Frontiers in Energy Research,
Journal Year:
2024,
Volume and Issue:
12
Published: Dec. 13, 2024
Introduction
In
the
domain
of
nuclear
power
plant
operations,
accurately
and
rapidly
predicting
future
states
is
crucial
for
ensuring
safety
efficiency.
Data-driven
methods
are
becoming
increasingly
important
parameter
forecasting.
While
Transformer
neural
networks
have
emerged
as
powerful
tools
due
to
their
self-attention
mechanisms
ability
capture
long-range
dependencies,
application
in
energy
field
remains
limited
capabilities
largely
untested.
Additionally,
models
highly
sensitive
data
complexity,
presenting
challenges
model
development
computational
Methods
This
study
proposes
a
feature
selection
method
that
integrates
clustering
mutual
information
techniques
reduce
dimensionality
training
before
applying
models.
By
identifying
key
physical
quantities
from
large
datasets,
we
refine
used
model,
which
then
optimized
using
Tree-structured
Parzen
Estimator
algorithm.
Results
Applying
this
dataset
shutdown
condition
plant,
demonstrate
effectiveness
proposed
“feature
+
Transformer”
approach:
(1)
The
achieved
high
accuracy
parameters,
with
such
temperature,
pressure,
water
level
attaining
normalized
root
mean
squared
error
below
0.009,
indicating
RMSE
0.9%
range
original
data,
reflecting
very
small
prediction
error.
(2)
effectively
reduced
input
minimal
impact
on
accuracy.
Discussion
results
information-based
provides
an
effective
strategy
encapsulates
operational
plant.
The Science of The Total Environment,
Journal Year:
2025,
Volume and Issue:
963, P. 178541 - 178541
Published: Jan. 17, 2025
The
role
of
biochar
in
reducing
greenhouse
gas
(GHG)
emissions
and
improving
soil
health
is
a
topic
extensive
research,
yet
its
effects
remain
debated.
Conflicting
evidence
exists
regarding
biochar's
impact
on
microbial-mediated
with
respect
to
different
GHGs.
This
study
systematically
examines
these
divergent
perspectives,
aiming
investigate
influence
GHG
agricultural
soils.
meta-analysis
includes
2594
paired
observations
from
157
studies
conducted
between
2000
2024.
It
was
found
that
increased
the
presence
amoA
nosZ
genes
by
39.4
%
41.7
%,
respectively,
while
abundance
nirS
gene
17.8
%.
led
13.1
decrease
N2O
emissions.
Nitrous
were
positively
associated
mean
annual
temperature
pyrolysis
dosage
inversely
related
pH,
nitrogen
fertilisation
rate,
pH
carbon
content.
Biochar
also
regulated
enzyme
activity
nutrient
cycle
microbial
biomass
carbon,
nitrogen,
phosphorus
16.6
23.9
50.2
leading
changes
community
diversity.
These
contributed
reduction
CO2
CH4
emissions,
particularly
when
fertiliser
applied
at
doses
below
21.4
t
ha-1
242.5
kg
ha-1,
as
predicted
machine
learning
models.
offers
an
overview
positive
amendments
mitigation.
key
predictive
factors
identified
could
help
optimise
production
targeted
amendments,
potentially
achieving
neutrality
agroecosystems.
Remote Sensing,
Journal Year:
2024,
Volume and Issue:
16(11), P. 1948 - 1948
Published: May 28, 2024
Monitoring
and
evaluation
of
soil
ecological
environments
are
very
important
to
ensure
saline–alkali
health
the
safety
agricultural
products.
It
is
foremost
importance
to,
within
a
regional
risk-reduction
strategy,
develop
useful
online
system
for
assessment
prediction
prevent
people
from
suffering
threat
sudden
disasters.
However,
traditional
manual
or
empirical
parameter
adjustment
causes
mismatch
hyperparameters
model,
which
cannot
meet
urgent
need
high-performance
properties
using
multi-dimensional
data
in
WebGIS
system.
To
this
end,
study
aims
monitoring
real-time
ecology
Yellow
River
Delta,
China.
The
applied
advanced
web-based
GIS,
including
front-end
back-end
technology
stack,
cross-platform
deployment
machine
learning
models,
database
embedded
multi-source
environmental
variables.
adopts
five-layer
architecture
integrates
functions
such
as
statistical
analysis,
assessment,
salt
prediction,
management.
visually
displays
results
air
quality,
vegetation
index,
area.
provides
users
with
risk
analyze
heavy
metal
pollution
soil.
Specially,
introduces
tree-structured
Parzan
estimator
(TPE)-optimized
model
achieve
accurate
salinity.
TPE–RF
had
highest
accuracy
(R2
=
94.48%)
testing
set
comparison
TPE–GBDT
exhibited
strong
nonlinear
relationship
between
variables
developed
can
provide
information
government
agencies
farmers,
great
significance
production
protection.
