The
air
conditioning
system
consumes
more
than
half
of
the
total
energy
demand
in
hub
airport
buildings.
To
enhance
efficiency
and
to
enable
intelligent
management,
it
is
vital
build
an
accurate
cold
load
prediction
model.
However,
current
models
face
challenges
dealing
with
dispersed
patterns
lack
interpretability
when
black
box
are
adopted.
tackle
these
challenges,
we
propose
a
novel
k-means-Temporal
Fusion
Transformer
(TFT)
based
hybrid
Specifically,
daily
grouped
using
improved
k-means
clustering
method
that
considers
both
input
feature
weights
dynamic
time
warping
(DTW)
distances.
Additionally,
statistical
features
output
inputted
into
TFT.
By
further
incorporating
context
information,
integration
data
between
different
schema
categories
achieved,
thus
reducing
errors
may
occur
during
transition
process.
As
result,
performance
significantly
improved.
Chongqing
Jiangbei
Airport
T3A
terminal
used
as
case
study,
experiments
conducted
cooling
from
No.1
station,
well
traffic
meteorological
station
data.
Results
compared
other
mainstream
models,
confirming
proposed
day-ahead
forecasting
model
achieves
improvements
several
indicators,
including
MAE,
MAPE,
CV-RMSE,
R2,
which
384
kW,
3%,
5%,
0.058
respectively.
Sustainability,
Journal Year:
2024,
Volume and Issue:
16(6), P. 2522 - 2522
Published: March 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.
Energy and Buildings,
Journal Year:
2023,
Volume and Issue:
303, P. 113812 - 113812
Published: Dec. 1, 2023
Real-time
nonintrusive
occupancy
estimation
can
maximize
the
use
of
existing
sensors
to
infer
occupant
information
in
buildings
with
advantages
fewer
privacy
concerns
and
extra
device
costs.
Recently,
many
deep
learning
architectures
have
proven
effective
estimating
directly
from
raw
sensor
data.
However,
some
handcrafted
features
manually
extracted
statistical
temporal
domains
might
convey
additional
for
estimation.
In
this
study,
a
novel
knowledge
fusion
network
is
proposed
integrate
two
streams,
i.e.
automatic
stream
architecture
manual
feature
engineering.
Moreover,
four
different
modules
are
investigated
optimize
design
network.
To
verify
effectiveness
network,
experiments
conducted
dataset
ASHRAE
Global
Occupant
Behavior
Database,
which
collected
an
office
space
records
indoor
environment
parameters,
occupant-building
interactions,
contextual
information.
The
results
demonstrate
superiority
outperforms
five
representative
algorithms.
Furthermore,
ablation
study
underscores
benefits
interaction
information,
showing
that
enhance
accuracy
by
3.47%
9.24%.
The
air
conditioning
system
constitutes
more
than
half
of
the
total
energy
demand
in
hub
airport
buildings.
To
enhance
efficiency
and
to
enable
intelligent
management,
it
is
vital
build
an
accurate
cold
load
prediction
model.
However,
current
models
face
challenges
dealing
with
dispersed
patterns
lack
interpretability
when
black
box
are
adopted.
tackle
these
challenges,
we
propose
a
novel
k-means-Temporal
Fusion
Transformer
(TFT)
based
hybrid
Specifically,
daily
grouped
using
improved
k-means
clustering
method
that
considers
both
input
feature
weights
dynamic
time
warping
(DTW)
distances.
Additionally,
statistical
features
output
inputted
into
TFT.
By
further
incorporating
context
information,
integration
data
between
different
schema
categories
achieved,
thus
reducing
errors
may
occur
during
transition
process.
As
result,
performance
significantly
improved.
Chongqing
Jiangbei
Airport
T3A
terminal
used
as
case
study,
experiments
conducted
cooling
from
No.1
station,
well
traffic
meteorological
station
data.
Results
compared
other
mainstream
models,
confirming
proposed
day-ahead
forecasting
model
achieves
improvements
several
indicators,
including
MAE,
MAPE,
CV-RMSE,
R2,
which
384
kW,
3%,
5%,
0.058
respectively.
Pharmaceutical
production
cannot
be
separated
from
industrial
refrigeration
systems.
The
high
dependence
on
central
air-conditioning
leads
to
a
large
proportion
of
energy
consumption
costs
in
costs.
Reducing
cost
is
the
common
goal
enterprises,
and
saving
carbon
reduction
government's
expectation
enterprises.
By
predicting
air-conditioning,
it
possible
combine
operation
sub-equipment
more
reasonable
way,
give
energy-saving
maintenance
management
suggestions,
help
factories
achieve
reduction.
However,
traditional
basic
prediction
models,
such
as
convolutional
neural
network
(CNN),
long
short-term
memory
(LSTM),
etc.,
have
errors
non-linear
scenarios.
In
order
get
better
results,
based
system
an
auxiliary
workshop
pharmaceutical
company
East
China,
this
study
improves
combines
proposes
Multiple
Kernel
Convolutional
Neural
Network-Bidirectional
Long
Short-Term
Memory-Attention
(MKCNN-BiLSTM-Attention)
method.
results
show
that
MKCNN-BiLSTM-Attention
model
are
reliable
compared
with
underlying
temporal
their
combined
model.
