Solar
energy
plays
an
important
role
in
the
future
system.
However,
inherent
uncertainty
of
solar
brings
great
difficulties
to
grid
connection
and
short-term
planning
dispatching.
Deep
learning
method
makes
it
possible
predict
with
its
powerful
ability,
but
huge
training
process
parameter
adjustment
bring
actual
deployment.
Therefore,
this
paper
proposes
a
new
lightweight
multi-modal
model
for
irradiance
prediction
based
on
knowledge
distillation
strategy,
which
greatly
reduces
complexity
while
ensuring
acceptable
accuracy,
facilitating
Firstly,
teacher
inputs
Informer
framework
is
built
guide
student
model.
Then,
constructed
obtain
same
input
reduced
trainable
parameters.
The
optimal
settings
loss
function
ratio
are
studied.
Results
show
that
can
reduce
parameters
inference
time
by
97.7%
52.5%,
respectively.
normalized
root
mean
square
error
24.87%
compared
without
distillation,
verifying
effectiveness
proposed
method.
soft
uses
loss,
0.3,
best
results
structure
3
residual
blocks
LSTM
layers
proved
be
task.
Energy Science & Engineering,
Journal Year:
2025,
Volume and Issue:
13(5), P. 2220 - 2230
Published: March 3, 2025
ABSTRACT
Reliable
and
accurate
predictions
of
solar
radiation
are
essential
for
the
supervision
operation
photovoltaic
power
generation
systems.
As
primary
media
involved
in
atmospheric
transfer,
aerosols
significantly
influence
global
horizontal
irradiance
(GHI).
The
composition,
shape,
number
density
distribution
vary
greatly,
resulting
significant
differences
their
optical
properties,
which
turn
affect
different
ways.
This
study
aims
to
explore
impact
types
on
predicting
GHI.
First,
we
expanded
data
within
a
fixed
region
by
incorporating
spatial
information
supplement
timescale
data.
Furthermore,
used
Informer
model
forecast
GHI
regions,
inputting
historical
aerosol
depth
(AOD),
meteorological
parameters,
Finally,
an
classification
classify
regions
calculated
types.
findings
suggest
that
impacts
predictive
performance
When
continental
subcontinental
dominated,
improved.
biomass‐burning
dominate,
accuracy
reduced.
Journal of Renewable and Sustainable Energy,
Journal Year:
2025,
Volume and Issue:
17(2)
Published: March 1, 2025
Developing
and
using
solar
energy
has
become
an
important
strategic
decision
for
sustainable
development
in
many
countries.
Short-term
changes
irradiance
can
affect
the
safety
stability
of
photovoltaic
thermal
power
plants,
so
accuracy
prediction
attracted
significant
attention.
This
paper
proposes
a
short-term
method
based
on
improved
complete
ensemble
empirical
mode
decomposition
with
adaptive
noise
partial
differential
equation
model.
Image
feature
information
is
obtained
from
ground-based
sky
images,
two
ordinary
(ODE)
networks
are
used
to
process
historical
exogenous
variables,
including
meteorological
images
information.
Using
ODE
solver,
temporal
pattern
target
sequence
serial
correlation
between
variables
obtained,
model
multivariate
time
series
established.
The
proposed
evaluated
public
dataset
California,
USA,
locally
collected
datasets.
experimental
results
show
that
high
significantly
improves
estimation
irradiance.
Journal of Renewable and Sustainable Energy,
Journal Year:
2025,
Volume and Issue:
17(2)
Published: March 1, 2025
The
precise
forecasting
of
photovoltaic
energy
generation
holds
paramount
importance
in
refining
scheduling
and
ensuring
safe
operation
extensive
power
stations.
However,
the
inherent
instability
volatility
pose
significant
challenges
to
prediction
accuracy.
To
address
this,
this
article
conducts
a
thorough
analysis
seasonal
characteristics
introduces
hybrid
model
based
on
ensemble
empirical
mode
decomposition
(EEMD)-improved
whale
optimization
algorithm
(IWOA)-bidirectional
long
short-term
memory
network
(BiLSTM)
algorithm.
This
leverages
multi-seasonal
meteorological
features
enhance
First,
EEMD
is
used
decompose
reconstruct
data
eliminate
its
volatility.
Second,
three
improved
strategies
are
proposed
for
position
update
different
stages
IWOA,
IWOA-optimized
Bidirectional
LSTM
established.
Finally,
operational
station
northwest
region
China
as
case
study
evaluate
performance
detail.
results
show
that
model's
accuracy
rate
ranges
from
97.1%
98.7%,
which
can
accurately
predict
improve
utilization
renewable
energy.
