have been developed in particular


for LA batteries. They can be grouped into two main categories:
physics-based and fatigue models or combination of them.
Physics-based models are intended to determine the impact of
operating conditions and strategies on the lifetime defining stress
factors on a physical or chemical basis. One of the most
exhaustive models is the "Weighted Ah-throughput" that was
developed in the framework of a EU Benchmarking Research
Project (ENK6-CT-2001-80576 - 2005) which combines the
model development with experimental verification, using both
specific lab tests of selected batteries as well as field test results
collected in the projects systems test database [3-5].
Fatigue models are basically heuristic and are based on a
mechanical view of the battery ageing assuming that operating
conditions impact in an incremental way, step by step, on the
battery lifetime. Those methods are mainly based on well
known CF vs. DOD graphs that are usually given by the
manufacturers. These graphs are based on well-defined test
conditions simply determined by discharging the battery with a
constant current to a certain depth of discharge and a
subsequent full charge with a given charging regime. Rainflow
method [6], UMass Battery model [7, 8] and Average Ahthroughput [9] are the most common reference models falling in
this category. They assume that cycling damage to a battery is
primarily a function of the depth of discharge to which the
battery is subjected. However, in real applications the operating
conditions typically deviate from these standard operating
conditions and the battery may be more or less damaging than
during the standard operating conditions. In addition, the
magnitude of the cycle was assumed to be more important than
the final SOC: the effect, for example, of going from 10 % to
30 % discharge and back was assumed to be approximately the
same as from going from 50% to 70%.
This paper proposes a new model of expected lifetime of LA
batteries in smart grid scenario falling in the category of "fatigue
models" and able to take into account the detrimental effect of
cycling the battery at different values of final SOC. Its main
characteristic is that it is easy to implement and it only requires
the well known CF vs. DOD graphs that are usually given by the
manufacturers. The results of two case studies proposed in [4]
demonstrate the accuracy of the proposed model despite its
simplicity. For both the case-studies, the results of the Weighted
Ah-throughput method [5] were chosen as reference.

1343

ENERGYCON 2014 May 13-16, 2014 Dubrovnik, Croatia

II. LITERATURE METHODS FOR LIFETIME ESTIMATION


A. Weighted Ah-throughput
A weighted Ah ageing model is based on the assumption
that the impact of a given Ah-throughput (Ah-T) on the
lifetime depends on the details of the conditions during the Ah
throughput. It is assumed that, under standard conditions, a
battery can achieve an overall Ah-T until the end of the
lifetime is reached [3]. Deviations from the standard
conditions result in a virtual increase (or decrease) of the
physical Ah-T. These deviations are taken into account so
determining that the battery is at the end of its lifetime once
the weighted Ah-T has exceeded the expected un-weighted
Ah-T measured under nominal operating conditions.
The major stress factors, except for the temperature, that
influence the lifetime of a lead-acid battery depend mainly on
the following phenomena:
Corrosion of the positive electrode leads to a reduction in
conductivity due to the formation of a corrosion layer between
the grid and the active mass;
Cycling at low SOC stresses the battery more than
cycling at high SOC;
Cycling the battery while acid stratification is present is
known to result in inhomogeneous current distribution along
the electrode;
Long periods without a full charge are known to be
detrimental because of sulphation and capacity loss.
The Ah-T needs to be weighted with factors that take into
account the stress effects.
During one simulation step, the voltage and the SOC are
determined on the basis of the battery parameters. The voltage
is calculated according to a modified Shepherd equation (as
shown in Appendix A) and the SOC is calculated by:

(1)

with .
Based on voltage and SOC, corrosion and degradation
parameters are determined and then used to change the battery
parameters and to determine the remaining capacity of the
battery, which is the main output of the model. The remaining
capacity, , is then the initial capacity, , minus the
capacity losses:
,

(2)

being the capacity loss caused by corrosion and


the capacity loss caused by degradation. Their
analytical expression are reported in Appendix A.
The functions used to represent the different effects are in
part of heuristic nature, but they are based on the knowledge
of real physical or chemical processes. The fact that the
parameters used in the equations are battery specific and
therefore differ depending on the battery type must be taken
into account and makes the model difficult to be applied for
technologies not considered in [3].

