Improved monthly runoff time series prediction by integrating ICCEMDAN and SWD with ELM
Abstract
Accurate
and
timely
runoff
prediction
is
a
powerful
basis
for
important
measures
such
as
water
resource
management
flood
drought
control,
but
the
stochastic
of
brought
by
environmental
changes
human
activities
poses
significant
challenge
to
obtaining
reliable
results.
This
paper
develops
secondary
decomposition
hybrid
mode.
In
first
stage
model
design,
improved
complete
ensemble
empirical
mode
with
adaptive
noise
(ICEEMDAN)
utilized
discover
frequencies
in
predicted
non-stationary
target
data
series,
where
inputs
are
decomposed
into
intrinsic
modal
functions.
second
stage,
swarm
(SWD)
required
decomposing
high-frequency
components
whose
time-shift
multi-scale
weighted
permutation
entropy
(TSMWPE)
values
remain
calibrated
be
high
sub-sequences,
further
identifying
establishing
attributes
that
will
incorporated
extreme
learning
machine
(ELM)
algorithm
order
simulate
respective
series
component
aggregated
comprehensive
tool
prediction.
The
shows
superior
accuracy,
Nash-Sutcliffe
efficiency
exceeds
0.95
qualification
rate
greater
than
0.93,
which
can
used
decision-making
system
design
an
efficient
accurate
generating
predictions,
especially
hydrological
problems
characterized
data.
Research Square (Research Square), Journal Year: 2024, Volume and Issue: unknown
Published: Sept. 5, 2024
Language: Английский
