Sustainability,
Год журнала:
2024,
Номер
16(23), С. 10721 - 10721
Опубликована: Дек. 6, 2024
Factory
aquaculture
faces
a
difficult
situation
due
to
its
high
running
costs,
with
one
of
the
main
contributing
factors
being
energy
consumption
workshops.
Accurately
predicting
power
load
recirculating
systems
(RAS)
is
critical
optimizing
use,
reducing
consumption,
and
promoting
sustainable
development
factory
aquaculture.
Adequate
data
can
improve
accuracy
prediction
model.
However,
there
are
often
missing
abnormal
in
actual
detection.
To
solve
this
problem,
study
uses
time-series
convolutional
network–temporal
sequence
generation
adversarial
network
(TCN-TimeGAN)
synthesize
multivariate
RAS
train
long
short-term
memory
(LSTM)
on
original
generated
predict
future
electricity
loads.
The
experimental
results
show
that
based
improved
TCN-TimeGAN
provide
more
comprehensive
coverage
distribution,
lower
discriminative
score
(0.2419)
predictive
(0.0668)
than
conventional
TimeGAN.
Using
for
prediction,
R2
reached
0.86,
which
represents
19%
improvement
over
ARIMA
Meanwhile,
compared
LSTM
GRU
without
augmentation,
mean
absolute
error
(MAE)
was
reduced
by
1.24
1.58,
respectively.
model
has
good
performance
generalization
ability,
benefits
saving,
production
planning,
term
sustainability
Deleted Journal,
Год журнала:
2024,
Номер
20(3s), С. 780 - 788
Опубликована: Апрель 4, 2024
This
research
about
presents
a
groundbreaking
approach
to
revolutionize
farming
through
the
integration
of
Web
Things
(IoT)
innovation
and
progressed
machine
learning
calculations.
Centering
on
improvement
execution
an
IoT-based
edit
checking
framework
coupled
with
Random
Forest
calculation
for
malady
expectation,
ponder
points
improve
agrarian
hones
relieve
trim
misfortunes
caused
by
infections
natural
components.
Real-time
information
collection
from
IoT
sensors
sent
in
rural
areas
empowers
comprehensive
vital
parameters
such
as
temperature,
mugginess,
soil
dampness,
light
concentrated.
The
analyzes
this
precisely
foresee
maladies,
giving
ranchers
significant
bits
knowledge
proactive
illness
administration.
Test
comes
illustrate
adequacy
proposed
approach,
show
accomplishing
exactness
92%,
93%,
review
91%,
F1-score
92%.
These
almost
defeat
customary
methodologies
existing
explore
works,
highlighting
potential
optimizing
alter
proficiency
ensuring
around
world
food
security.
Sustainability,
Год журнала:
2024,
Номер
16(23), С. 10721 - 10721
Опубликована: Дек. 6, 2024
Factory
aquaculture
faces
a
difficult
situation
due
to
its
high
running
costs,
with
one
of
the
main
contributing
factors
being
energy
consumption
workshops.
Accurately
predicting
power
load
recirculating
systems
(RAS)
is
critical
optimizing
use,
reducing
consumption,
and
promoting
sustainable
development
factory
aquaculture.
Adequate
data
can
improve
accuracy
prediction
model.
However,
there
are
often
missing
abnormal
in
actual
detection.
To
solve
this
problem,
study
uses
time-series
convolutional
network–temporal
sequence
generation
adversarial
network
(TCN-TimeGAN)
synthesize
multivariate
RAS
train
long
short-term
memory
(LSTM)
on
original
generated
predict
future
electricity
loads.
The
experimental
results
show
that
based
improved
TCN-TimeGAN
provide
more
comprehensive
coverage
distribution,
lower
discriminative
score
(0.2419)
predictive
(0.0668)
than
conventional
TimeGAN.
Using
for
prediction,
R2
reached
0.86,
which
represents
19%
improvement
over
ARIMA
Meanwhile,
compared
LSTM
GRU
without
augmentation,
mean
absolute
error
(MAE)
was
reduced
by
1.24
1.58,
respectively.
model
has
good
performance
generalization
ability,
benefits
saving,
production
planning,
term
sustainability
