KAN–CNN: A Novel Framework for Electric Vehicle Load Forecasting with Enhanced Engineering Applicability and Simplified Neural Network Tuning
Electronics,
Год журнала:
2025,
Номер
14(3), С. 414 - 414
Опубликована: Янв. 21, 2025
Electric
Vehicle
(EV)
load
forecasting
is
critical
for
optimizing
resource
allocation
and
ensuring
the
stability
of
modern
energy
systems.
However,
traditional
machine
learning
models,
predominantly
based
on
Multi-Layer
Perceptrons
(MLPs),
encounter
substantial
challenges
in
modeling
complex,
nonlinear,
dynamic
patterns
inherent
EV
charging
data,
often
leading
to
overfitting
high
computational
costs.
To
overcome
these
limitations,
this
study
introduces
KAN–CNN,
a
novel
hybrid
architecture
that
integrates
Kolmogorov–Arnold
Networks
(KANs)
into
frameworks,
specifically
Convolutional
Neural
(CNNs).
By
combining
spatial
feature
extraction
strength
CNNs
with
adaptive
nonlinearity
KAN,
KAN–CNN
achieves
superior
representation
flexibility.
The
key
innovations
include
bottleneck
KAN
convolutional
layers
reducing
parameter
complexity,
Self-Attention
Network
Global
Nonlinearity
(Self-KAGN)
Attention
enhance
global
dependency
modeling,
Focal
KAGN
Modulation
refinement.
Furthermore,
regularization
techniques
such
as
L1/L2
penalties,
dropout,
Gaussian
noise
injection
are
utilized
model’s
robustness
generalization
capability.
When
applied
forecasting,
demonstrates
prediction
accuracy
comparable
state-of-the-art
methods
while
significantly
overhead
simplifying
tuning.
This
work
bridges
gap
between
theoretical
practical
applications,
offering
robust
efficient
solution
system
challenges.
Язык: Английский
SOH-KLSTM: A hybrid Kolmogorov-Arnold Network and LSTM model for enhanced Lithium-ion battery Health Monitoring
Journal of Energy Storage,
Год журнала:
2025,
Номер
122, С. 116541 - 116541
Опубликована: Апрель 15, 2025
Язык: Английский
Advancing Real-Estate Forecasting: A Novel Approach Using Kolmogorov–Arnold Networks
Algorithms,
Год журнала:
2025,
Номер
18(2), С. 93 - 93
Опубликована: Фев. 7, 2025
Accurately
estimating
house
values
is
a
critical
challenge
for
real-estate
stakeholders,
including
homeowners,
buyers,
sellers,
agents,
and
policymakers.
This
study
introduces
novel
approach
to
this
problem
using
Kolmogorov–Arnold
networks
(KANs),
type
of
neural
network
based
on
the
theorem.
The
proposed
KAN
model
was
tested
two
datasets
demonstrated
superior
performance
compared
existing
state-of-the-art
methods
predicting
prices.
By
delivering
more
precise
price
forecasts,
supports
improved
decision-making
stakeholders.
Additionally,
results
highlight
broader
potential
KANs
addressing
complex
prediction
tasks
in
data
science.
aims
provide
an
innovative
effective
solution
accurate
estimation,
offering
significant
benefits
industry
beyond.
Язык: Английский