Processes,
Journal Year:
2024,
Volume and Issue:
12(8), P. 1749 - 1749
Published: Aug. 20, 2024
Accurate
energy
consumption
prediction
is
crucial
for
addressing
scheduling
problems.
Traditional
machine
learning
models
often
struggle
with
small-scale
datasets
and
nonlinear
data
patterns.
To
address
these
challenges,
this
paper
proposes
a
hybrid
grey
model
based
on
stacked
LSTM
layers.
This
approach
leverages
neural
network
structures
to
enhance
feature
harnesses
the
strengths
of
in
handling
data.
The
trained
using
Adam
algorithm
parameter
optimization
facilitated
by
grid
search
algorithm.
We
use
latest
annual
coal,
electricity,
gasoline
Henan
Province
as
application
background.
model’s
performance
evaluated
against
nine
fifteen
four
metrics.
Our
results
show
that
proposed
achieves
smallest
errors
across
all
metrics
(RMSE,
MAE,
MAPE,
TIC,
U1,
U2)
compared
other
15
system
9
during
testing
phase,
indicating
higher
accuracy
stronger
generalization
performance.
Additionally,
study
investigates
impact
different
layers
performance,
concluding
while
increasing
number
initially
improves
too
many
lead
overfitting.
International Journal of Sustainable Development & World Ecology,
Journal Year:
2024,
Volume and Issue:
31(5), P. 523 - 536
Published: Jan. 7, 2024
China,
the
world's
largest
CO2
emitter,
has
pledged
to
reduce
its
carbon
intensity
by
18%
2025,
which
requires
accurate
forecasting
of
emissions
and
their
drivers.
However,
existing
gray
models
have
limitations
in
dealing
with
fluctuating
data
or
long-time
series
data,
they
often
suffer
from
overfitting
poor
generalization
ability.
Moreover,
there
is
a
lack
research
judgment
on
future
changes
emission
To
address
these
issues,
this
study
proposes
fractional
order
adaptive
rolling
model
(RFAGM(1,1))
that
optimizes
background
generation
incorporates
mechanism.
We
apply
RFAGM(1,1)
forecast
China's
emissions,
GDP,
population,
consumption
raw
coal,
crude
oil,
natural
gas
2020
2025.
Our
results
show
achieves
significantly
higher
accuracy
than
standard
models,
except
for
population.
The
projections
indicate
China
will
meet
reduction
target
Furthermore,
LMDI
decomposition
reveals
economic
growth
population
positive
cumulative
impacts
(245.68%
11.95%,
respectively),
while
energy
structural
negative
(−151.60%
−6.02%,
respectively).
improved
enables
evaluation
climate
policies,
factor
analysis
provides
valuable
insights
developing
evidence-based
strategies
achieve
peaking
neutrality
goals
2030/2060.
With
the
fast
development
of
renewable
energy,
a
large
amount
energy
is
integrated
into
power
system.
However,
intermittency
and
volatility
sources
such
as
solar
wind
may
pose
huge
challenges
to
system
scheduling.
In
order
reduce
impact
on
operation
improve
autonomy
This
paper
proposes
reinforcement
learning
scheduling
method
for
based
data-driven
pretraining.
Firstly,
it
utilizes
ge2e
encode
wind,
photovoltaic,
load,
performs
pre
training
obtain
embedding
vectors
different
sourcesThen,
are
used
input
features
state,
method-SAC.
able
keep
secure
real-time
with
highly
volatile
loads
energy.
