Journal of The Electrochemical Society,
Journal Year:
2024,
Volume and Issue:
171(11), P. 110528 - 110528
Published: Nov. 1, 2024
To
accurately
predict
the
state
of
health
(SOH)
lithium-ion
batteries
and
improve
safety
reliability
battery
management
systems,
a
new
SOH
estimation
method
based
on
fusion
features
(HFs)
adaptive
boosting
integrated
grey
wolf
optimizer
to
optimize
back
propagation
neural
network
(Adaboost-GWO-BP)
is
proposed.
First,
five
kinds
multi-type
HFs
were
extracted
from
charging
process,
correlation
between
proposed
was
verified
by
Pearson
Spearman
coefficients.
Then,
indirect
feature
(IHF)
obtained
multidimensional
scaling
dimensionality
reduction
reduce
data
redundancy
SOH.
The
GWO-BP
model
then
used
establish
nonlinear
mapping
relationship
IHF
In
order
overcome
problem
low
accuracy
in
single
model,
Adaboost
algorithm
ensemble
learning
introduced
enhance
estimation.
Finally,
NASA
dataset,
compared
with
other
models.
comparative
experiments,
mean
absolute
error
root
square
for
less
than
0.81%
1.26%,
which
has
higher
Batteries,
Journal Year:
2024,
Volume and Issue:
10(12), P. 433 - 433
Published: Dec. 6, 2024
Accurate
assessment
of
battery
State
Health
(SOH)
is
crucial
for
the
safe
and
efficient
operation
electric
vehicles
(EVs),
which
play
a
significant
role
in
reducing
reliance
on
non-renewable
energy
sources.
This
study
introduces
novel
SOH
estimation
method
combining
Kolmogorov–Arnold
Networks
(KAN)
Long
Short-Term
Memory
(LSTM)
networks.
The
based
fully
charged
characteristics,
extracting
key
parameters
such
as
voltage,
temperature,
charging
data
collected
during
cycles.
Validation
was
conducted
under
temperature
range
10
°C
to
30
different
charge–discharge
current
rates.
Notably,
variations
were
primarily
caused
by
seasonal
changes,
enabling
experiments
more
realistically
simulate
battery’s
performance
real-world
applications.
By
enhancing
dynamic
modeling
capabilities
capturing
long-term
temporal
associations,
experimental
results
demonstrate
that
achieves
highly
accurate
various
conditions,
with
low
mean
absolute
error
(MAE)
root
square
(RMSE)
values
coefficient
determination
(R2)
exceeding
97%,
significantly
improving
prediction
accuracy
efficiency.
