Inter
national
J
our
nal
of
Electrical
and
Computer
Engineering
(IJECE)
V
ol.
8,
No.
1,
February
2018,
pp.
497
–
504
ISSN:
2088-8708
497
I
ns
t
it
u
t
e
o
f
A
d
v
a
nce
d
Eng
ine
e
r
i
ng
a
nd
S
cie
nce
w
w
w
.
i
a
e
s
j
o
u
r
n
a
l
.
c
o
m
Solar
Photo
v
oltaic
P
o
wer
F
or
ecasting
in
J
ordan
using
Artificial
Neural
Netw
orks
Mohammad
H.
Alomari
1
,
J
ehad
Adeeb
2
,
and
Ola
Y
ounis
3
1
Electrical
Engineering
Department,
Applied
Science
Pri
v
ate
Uni
v
ersity
,
Amman,
Jordan
2
Rene
w
able
Ener
gy
Center
,
Applied
Science
Pri
v
ate
Uni
v
ersity
,
Amman,
Jordan
3
School
of
Electrical
Engineering,
Electronics
and
Computer
Science,
Uni
v
ersity
of
Li
v
erpool,
United
Kingdom
Article
Inf
o
Article
history:
Recei
v
ed:
Jul
21,
2017
Re
vised:
No
v
8,
2017
Accepted:
Dec
3,
2017
K
eyw
ord:
Solar
photo
v
oltaic
solar
irradiance
PV
po
wer
forecasting
machine
learning
artificial
neural
netw
orks
ABSTRA
CT
In
this
paper
,
Artificial
Neural
Netw
orks
(ANNs)
are
used
to
study
the
correlations
between
solar
irradiance
and
solar
photo
v
oltaic
(PV)
output
po
wer
which
can
be
used
for
the
de-
v
elopment
of
a
real-time
prediction
model
to
predict
the
ne
xt
day
produced
po
wer
.
Solar
irradiance
records
were
measured
by
ASU
weather
station
located
on
the
campus
of
Ap-
plied
Science
Pri
v
ate
Uni
v
ersity
(ASU),
Amman,
Jordan
and
the
solar
PV
po
wer
outputs
were
e
xtracted
from
the
installed
264KWp
po
wer
plant
at
the
uni
v
ersity
.
Intensi
v
e
training
e
xperiments
were
carried
out
on
19249
records
of
data
to
find
the
optimum
NN
configura-
tions
and
the
testing
results
sho
w
e
xcellent
o
v
erall
performance
in
the
prediction
of
ne
xt
24
hours
output
po
wer
in
KW
reaching
a
Root
Mean
Square
Error
(RMSE)
v
alue
of
0.0721.
This
research
sho
ws
that
machine
learning
algorithms
hold
some
promise
for
the
predic-
tion
of
po
wer
production
based
on
v
arious
weather
conditions
and
measures
which
help
in
the
management
of
ener
gy
flo
ws
and
the
optimisation
of
inte
grating
PV
plant
s
into
po
wer
systems.
Copyright
c
2018
Institute
of
Advanced
Engineering
and
Science
.
All
rights
r
eserved.
Corresponding
A
uthor:
Mohammad
H.
Alomari
Electrical
Engineering
Department
Applied
Science
Pri
v
ate
Uni
v
ersity
166
Amman
11931
Jordan
+962
560
9999
Ext
1165
m
alomari@asu.edu.jo
1.
INTR
ODUCTION
The
importance
of
solar
Photo
v
oltaic
(PV)
systems
is
increasing
with
the
ongoing
industrial
gro
wth
and
the
increased
ener
gy
demand
for
de
v
eloped
and
de
v
eloping
countries
[1,
2].
Ener
gy
production
by
PV
systems
is
becoming
one
of
the
mai
n
rene
w
able
ener
gy
sources
as
it
turns
the
po
wer
of
the
sun
into
electricity
and
this
can
be
done
repeatedly
without
causing
an
y
damage
to
the
en
vironment.
The
term
“Photo
v
oltaic”
is
first
used
in
English
since
1849
as
the
process
of
light
con
v
ersion
into
electricity
[3].
Solar
PV
po
wer
plants
are
installed
in
tw
o
modes:
grid-connected
and
a
stand-alone
(Of
f-Grid)
[4].
Of
f-Grid
systems
are
used
for
isolated
or
remote
areas
that
are
normally
on
smaller
scale.