978-1-4799-2449-3/14/$31.00 2014 IEEE

B. Rainflow method
This method, which uses the Downings algorithm, is based
on the assumption that the number of cycles a battery can
tolerate is a function only of the depth of discharge [6].
To use the Rainflow cycle counting algorithm it is
necessary to convert the temporal sequence of the SOC in a
sequence of peaks and valleys as described in [6] and shown
in Fig. 1.
DODi

SOCin,i
SOCfin,i

i-th
cycle

Fig. 1. A generic SOC (black) and the input of the Rainflow algorithm (red).

The number of cycles per year corresponding to each of


the ranges of DOD is determined. For each interval, there
will be a number of Cycles to Failure ( ) obtained from the
manufacturer's curves reported in Fig. 2 and extended to
ranges (from 0 to 10% and from 80 to 100%) usually not
covered.

Fig. 2. Cycles to Failure vs. DOD [%] obtained by manufacturers.

The fraction of life used up during a given cycle,


corresponding to a given class, is 1/ . After sufficient cycles
have passed that the fractions multiplied by the number of
cycles reaches to 1.0, the battery is assumed to be spent.
Battery duration, in years, will then be
 

(3)

Several authors noticed [3] that this method is simple and


quick to implement. However, hypothesis made are, at least,
questionable: there is no interaction between several ageing
events, or an event will impact lifetime equally at the
beginning or at the end of life of the system.
C. UMass Battery model
The UMass battery model, also known as the Kinetic
Battery Model, consists of three separate parts: capacity,
voltage and lifetime. The three parts are physically based
structures with constants suitable determined from test data
and they are defined in reference [7] and [8].
The capacity model is based on the concept that the battery
capacity is distributed over two wells as in Fig. 3: the
available charge and the bound charge. A fraction of the
total capacity is put in the available charge well and a fraction
( ) in the bound charge well.

1344

ENERGYCON 2014 May 13-16, 2014 Dubrovnik, Croatia

CFeq

Fig. 4. Equivalent Full Cycles vs. DOD.

Fig. 3. Model of the battery charge in Kinetic Battery Model.

The available charge well, which is immediately accessible,


supplies the load whereas the bound-charge well supplies only
the available charge well at a rate which depends on the height
difference between the two wells, and on a parameter k. The
heights of the two wells are respectively: and
. When a load is applied to the battery, the
available charge is reduced, and the difference between the
two wells grows. When the load is removed, the electrons
flow from the bound-charge well to the available-charge well
until the two charges are equal.
The flows of available and bound charge during a constant
discharge or charge current are given by equations (4).

(4)

whose solution, with initial conditions and


, together with the voltage model give the
capacity model.
The battery is considered at the end of its life when there is no
charge left in the available charge well.
The battery lifetime model is based on empirical curves
obtained interpolating the points of Fig. 2:
.

III. PROPOSED METHOD


Reference is made to specific operating conditions
determined in the design stage where a reference value of the
depth of discharge,
, is chosen. The corresponding
, is taken from Fig. 2.
number of cycle to failure,
The proposed method is based on the concept that the
fraction of life used up during the i-th cycle is calculated as
the ratio of the energy delivered to the load, , to the energy
that would be delivered during a complete reference cycle,

, multiplied by a stress factor, , introduced in order


to take into account the effects of the final at the end of a
given cycle.
The degradation of the life during the i-th cycle, Di,
expressed as fraction of number of cycles, will be:

 (8)

is function of the SOC value at the end of the cycle,


(see also Fig. 1). Information about can be
derived again from the manufacturers' curves of Fig. 2; the
test performed, for different values, always start from
and end for . Fig. 5
reports the Cycles to Failure vs. .

