Electronics,
Год журнала:
2023,
Номер
12(21), С. 4443 - 4443
Опубликована: Окт. 29, 2023
Global
warming-induced
extreme
tropical
storms
disrupt
the
operation
of
offshore
wind
farms,
causing
power
ramp
events
and
threatening
safety
interconnected
onshore
grid.
In
order
to
attenuate
impact
these
ramps,
this
paper
proposes
an
integrated
strategy
for
forecasting
controlling
ramps
in
farms.
First,
characteristics
during
are
studied,
a
general
control
framework
is
established.
Second,
prediction
scheme
designed
based
on
minimal
gated
memory
network
(MGMN).
Third,
by
taking
into
account
results
uncertainties,
chance-constraint
programming-based
optimal
developed
simultaneously
maximize
absorption
minimize
costs.
Finally,
we
use
real-world
farm
data
validate
effectiveness
proposed
strategy.
Energies,
Год журнала:
2025,
Номер
18(3), С. 580 - 580
Опубликована: Янв. 26, 2025
Accurate
wind
power
forecasting
is
crucial
for
optimizing
grid
scheduling
and
improving
utilization.
However,
real-world
time
series
exhibit
dynamic
statistical
properties,
such
as
changing
mean
variance
over
time,
which
make
it
difficult
models
to
apply
observed
patterns
from
the
past
future.
Additionally,
execution
speed
high
computational
resource
demands
of
complex
prediction
them
deploy
on
edge
computing
nodes
farms.
To
address
these
issues,
this
paper
explores
potential
linear
constructs
NFLM,
a
linear,
lightweight,
short-term
model
that
more
adapted
characteristics
data.
The
captures
both
long-term
sequence
variations
through
continuous
interval
sampling.
mitigate
interference
features,
we
propose
normalization
feature
learning
block
(NFLBlock)
core
component
NFLM
processing
sequences.
This
module
normalizes
input
data
uses
stacked
multilayer
perceptron
extract
cross-temporal
cross-dimensional
dependencies.
Experiments
with
two
real
farms
in
Guangxi,
China,
showed
compared
other
advanced
methods,
MSE
24-step
ahead
respectively
reduced
by
23.88%
21.03%,
floating-point
operations
(FLOPs)
parameter
count
only
require
36.366
M
0.59
M,
respectively.
results
show
can
achieve
good
accuracy
fewer
resources.
IET Renewable Power Generation,
Год журнала:
2024,
Номер
unknown
Опубликована: Фев. 23, 2024
Abstract
The
growing
integration
of
renewable
energy
sources
into
the
power
grid
has
introduced
unprecedented
uncertainty.
Ensuring
an
appropriately
scheduled
reserve
is
essential
to
accommodate
energy's
intermittent
and
volatile
nature.
This
study
introduces
innovative
approach
ultra‐short‐term
wind
forecasting,
which
relies
on
feature
engineering
a
hybrid
model.
effectiveness
this
proposed
method
showcased
through
case
involving
utility‐scale
farm
in
Inner
Mongolia,
China.
findings
indicate
that
model,
combines
XGBoost
(Extreme
Gradient
Boosting)
algorithm
LSTM
(Long
Short‐Term
Memory)
network
with
KDJ
(Stochastic
Oscillator),
MACD
(Moving
Average
Convergence
Divergence),
achieves
highest
forecasting
accuracy.
Specifically,
model
yields
normalized
mean
absolute
error
0.0396
for
forecasting.
modelling
process
takes
approximately
550
s.
Furthermore,
suggested
employed
predict
speed
USA.
experimental
results
consistently
maintains
dependable
performance
across
various
raw
datasets;
it
suitable
use
system
operations.