On
the
other
hand,
grid-connected
systems
are
widely
operat
ed
and
the
y
are
pro
v
en
to
be
hugely
beneficial
b
ut
the
y
were
kno
wn
as
uncertain
systems,
uncontrollable,
and
non-scheduling
po
wer
source
[5].
This
is
because
such
type
of
po
wer
production
depends
on
the
v
ariable
weather
conditions
according
to
the
geographical
area
of
the
system.
T
o
maintain
a
stable
po
wer
quality
and
scheduling
and
impro
v
e
in
v
estment
feasabi
lity
,
man
y
studies
were
reported
in
the
literature
suggesting
dif
ferent
modeling,
simulation,
and
prediction
methods
for
the
e
xpected
po
wer
production
of
solar
PV
plants
[6,
7].
In
[8],
the
accurac
y
of
one-day
ahead
predi
ction
for
the
po
wer
produced
by
1MW
PV
System
is
compared
for
tw
o
methods,
Support
V
ector
Machines
(SVM)
and
M
ultilayer
Perceptron
(MP)
Artificial
Neural
Netw
orks
(ANNs).
It
w
as
found
that
the
tw
o
algorithms
approximately
obtained
almost
the
same
accurac
y
with
0.07
KWh/m
2
and
0.11
KWh/m
2
Mean
Absolute
Error
(MAE)
and
Root
Mean
Square
Error
(RMSE),
respecti
v
ely
.
V
arious
forecasting
methods
of
PV
po
wer
output
were
re
vie
wed
in
[9].
It
w
as
demonstrated
that
an
y
model
J
ournal
Homepage:
http://iaescor
e
.com/journals/inde
x.php/IJECE
I
ns
t
it
u
t
e
o
f
A
d
v
a
nce
d
Eng
ine
e
r
i
ng
a
nd
S
cie
nce
w
w
w
.
i
a
e
s
j
o
u
r
n
a
l
.
c
o
m
,
DOI:
10.11591/ijece.v8i1.pp497-504
Evaluation Warning : The document was created with Spire.PDF for Python.
498
ISSN:
2088-8708
uses
numerically
predicted
weather
data
will
not
tak
e
into
account
the
ef
fect
of
cloud
co
v
er
and
cloud
formation
when
initializing,
therefore
sk
y
imaging
and
satellite
data
methods
used
to
predict
the
PV
po
wer
output
with
higher
accurac
y
.
The
article
also
outlined
some
k
e
y
f
actors
af
fecting
the
accurac
y
of
prediction,
such
as
forecast
horizon,
forecasting
interv
al
width,
system
size
and
PV
panels
mounting
method
(fix
ed
or
tracking).
The
aim
of
the
w
ork
published
in
[10]
w
as
to
study
the
ef
fect
of
forecast
horizon
on
the
accurac
y
of
the
method
used
to
predict
the
PV
po
wer
production,
which
w
as
Support
V
ector
Re
gression
(SVR)
using
numerically
predicted
weather
data.
T
w
o
forecast
horizons
studied:
up
to
2
and
25
hours
ahead.
As
e
xp
e
cted,
the
forecasting
of
up
to
2
hours
ahead
w
as
more
accurate
with
RMSE
and
MAE
increased
13%
and
17%,
respecti
v
ely
,
when
the
forecast
horizon
w
as
up
to
25
hours
ahead.
The
authors
of
[11]
de
v
eloped
and
v
alidated
a
model
that
adapted
an
ANN
with
tapped
delay
lines
and
b
uilt
for
one
day
ahead
forecasting.
The
inputs
were
the
irradiation
and
the
sampling
hours.
The
model
achie
v
ed
seasonal
MAE
ranging
from
12.2%
to
26%
in
spring
and
autumn,
respecti
v
ely
.
The
research
w
ork
of
[12]
compared
tw
o
short-term
forecasting
models:
the
analytical
PV
po
wer
forecasting
model
(APVF)
and
the
MP
PV
forecasting
model
(MPVF),
with
both
of
the
models
using
numerically
predicted
weather
data
and
past
hourly
v
al
u
e
s
for
PV
electric
po
wer
production.
The
tw
o
models
achie
v
ed
similar
results
(RMSE
v
arying
between
11.95%
and
12.10%)
with
forecast
horizons
co
v
ering
all
daylight
hours
of
one
day
ahead,
thus
the
models
demonstrated
their
applicability
for
PV
electric
po
wer
prediction.
