Processes,
Год журнала:
2023,
Номер
11(8), С. 2323 - 2323
Опубликована: Авг. 2, 2023
Feature
data
refer
to
direct
measurements
of
specific
features,
while
feature
residuals
represent
the
deviations
between
these
and
their
corresponding
benchmark
values.
Both
types
information
offer
unique
insights
into
system’s
behavior.
However,
conventional
diagnostic
systems
often
struggle
effectively
integrate
utilize
both
concurrently.
To
address
this
limitation
improve
performance,
a
hybrid
method
based
on
Bayesian
network
(BN)
is
proposed.
This
enables
parallel
fusion
within
unified
model,
comprehensive
framework
for
developing
also
given.
In
BN,
symptom
layer
consists
residual
nodes
representing
measured
data.
By
applying
proposed
two
chillers
comparing
it
with
state-of-the-art
existing
methods,
we
demonstrate
its
effectiveness
superiority.
The
results
highlight
that
not
only
accommodates
absence
either
type
but
leverages
them
enhance
performance.
Compared
using
single
node,
achieves
maximum
improvement
24.5%
in
accuracy,
significant
enhancements
F-measure
observed
refrigerant
leakage
fault
(34.5%)
excessive
lubricant
(32.8%),
respectively.
Energy Reports,
Год журнала:
2023,
Номер
11, С. 97 - 114
Опубликована: Ноя. 27, 2023
Wind
power
forecasting
plays
a
significant
role
in
regulating
the
peak
and
frequency
of
system,
which
can
improve
wind
receiving
capacity.
Despite
plenty
methods
have
been
proposed
to
fortify
accuracy
forecasting,
existing
models
do
not
consider
reconstruction
missing
data
extract
spatiotemporal
features
from
data.
To
address
these
issues,
this
study
proposes
an
improved
long
short-term
memory
(LSTM)
network
based
method
reconstruct
capture
In
order
model,
multiple
imputation
technique
(MIT)
is
first
developed
fill
up
samples
with
reconstructed
by
analyzing
correlation
among
variables
raw
Secondly,
exploit
spatial
temporal
reduce
low
computation
complexity,
new
parallel
convolutional
involving
dilated
convolution
causal
established
for
extraction.
Finally,
further
performance,
LSTM
applied
long-term
trends
reveal
internal
relations
derived
features.
The
experimental
results
on
benchmark
dataset
both
demonstrate
that
obtain
better
performance.
Results in Engineering,
Год журнала:
2024,
Номер
22, С. 102170 - 102170
Опубликована: Апрель 24, 2024
Sensor
faults
are
a
common
type
of
failure
in
heat
pump
systems,
which
can
seriously
affect
the
normal
operation
systems.
Self-correction
sensor
fault
system
is
crucial.
State-of-the-art
correction
methods
based
on
data-driven
and
physical
models
face
challenges,
such
as
need
for
co-located
sensors,
accurate
models,
large
amount
labeled
data,
greatly
limiting
their
applicability.
This
paper
proposes
using
machine
learning
self-correction.
Firstly,
data
self-correction
strategy
convolutional
autoencoder
introduced.
Furthermore,
an
artificial
sample
generation
proposed
to
address
scarcity
training
model.
The
results
demonstrate
that
method
effectively
self-corrects
both
single
multiple
faults.
Simultaneously,
thermal
diagnosis
evaluations
reveal
over
90%
accuracy
corrected
with
maximum
diagnostic
improvement
53.5%.
study
shows
number
parameters
crucial
effective
correction,
underscoring
over-constraint
essential
successful
Sensors,
Год журнала:
2023,
Номер
23(13), С. 5958 - 5958
Опубликована: Июнь 27, 2023
Due
to
energy
constraints
and
people’s
increasing
requirements
for
indoor
thermal
comfort,
improving
efficiency
while
ensuring
comfort
has
become
the
focus
of
research
in
design
operation
HVAC
systems.
This
study
took
office
rooms
with
few
people
occupying
them
Wuhan
as
object.
The
EnergyPlus-Fluent
co-simulation
method
was
used
impact
12
forms
air
distribution
on
environment
air-conditioner
consumption.
results
indicate
that
3
m/s
supply
velocity
45°
angle
are
more
suitable
case
model
this
study.
paper
provides
a
reference
environments
offices
them.