(5)

The Rainflow algorithm is used to count the charge/discharge


cycles and equation (3) is used to predict the battery life.
D. Average Ah-throughput
Fig. 5. Cycles to Failure vs. SOCfin [%] obtained by manufacturers curves.
The simplest of all the models for battery lifetime
It is assumed that the stress effect of a given cycle is in
prediction in literature is the Average Ah-throughput recalled
:
inverse proportion to the ratio of to
in reference [8].

.
(9)

Starting from Fig. 2, it is possible to define, for each range i,

the corresponding equivalent full cycles:

(6)

Once known the number of cycles per year of the battery and
the average value of the equivalent full cycles, , (Fig. 4),
the expected lifetime, in years, is given by equation (7).

978-1-4799-2449-3/14/$31.00 2014 IEEE

(7)

Fig. 6 reports the stress factor versus for different


choices of
. For a standard cycle, the weighting-factor
will be 1. On the contrary, different operating
conditions depth of discharge bigger than
will be
weighted with a factor greater than 1. Instead, operating
conditions less severe depth of discharge smaller than

will be weighted with a factor lower than 1.

1345

ENERGYCON 2014 May 13-16, 2014 Dubrovnik, Croatia

A series of 35 PV profile;
A discharge for one hour to reach SOC=50%;
A series of 35 PV profile;
A charge at for three hours to reach SOC=80%;
A series of 35 PV profile.
PV Profile

.
Fig. 6. Stress factor versus for different choices of

The procedure proposed to estimate the battery lifetime


starts from the preliminary analysis of the temporal sequence
of SOC, so that charge/discharge cycles are singled out for
the reference year. Then, for each cycle i the following
sequential steps are repeated:
1) calculate the energy as:
;
2) the ratio

(10)

is determined given that:

(11)

3) is calculated from Fig. 6;


4) is evaluated by means of (8);

 

Fig. 8. SOC with PV profile.

(12)

IV. CASE STUDIES


In this section two case studies are presented. The behavior
of a OPzS battery (flooded electrolyte, where the positive
electrode is a tubular-plate and the negative is a flat-plane
electrode) with  was simulated using the software
MATLAB. The parameters used are reported in Appendix B
[5]. The battery was subjected to two current profiles that
simulate its operation in the presence of a photovoltaic system
in the first case and of a wind generator in the latter case. The
test profiles chosen were taken from the experimental activity
reported in [4]. For both the case-studies, the results of the
Weighted Ah-throughput method, verified also experimentally,
were chosen as reference; in particular, the results reported in
[5] that contain an enhanced version of the Weighted Ahthroughput method are considered.
A. PV Profile Test
The test current profile represents the operating conditions
of a battery used in a PV system. Fig. 7 reports an extract
from the PV current profile versus the time while Fig. 8 the
corresponding SOC time profile determined using (1).
Batteries were first subjected to five discharge/charge
cycles to fully charge and discharge the battery. The tests then
consisted of blocks of repeated PV profile interspersed with
fully charge/discharge cycles (capacity test) needed to check
the available capacity during the experimental verification
conducted in [4]. The PV blocks are constructed as it follows:

Fig. 7. PV current profile.

A discharge at for two hours to reach SOC=80%;

978-1-4799-2449-3/14/$31.00 2014 IEEE

Using a cycle counting algorithm it is possible to count the


different discharge/charge cycles to which the battery is
subjected and then represent their year frequency with the
histogram of Fig. 9. It is possible to observe that the majority
of the cycles have DOD=30%, while the weighted average
was calculated as 31%. As a consequence, it was assumed

.

Fig. 9. Number of cycles per year vs. DOD [%] in the PV operation profile.