A
ne
w
Ph
ysical
Hybrid
ANN
(PHANN)
method
w
as
proposed
in
[13]
to
impro
v
e
the
accurac
y
of
the
standard
ANN
method.
The
h
ybrid
method
is
based
on
ANN
and
clear
sk
y
curv
es
for
a
PV
plant.
The
PHANN
method
reduced
the
Normalized
MAE
(NMAE)
and
the
W
eighted
MAE
(WMAE)
by
almost
50%
in
man
y
days
compared
to
the
standard
ANN
method.
In
[14],
the
PV
ener
gy
production
for
the
ne
xt
day
with
15-minutes
interv
als
w
as
accurately
predicted
with
a
SVM
model
that
uses
historical
data
for
solar
irradi
ance,
ambient
temperature
and
past
ener
gy
production.
The
method
demonstrated
v
ery
good
accurac
y
with
R
2
correlation
coef
ficients
of
more
than
90%,
and
the
coef
ficient
w
as
strongly
dependent
on
the
quality
of
the
weather
forecast.
A
model
using
multilayer
perceptron-based
ANN
w
as
proposed
in
[5]
for
one
day
ahead
forecasting.
The
daily
solar
po
wer
output
and
atmospheri
c
temperature
for
70
days
used
for
training
the
ANN.
F
or
the
dif
ferent
settings
of
the
ANN
model
(number
of
hidden
layers,
acti
v
ation
function
and
learning
rule),
the
minimum
MAPE
achie
v
ed
w
as
0.855%.
In
this
research
w
ork,
ANNs
were
optimized
to
find
the
best
learning
configurations
and
map
the
a
v
ailable
solar
irradiance
records
into
the
generated
solar
PV
po
wer
.
The
proposed
system
pro
vides
real-time
ne
xt-day
predic-
tions
for
the
output
po
wer
based
on
the
kno
wledge
e
xtracted
from
the
a
v
ailable
historical
data.
These
predictions
can
be
used
by
man
y
ener
gy
management
systems
[15]
and
po
wer
control
systems
of
grid-tied
PV
plants
[16].
2.
PV
SYSTEMS
AND
D
A
T
A
The
data
used
in
this
research
were
collected
from
the
e
xisting
weather
stat
ion
and
solar
PV
plants
at
Applied
Science
Pri
v
ate
Uni
v
ersity
(ASU)
as
depicted
in
the
map
of
Figure
1.
Figure
1.
A
map
sho
wing
part
of
ASU’
s
campus.
There
are
four
separate
PV
systems
installed
at
the
uni
v
ersity
campus
for
a
total
generation
capacity
of
IJECE
V
ol.
8,
No.
1,
February
2018:
497
–
504
Evaluation Warning : The document was created with Spire.PDF for Python.
IJECE
ISSN:
2088-8708
499
550KWp:
three
rooftop
mounted
solar
systems
and
one
ground
mounted
test
field.
In
this
w
ork,
the
po
wer
production
data
e
xtracted
from
the
PV
system
ASU09
(F
aculty
of
Engineering)
[17]
is
correlated
with
the
solar
irradiance
mea-
sured
for
the
same
period
by
the
weather
station
[18]
which
is
located
about
175m
from
the
engineering
b
uilding
(see
Figure
1).
2.1.
PV
ASU09:
F
aculty
of
Engineering
The
lar
gest
PV
system
is
installed
on
top
of
the
f
aculty
of
engineering
b
uilding
with
a
capacity
of
264KWp.
It
consists
of
14
SMA
sunn
y
tripo
wer
in
v
erters
(17KW
and
10KW)
connected
with
Y
ingli
Solar
(YL
245P-29b-PC)
panels
that
are
tilted
by
11and
oriented
36(S
to
E).
The
dataset
used
in
this
research
w
as
created
using
all
reported
solar
irradiance
and
PV
po
wer
records
between
15
May
2015
and
30
September
2017.
This
consists
of
19800
PV
po
wer
and
20808
weather
station
records
with
one
hour
frequenc
y
.
3.
THE
PR
OPOSED
PREDICTION
SYSTEM
3.1.
Pr
epr
ocessing
As
sho
wn
in
Figure
2,
the
first
stage
of
our
system
is
to
mak
e
sure
that
all
data
entries
are
consistent
and
a
v
ailable
for
both
solar
irradiance
and
PV
po
wer
per
instance
of
time.
