Building Services Engineering Research and Technology,
Год журнала:
2024,
Номер
45(6), С. 775 - 794
Опубликована: Авг. 18, 2024
The
regulation
strategy
of
a
district
heating
system
is
adjusted
based
on
accurate
heat
load
prediction,
which
not
only
effectively
reduces
energy
consumption
but
also
improves
efficiency
and
user
comfort.
In
order
to
improve
the
accuracy
forecasting,
forecasting
model
considering
two-dimensional
change
time
series
introduced
in
this
paper.
Firstly,
original
data
denoised
by
SVMD
decomposition,
several
stationary
regular
modal
components
are
obtained.
Then,
three
strategies
were
used
enhance
BWO
algorithm,
IBWO-TimesNet
prediction
was
established
extract
hidden
information
from
perspective.
Finally,
performance
evaluated
detail
through
case
analysis.
results
show
that
MAE,
RMSE
R2
SVMD-IBWO-TimesNet
0.647,
1.190
99.1%,
respectively.
Compared
with
other
mainstream
models,
has
higher
accuracy.
addition,
even
if
training
samples
reduced,
can
still
predict
strong
generalization
ability.
Therefore,
verified,
provides
reference
for
control
load.
Practical
application
Heat
vital
task
particularly
relation
its
impact
management
building
efficiency.
contribution
paper
provide
advanced
algorithms
This
insight
derived
modelling
will
assist
professionals
pursuit
more
needs
buildings,
thereby
optimizing
design
operation
systems.
practical
technology
could
save
costs,
reduce
carbon
emissions,
comfort
sustainability
buildings.
Sustainability,
Год журнала:
2024,
Номер
16(6), С. 2522 - 2522
Опубликована: Март 19, 2024
Accurately
predicting
the
cold
load
of
industrial
buildings
is
a
crucial
step
in
establishing
an
energy
consumption
management
system
for
constructions,
which
plays
significant
role
advancing
sustainable
development.
However,
due
to
diverse
influencing
factors
and
complex
nonlinear
patterns
exhibited
by
data
buildings,
these
loads
poses
challenges.
This
study
proposes
hybrid
prediction
approach
combining
Improved
Snake
Optimization
Algorithm
(ISOA),
Variational
Mode
Decomposition
(VMD),
random
forest
(RF),
BiLSTM-attention.
Initially,
ISOA
optimizes
parameters
VMD
method,
obtaining
best
decomposition
results
data.
Subsequently,
RF
employed
predict
components
with
higher
frequencies,
while
BiLSTM-attention
utilized
lower
frequencies.
The
final
are
obtained
predictions.
proposed
method
validated
using
actual
from
building,
experimental
demonstrate
its
excellent
predictive
performance,
making
it
more
suitable
constructions
compared
traditional
methods.
By
enhancing
accuracy
not
only
improves
efficiency
but
also
promotes
reduction
carbon
emissions,
thus
contributing
development
sector.
IEEE Internet of Things Journal,
Год журнала:
2024,
Номер
11(17), С. 28307 - 28319
Опубликована: Май 15, 2024
The
evolution
of
Internet-of-Things
(IoT)
is
fostering
the
use
intelligent
controls
for
energy
conservation.
Yet,
efficacy
these
strategies
largely
tied
to
diverse
load
forecasting
algorithms.
Given
significant
contribution
heating,
ventilation,
and
air-conditioning
(HVAC)
systems
global
consumption,
accurate
HVAC
power
usage
crucial
improving
overall
efficiency.
However,
real-world
forecasting,
bolstered
by
various
IoT
devices,
complicated
multiple
factors:
data
variability,
fluctuations,
electronic
phenomena
(e.g.,
zero
drifts),
increased
time
complexity
larger
model
sizes
required
manage
accumulating
historical
data.
To
address
challenges,
we
first
present
an
in-depth
measurement
study
on
characteristics
at
a
minute
scale
based
collected
in
six
locations.
We
propose
HALO,
transformer-based
framework
specifically
designed
load.
HALO
incorporates
adaptive
pre-processing
stage
local-global-scale
stage,
enabling
precise
optimization
utilization.
Evaluation
traces
from
prototype
application
demonstrates
that
proposed
significantly
outperforms
existing
models.
Future Internet,
Год журнала:
2024,
Номер
16(6), С. 192 - 192
Опубликована: Май 31, 2024
Accurate
short-term
load
forecasting
(STLF)
plays
an
essential
role
in
sustainable
energy
development.
Specifically,
companies
can
efficiently
plan
and
manage
their
generation
capacity,
lessening
resource
wastage
promoting
the
overall
efficiency
of
power
utilization.
However,
existing
models
cannot
accurately
capture
nonlinear
features
electricity
data,
leading
to
a
decline
performance.
To
relieve
this
issue,
paper
designs
innovative
method,
named
Prophet–CEEMDAN–ARBiLSTM,
which
consists
Prophet,
Complete
Ensemble
Empirical
Mode
Decomposition
with
Adaptive
Noise
(CEEMDAN),
residual
Bidirectional
Long
Short-Term
Memory
(BiLSTM)
network.
firstly
employs
Prophet
method
learn
cyclic
trend
from
input
aiming
discern
influence
these
on
load.
Then,
adopts
CEEMDAN
decompose
series
yield
components
distinct
modalities.
In
end,
advanced
BiLSTM
(ARBiLSTM)
block
as
above
extracted
obtain
results.
By
conducting
multiple
experiments
New
England
public
dataset,
it
demonstrates
that
Prophet–CEEMDAN–ARBiLSTM
achieve
better
performance
compared
Prophet-based
ones.