Table I shows the expected lifetime estimated using the


methods described in Section II and III together with an
heuristic indication about the overall "complexity" of the
methods.
TABLE I
ESTIMATED LIFETIMES FOR PV PROFILE OPERATION

Methods
Weighted Ah-throughput [5]
Rainflow method
UMass Battery model [4]
Average Ah-throughput
Proposed method

Lifetime (years)
1.63 (*)
2.78
2.05
0.80
1.61

Complexity
+++++
++
++++
+
++

(*) The results of the experimental validation are reported in [5].

1346

ENERGYCON 2014 May 13-16, 2014 Dubrovnik, Croatia

It is possible to observe that:

The Rainflow method overestimates of about 70% the


reference lifetime of 1.63 years;
The UMass battery model overestimates of about 25%
the reference lifetime;
The Average Ah-throughput method underestimates of
about 50% the reference lifetime;
The proposed method approximate the reference method
with an error of -1.5% and its complexity is very lower.

B. Wind Profile Test


The test current profile represents the operating conditions of
a battery used in a wind system. Fig. 10 reports an extract
from the Wind current profile versus the time while Fig. 11
the corresponding SOC time profile.
The batteries were first subjected to five
discharge/charge cycles to fully charge and discharge the
battery. The tests then consisted of blocks of repeated wind
profile interspersed with fully charge/discharge cycles.
The wind blocks are constructed following the instructions
from reference [4] as it follows:
A discharge at for one hour to reach SOC=90%;
A series of 50 wind profile;

The UMass battery model underestimates of about 25%


the reference lifetime;
The Average Ah-throughput method underestimates of
about 28% the reference lifetime;
The proposed method approximates the reference
method with an error of -4.1%.

Fig. 12. Number of cycles per year vs. DOD [%] in the wind operation
profile.
TABLE II
ESTIMATED LIFETIMES FOR WIND PROFILE OPERATION

Methods
Weighted Ah-throughput [5]
Rainflow method
UMass Battery model [4]
Average Ah-throughput
Proposed method

Wind Profile

Lifetime (years)
1.33 (*)
2.12
1.00
0.95
1.27

Complexity
+++++
++
++++
+
++

(*) The results of the experimental validation are reported in [5].

Fig. 10. Wind current profile.

Fig. 11. SOC with wind profile.

Again, using a cycle counting algorithm it is possible to


count the discharge/charge cycles to which the battery is
subjected and then represent them with the histogram in Fig.
12. It is possible to observe that the majority of the cycles
have DOD=10% and DOD=80% while the weighted average
was calculated as 47.8%. As a consequence, it was assumed

.
Table II is the equivalent of Table I.
It is possible to observe that:

V. CONCLUSION
A new model of expected lifetime of LA batteries in smart
grid scenario has been proposed. The model is able to take
into account the detrimental effects of cycling the battery at
different values of final SOC. Its main characteristic is that it is
easy to implement and it only requires the well known
CF vs. DOD graphs that are usually given by the manufacturers.
Two case studies have demonstrated the accuracy of the
proposed model despite its simplicity.
The simplicity of the model which needs only the
knowledge of the CF vs DOD curves makes it possible to
extend it at different battery technologies.
Finally, it seems possible to include the model in an
adaptive control system implemented in the battery interface
converter for real time monitoring of the health of the battery
to determine the optimal operating conditions and to plan the
replacement intervals. This feature seems very attractive in the
new Smart Grid scenario where the use of energy storage
systems to optimize distributed generation energy production
is considered.
ACKNOWLEDGMENT
This work was supported by the Italian Ministry of
University and Research under Grant PON03PE_00159_6.