Figure
2.
A
block
diagram
for
the
proposed
system.
A
filter
w
as
designed
to
remo
v
e
out
an
y
irradiance
record
where
no
P
V
po
wer
v
alue
is
reported
at
the
same
time.
In
addition,
man
y
records
were
not
reported
correctly
because
of
some
netw
ork
connection
disruptions
and
in
some
cases
this
w
as
caused
by
an
in
v
erter
f
ailure.
An
irradiance
record
is
associated
with
a
solar
PV
output
po
wer
v
alue
at
each
hour
for
a
total
of
19249
samples
as
depicted
in
Figure
3.
As
sho
wn
in
Figure
2,
the
dataset
is
then
normalized
between
0
and
1
for
a
better
machine
learning
performance.
3.2.
Artificial
Neural
Netw
orks
ANNs
is
a
m
achine
learning
algorithm
that
interconnects
non-linear
elements
through
adjustable
weights.
The
structure
of
ANN
consists
of
three
layers:
input,
hidden,
and
output
layers
as
illustrated
in
Figure
4
[19].
The
input
layer
recei
v
es
the
ra
w
data,
and
then
these
inputs
are
processed
in
the
hidden
layer
to
be
finally
sent
as
computed
information
from
the
output
layer
[5].
Using
neural
netw
ork
learning
met
hod
s
pro
vide
a
rob
ust
algorithm
to
interpret
real-w
orld
sensor
data
[20],
and
it
has
been
widely
used
in
the
field
of
solar
ener
gy
[21].
Artificial
intelligence
techniques
can
be
used
for
sizing
PV
systems:
stand-alone
PVs,
grid-connected
PV
systems,
and
PV
-wind
h
ybrid
systems
[22].
There
are
man
y
learning
algorithms
that
can
be
used
in
our
w
ork
[23,
24,
25],
b
ut
it
w
as
sho
wn
in
the
literature
that
ANN
systems
were
pro
v
en
to
pro
vide
e
xcellent
prediction
and
classification
results
in
similar
applications
such
as
[26]
and
[27].
3.3.
ANN
Experiments
and
Optimisation
In
this
research
w
ork,
an
ANNs
netw
ork
model
w
as
created
with
fi
v
e
inputs
representing
the
solar
irradiance
(
I
r
r
)
records
at
the
same
time
of
the
pre
vious
fi
v
e
days
that
are
associated
with
a
current
solar
PV
output
po
wer
(
P
)
which
represents
the
tar
get
function
(output
node).
So,
if
the
mean
po
wer
v
alue
for
the
hour
h
on
day
d
is
represented
Solar
PV
P
ower
F
or
ecasting
in
J
or
dan
using
Artificial
...
(Alomari)
Evaluation Warning : The document was created with Spire.PDF for Python.
500
ISSN:
2088-8708
Figure
3.
The
associated
PV
po
wer
and
irradiance
data
(0
on
the
time
axis
corresponds
to
15
May
2015).
Figure
4.
The
structure
of
ANN.
by
P
h
(
d
)
,
then
it
is
associated
with
the
irradiance
v
alues
at
the
same
hour
h
for
the
pre
vious
fi
v
e
days:
I
r
r
h
(
d
1)
,
I
r
r
h
(
d
2)
,
I
r
r
h
(
d
3)
,
I
r
r
h
(
d
4)
,
I
r
r
h
(
d
5)
.
All
training
and
testing
e
xperiments
were
carried
out
using
the
MA
TLAB
ANNs
toolbox
with
the
aid
of
the
back-propag
ation
learning
algorithm
[28].
T
o
optimize
the
model
performance,
the
number
of
hidden
layers
w
as
IJECE
V
ol.
8,
No.
1,
February
2018:
497
–
504
Evaluation Warning : The document was created with Spire.PDF for Python.
IJECE
ISSN:
2088-8708
501
incremented
from
1
to
30
and
at
each
v
alue
of
hidden
layers,
ten
e
xperiments
were
carried
out
using
a
dif
ferent
set
of
randomly
mix
ed
samples
consisting
80%
of
the
samples
(15399
samples)
for
training,
5%
for
v
alidation,
and
15%
for
testing.
The
a
v
erage
RMSE
for
each
of
ten
e
xperiments
is
calculated
to
e
v
aluate
the
performance
per
specific
number
of
hidden
layers.
