2022 IEEE International Conference on Environment and Electrical Engineering and 2022 IEEE Industrial and Commercial Power Systems Europe (EEEIC / I&CPS Europe),
Journal Year:
2023,
Volume and Issue:
unknown, P. 1 - 6
Published: June 6, 2023
Wind
power
forecasting
is
a
crucial
aspect
of
re-newable
energy
production,
as
it
helps
to
optimize
output
and
ensure
grid
stability.
In
recent
years,
Transformer-based
language
models
such
ChatGPT
have
been
successfully
used
in
natural
processing
tasks,
but
their
application
wind
remains
largely
unexplored.
this
article,
we
propose
using
Transformer
model,
the
core
ChatGPT,
improve
accuracy
forecasting.
Using
self-attention
mechanism,
developed
model
can
capture
complex
temporal
relationships
large-scale
time
series
data.
Furthermore,
proposed
method
evaluated
on
test
set
various
performance
metrics.
Results
show
that
our
outperforms
traditional
models,
achieving
higher
accuracy.
Our
findings
suggest
significant
potential
for
improving
ultimately
contributing
more
sustainable
future.
Energy Engineering,
Journal Year:
2024,
Volume and Issue:
121(2), P. 359 - 376
Published: Jan. 1, 2024
The
fluctuation
of
wind
power
affects
the
operating
safety
and
consumption
electric
grid
restricts
connection
on
a
large
scale.
Therefore,
forecasting
plays
key
role
in
improving
economic
benefits
grid.
This
paper
proposes
predicting
method
based
convolutional
graph
attention
deep
neural
network
with
multi-wind
farm
data.
Based
mechanism,
extracts
spatial-temporal
characteristics
from
data
multiple
farms.
Then,
combined
network,
model
is
constructed.
Finally,
trained
quantile
regression
loss
function
to
achieve
deterministic
probabilistic
prediction
A
dataset
U.S.
taken
as
an
example
demonstrate
efficacy
proposed
model.
Compared
selected
baseline
methods,
achieves
best
performance.
point
errors
(i.e.,
root
mean
square
error
(RMSE)
normalized
absolute
percentage
(NMAPE))
are
0.304
MW
1.177%,
respectively.
And
comprehensive
performance
continuously
ranked
probability
score
(CRPS))
0.580.
Thus,
significance
feature
extraction
module
self-evident.
Energies,
Journal Year:
2025,
Volume and Issue:
18(7), P. 1849 - 1849
Published: April 6, 2025
Accurate
wind
power
forecasting
(WPF)
is
crucial
to
enhance
availability
and
reap
the
benefits
of
integration
into
grids.
The
time
lag
generation
lags
speed
changes,
especially
in
ultra-short-term
forecasting.
prediction
model
sensitive
outliers
sudden
changes
input
historical
meteorological
data,
which
may
significantly
affect
robustness
WPF
model.
To
address
this
issue,
paper
proposes
a
novel
hybrid
machine
learning
for
highly
accurate
raw
data
were
filtered
classified
with
local
outlier
factor
(LOF)
voting
tree
(VT)
obtain
subset
inputs
best
relevance.
time-varying
properties
fluctuating
sub-signals
sequences
analyzed
optimized
variational
mode
decomposition
(OVMD)
algorithm.
Northern
Goshawk
optimization
(NGO)
algorithm
was
improved
by
incorporating
logical
chaotic
initialization
strategy
adaptive
inertia
weights.
NGO
used
optimize
least
squares
support
vector
regression
(LSSVR)
improve
computational
results.
proposed
compared
traditional
models,
deep
other
models.
experimental
results
show
that
has
an
average
R2
0.9998.
MSE,
MAE,
MAPE
are
as
low
0.0244,
0.1073,
0.3587,
displayed
WPF.
