Advances in Building Energy Research,
Год журнала:
2024,
Номер
18(2), С. 105 - 125
Опубликована: Март 3, 2024
Thermal
comfort
in
the
building
affects
occupants'
health,
productivity,
and
electricity
use.
Predicting
thermal
advance
will
be
helpful.
Nowadays,
a
widely
used
model
for
prediction
is
predicted
mean
vote
(PMV).
However,
studies
have
found
discrepancies
between
PMV
sensation
votes.
In
this
paper,
comparative
study
was
made
by
developing
office
models
using
linear
estimation
machine
learning
(ML)
algorithms.
Fanger's
six
parameters
nine
other
are
considered
ML
input
parameters.
These
divided
into
two
groups:
psychology
group
with
parameter's
preference,
acceptance,
overall
comfort,
air
movement
vote,
humidity
personal
of
gender
age.
The
predictive
developed
showed
validated
coefficient
correlation
more
than
0.88
both
categories.
For
ML,
training
evaluation
been
done
five
models:
Artificial
Neural
Network
(ANN),
Support
Vector
Machine
(SVM),
Random
Forest
(RF),
Decision
Tree
(DT),
K-Nearest
Neighbour
(KNN).
findings
that
new
significantly
better
prediction.
RF
has
lowest
average
error,
0.706
when
including
all
psychological
Engineering Applications of Artificial Intelligence,
Год журнала:
2022,
Номер
115, С. 105254 - 105254
Опубликована: Авг. 2, 2022
The
building
internet
of
things
(BIoT)
is
quite
a
promising
concept
for
curtailing
energy
consumption,
reducing
costs,
and
promoting
transformation.
Besides,
integrating
artificial
intelligence
(AI)
into
the
BIoT
essential
data
analysis
intelligent
decision-making.
Thus,
data-driven
approaches
to
infer
occupancy
patterns
usage
are
gaining
growing
interest
in
applications.
Typically,
analyzing
big
gathered
by
networks
helps
significantly
identify
causes
wasted
recommend
corrective
actions.
Within
this
context,
aids
improvement
efficacy
management
systems,
allowing
reduction
consumption
while
maintaining
occupant
comfort.
Occupancy
might
be
collected
using
variety
devices.
Among
those
devices
optical/thermal
cameras,
smart
meters,
environmental
sensors
such
as
carbon
dioxide
(CO2),
passive
infrared
(PIR).
Even
though
latter
methods
less
precise,
they
have
generated
considerable
attention
owing
their
inexpensive
cost
low
invasive
nature.
This
article
provides
an
in-depth
survey
strategies
used
analyze
sensor
determine
occupancy.
article's
primary
emphasis
on
reviewing
deep
learning
(DL),
transfer
(TL)
detection.
work
investigates
detection
develop
efficient
system
processing
providing
accurate
information.
Moreover,
paper
conducted
comparative
study
readily
available
algorithms
optimal
method
regards
training
time
testing
accuracy.
main
concerns
affecting
current
terms
privacy
precision
were
thoroughly
discussed.
For
detection,
several
directions
provided
avoid
or
reduce
problems
employing
forthcoming
technologies
edge
devices,
Federated
learning,
Blockchain-based
IoT.
Sustainable Cities and Society,
Год журнала:
2022,
Номер
85, С. 104059 - 104059
Опубликована: Июль 19, 2022
Smart
cities
attempt
to
reach
net-zero
emissions
goals
by
reducing
wasted
energy
while
improving
grid
stability
and
meeting
service
demand.
This
is
possible
adopting
next-generation
systems,
which
leverage
artificial
intelligence,
the
Internet
of
things
(IoT),
communication
technologies
collect
analyze
big
data
in
real-time
effectively
run
city
services.
However,
training
machine
learning
algorithms
perform
various
energy-related
tasks
sustainable
smart
a
challenging
science
task.
These
might
not
as
expected,
take
much
time
training,
or
do
have
enough
input
generalize
well.
To
that
end,
transfer
(TL)
has
been
proposed
promising
solution
alleviate
these
issues.
best
authors'
knowledge,
this
paper
presents
first
review
applicability
TL
for
systems
well-defined
taxonomy
existing
frameworks.
Next,
an
in-depth
analysis
carried
out
identify
pros
cons
current
techniques
discuss
unsolved
Moving
on,
two
case
studies
illustrating
use
(i)
prediction
with
mobility
(ii)
load
forecasting
sports
facilities
are
presented.
Lastly,
ends
discussion
future
directions.
Building and Environment,
Год журнала:
2023,
Номер
234, С. 110148 - 110148
Опубликована: Март 1, 2023
Personal
thermal
comfort
models
are
used
to
predict
individual-level
responses
inform
design
and
control
decisions
of
buildings
achieve
optimal
conditioning
for
improved
energy
efficiency.
However,
the
development
data-driven
requires
collecting
a
large
amount
sensor-related
measurements
user-labelled
data
(i.e.,
user
feedback)
accurate
predictions,
which
can
be
highly
intrusive
labour
intensive
in
real-world
applications.
In
this
work,
we
propose
hybrid
active
learning
framework
reduce
collection
costs
developing
data-efficient
robust
personal
that
users'
air
movement
preferences.
Through
proposed
framework,
evaluated
performance
two
algorithms
Uncertainty
Sampling
Query-by-Committee)
labelling
strategies
(Independent
Joint
Labelling
strategies)
reduction
effort
modelling.
The
effectiveness
was
demonstrated
on
dataset
involving
58
participants
collected
over
10
working
days
with
2,727
under
16
conditions.
final
results
showed
46%
35%
preference
models,
respectively,
increasing
reductions
occurring
time
when
encountering
new
users.
insights
gained
study,
future
studies
adopt
as
viable
effective
solution
address
high
cost
while
maintaining
model's
scalability
predictive
performance.
Personal
thermal
comfort
models
are
a
paradigm
shift
in
predicting
how
building
occupants
perceive
their
environment.
Previous
work
has
critical
limitations
related
to
the
length
of
data
collected
and
diversity
spaces.
This
paper
outlines
longitudinal
field
study
comprising
20
participants
who
answered
Right-Here-Right-Now
surveys
using
smartwatch
for
180
days.
We
more
than
1080
field-based
per
participant.
Surveys
were
matched
with
environmental
physiological
measured
variables
indoors
homes
offices.
then
trained
tested
seven
machine
learning
participant
predict
preferences.
Participants
indicated
58%
time
want
no
change
environment
despite
completing
75%
these
at
temperatures
higher
26.6°C.
All
but
one
personal
model
had
median
prediction
accuracy
0.78
(F1-score).
Skin,
indoor,
near
body
temperatures,
heart
rate
most
valuable
accurate
prediction.
found
that
≈250–300
points
needed
We,
however,
identified
strategies
significantly
reduce
this
number.
Our
provides
quantitative
evidence
on
improve
models,
prove
benefits
wearable
devices
preference,
validate
results
from
previous
studies.