A
total
of
300
sets
of
training,
v
alidation,
and
testing
e
xperiments
were
handled
and
the
best
ANN
config-
urations
were
found
to
pro
vide
an
a
v
erage
RMSE
of
0.0721
and
best
v
alidat
ion
MSE
of
0.0053397
using
22
hidden
layers
for
the
testing
performance
illustrated
in
Figure
5
and
Figure
6.
These
results
are
v
ery
good
compared
to
the
methods
and
measures
reported
in
the
literature
and
related
to
the
current
research.
Figure
5.
Correlation
coef
ficients
calculations.
A
tw
o-days
prediction
for
the
PV
ener
gy
production
for
23
and
24
May
2015
w
as
simulated
using
our
model
(see
Figure
7
(left))
and
the
system
pro
vided
a
RMSE=0.0234
and
correlation
coef
ficient
of
R=0.9983
which
means
an
almost
perfect
linear
relationship
between
solar
irradiation
and
the
output
po
wer
generated.
In
addition
a
ten-days
simulation
for
the
duration
from
20
to
30
July
2015
pro
vided
RM
SE=0.0333
and
R=0.9965
as
illustrated
in
Figure
7
(right).
4.
CONCLUSIONS
In
this
w
ork,
a
machine
learning
model
is
proposed
to
analyses
historical
solar
PV
output
po
wer
and
solar
irradiance
data
to
pro
vide
a
set
of
decision
rules
that
represent
a
proper
prediction
system.
All
data
records
in
the
duration
from
16
May
2015
to
30
September
2017
were
used
in
this
research
w
ork
and
the
ANNs-based
system
pro
vided
promising
results.
W
e
belie
v
e
that
this
w
ork
is
the
first
to
predict
the
ne
xt-day
solar
PV
output
po
wer
using
real
time
irradiation
data
measured
accurately
at
a
weather
station
that
is
located
at
the
same
geographical
area
of
the
PV
plants.
Solar
PV
P
ower
F
or
ecasting
in
J
or
dan
using
Artificial
...
(Alomari)
Evaluation Warning : The document was created with Spire.PDF for Python.
502
ISSN:
2088-8708
Figure
6.
ANN
e
xperiments
using
22
hidden
layers.
Figure
7.
Measured
and
forecasted
PV
ener
gy
production
for
23-24
May
2015
(left)
and
20-30
July
2015
(right).
A
CKNO
WLEDGMENT
The
authors
w
ould
lik
e
to
ackno
wledge
the
financial
support
recei
v
ed
from
Applied
Science
Pri
v
ate
Uni
v
er
-
sity
that
helped
in
accomplishing
the
w
ork
of
this
article.
REFERENCES
[1]
W
.
Hof
fmann,
“Pv
solar
electricity
industry:
Mark
et
gro
wth
and
perspecti
v
e,
”
Solar
Ener
gy
Materials
and
Solar
Cells
,
v
ol.
90,
no.
18,
pp.
3285
–
3311,
2006.
[2]
I.
E.
Agenc
y
,
“T
echnology
roadmap:
Solar
photo
v
oltaic
ener
gy
2014
edition,
”
P
aris,
France,
2014,
last
accessed:
IJECE
V
ol.
8,
No.
1,
February
2018:
497
–
504
Evaluation Warning : The document was created with Spire.PDF for Python.
IJECE
ISSN:
2088-8708
503
2017-10-2.
[Online].
A
v
ailable:
https://goo.gl/6opzZ8
[3]
A.
Smee,
Elements
of
electr
o-biolo
gy
,:
or
the
voltaic
mec
hanism
of
man;
of
electr
o-patholo
gy
,
especially
of
the
nervous
system;
and
of
electr
o-ther
apeutics
.
London,
UK:
Longman,
Bro
wn,
Green,
and
Longmans,
1849.
[4]
V
.
Karthik
e
yan,
S.
Rajasekar
,
V
.
Das,
P
.
Karuppanan,
and
A.
K.
Singh,
Grid-Connected
and
Of
f-Grid
Solar
Photo
voltaic
System
.
Cham:
Springer
International
Publishing,
2017,
pp.
125–157.
[5]
R.
Muhammad
Ehsan,
S.
P
.
Simon,
and
P
.
R.