The Rainflow method overestimates of about 60% the


reference lifetime of 1.33 years;

978-1-4799-2449-3/14/$31.00 2014 IEEE

1347

ENERGYCON 2014 May 13-16, 2014 Dubrovnik, Croatia

REFERENCES

[1] V. Carpentiero, R. Langella, A. Testa, Hybrid wind-diesel stand-alone


system sizing accounting for component expected life and fuel price
uncertainty, Electric Power Systems Research, Vol. 88, pp. 69-77, 2012.
[2] R. Langella, A. Testa, Analysis of the Cost of Electric Energy Discharged
to the Grid by Some Energy Storage Systems, in IEEE Energycon 2012,
Florence, Italy, September 2012.
[3] J. Schiffer, D. U. Sauer, H. Bindner, T. Cronin, P. Lundsager and R. Kaiser,
Model prediction for ranking lead-acid batteries according to expected
lifetime in renewable energy systems and autonomous power-supply
systems, Journal of Power Sources, 168, pp. 66-78, 2007.
[4] H. Bindner, T. Cronin, P. Lundsager, J. F. Manwell, U. Abdulwahid and I.
Baring-Gould, Lifetime modelling of lead acid batteries, Ris National
Laboratory, Roskilde, Denmark, Ris-R Report, 2005.
[5] A. Andersson, Battery lifetime modelling, Ume University, Sweden,
Ris National Laboratory, Roskilde, Denmark, Jan. 2006.
[6] S. D. Downing and D. F. Socie, Simple Rainflow counting algorithms,
International Journal of Fatigue, Volume 4, 31-40, January 1982.
[7] M. R. Jongerden, B. R. Haverkort, Which battery model to use?, UK-PEW
2008, Tech. Report DComp., Imperial College London, pp. 76-88, 2008.
[8] J. F. Manwell, A. Rogers, G. Hayman, C. T. Avelar, J. G. McGowan, U.
Abdulwahid and K. Wu, Hybrid2 A hybrid system simulation model.
Theory manual, Renewable Energy Research Laboratory, University of
Massachusetts, 2006.
[9] R. Dufo-Lpez and J. L. Bernal-Augustn, Influence of mathematical
models in design of PV-Diesel systems, Energy Conversion and
Management, 49, pp. 820-831, 2008.

APPENDIX A
The state of the battery, represented by its voltage is
calculated with a modified Shepherd equation (A.1):

where

where ,

The capacity reduction due to the degradation of the active


mass is given by equation (A.9)

where

(A.1)

(A.2)

(A.3)

The capacity reduction due to the corrosion is so obtained:


(A.4)

where and .
The weighting factor for SOC is calculated from (A.5):

978-1-4799-2449-3/14/$31.00 2014 IEEE

(A.9)

and .

TABLE III - BATTERY PARAMETERS

(A.8)

APPENDIX B

 

 

(A.7)

The dimension of the corrosion layer between the grid and the
active mass is evaluated by (A.3):

Finally, the Ah are weighted with the SOC factor and the
acid stratification factor as in equation (A.8):



and


 

The gassing current is determined by (A.7):

The corrosion of the positive electrode is obtained by (A.2):




 




 

and
.

The acid degradation factor depends on gassing process and


bad charges:

(A.6)

(A.5)

1348

Par.s

Description

Value

SOC (0)

g
/
/

Nominal Voltage
Nominal Capacity
Reference Voltage
Initial State of Charge
Initial Capacity
Open Circuit Voltage
Open Circuit Corrosion Voltage
Variation coeff. of with SOC
Internal Resistance (Ah)
Variation coeff. of the resistance
with SOC
Normalized Capacity
Gassing Cur. @ and 298K
Nominal Gassing Voltage
Voltage Coefficient for
Temperature Coeff. for
Corrosion Rate
Lifetime in years
SOC Weight-factor with SOC=0
SOC Weight-factor with
Coefficient for
IEC Number of Cycles
Increment (Decrement) Factor
for Acid Stratification

2V
50 Ah
2.3 V
100 %
86.2 Ah
2.1 V
1.716 V
0.054 V
0.43609/0.37885
0.36488/0.28957

1.001/1.642
0.02 A
2.23 V
5 V-1
0.06 k-1
0.0435
12
8
3
5
1200
1/30 (0.02)

Das könnte Ihnen auch gefallen