V
enkatesw
aran,
“Day-ahead
forecasting
of
solar
photo
v
oltaic
output
po
wer
using
multilayer
perceptron,
”
Neur
al
Computing
and
Applications
,
v
ol.
28,
no.
12,
pp.
3981–3992,
Dec
2017.
[6]
M.
Elyaqouti,
L.
Bouhouch,
and
A.
Ihlal,
“Modelling
and
predict
ing
of
the
characteristics
of
a
photo
v
oltaic
generator
on
a
horizontal
and
tilted
surf
ace,
”
International
J
ournal
of
Electrical
and
Computer
Engineering
,
v
ol.
6,
no.
6,
pp.
2557–2576,
2016,
cited
By
0.
[7]
S.
Mank
our
,
A.
Belarbi,
and
M.
Benmessaoud,
“Modeling
and
simul
ation
of
a
photo
v
oltaic
field
for
13
kw
,
”
International
J
ournal
of
Electrical
and
Computer
Engineering
,
v
ol.
7,
no.
6,
pp.
3271–3281,
2017,
cited
By
0.
[8]
J.
G.
da
Silv
a
F
onseca
Junior
,
T
.
Oozekia,
T
.
T
akashimaa,
G.
K
oshimizub,
Y
.
Uchidab,
and
K.
Ogimoto,
“F
orecast
of
po
wer
production
of
a
photo
v
oltaic
po
wer
plant
in
japan
with
multilayer
perceptron
artificial
neural
netw
orks
and
support
v
ector
machines,
”
in
26th
Eur
opean
Photo
voltaic
Solar
Ener
gy
Confer
ence
and
Exhibition
.
WIP-
Rene
w
able
Ener
gies,
Sep.
2011,
pp.
4237–4240.
[9]
A.
T
uoh
y
,
J.
Zack,
S.
Haupt,
J.
Sharp,
M.
Ahlstrom,
S.
Dise,
E.
Grimit,
C.
Moehrlen,
M.
Lange,
M.
Gar
-
cia
Casado,
J.
Black,
M.
Marquis,
and
C.
Collier
,
“Solar
forecasti
ng:
Methods,
challenges,
and
performance,
”
IEEE
P
ower
and
Ener
gy
Ma
gazine
,
v
ol.
13,
pp.
50–59,
11
2015.
[10]
J.
G.
da
Silv
a
F
onseca,
T
.
Oozeki,
T
.
T
akashima,
G.
K
oshimizu,
Y
.
Uchida,
and
K.
Ogimoto,
“Photo
v
oltaic
po
wer
production
forecasts
with
support
v
ector
re
gression:
A
study
on
the
forecast
horizon,
”
in
2011
37th
IEEE
Photo
voltaic
Specialists
Confer
ence
,
June
2011,
pp.
002
579–002
583.
[11]
M.
Cococcioni,
E.
D’Andrea,
and
B.
Lazzerini,
“One
day-ahead
forecasting
of
ener
gy
production
in
solar
pho-
to
v
oltaic
installations:
An
empirical
study
,
”
Intellig
ent
Decision
T
ec
hnolo
gies
,
v
ol.
6,
no.
3,
pp.
197–210,
Aug.
2012.
[12]
C.
Monteiro,
L.
A.
Fernandez-Jimenez,
I.
J.
Ramirez-Rosado,
and
P
.
M.
Lara-Santillan,
“Short-term
forecast-
ing
models
for
photo
v
oltaic
plants:
Analytical
v
ersus
soft-computing
techniques,
”
Mathematical
Pr
oblems
in
Engineering
,
v
ol.
2013,
pp.
1–9,
2013.
[13]
A.
Dolara,
F
.
Grimaccia,
S.
Le
v
a,
M.
Mussetta,
and
E.
Ogliari,
“
A
ph
ysical
h
ybrid
artificial
neural
netw
ork
for
short
term
forecasting
of
pv
plant
po
wer
output,
”
Ener
gies
,
v
ol.
8,
no.
2,
pp.
1138–1153,
2015.
[14]
R.
Leone,
M.
Pietrini,
and
A.
Gio
v
annelli,
“Photo
v
oltaic
ener
gy
production
forecast
using
support
v
ector
re
gres-
sion,
”
Neur
al
Computing
and
Applications
,
v
ol.
26,
no.
8,
pp.
1955–1962,
No
v
.
2015.
[15]
V
.
Jyothi,
T
.
Muni,
and
S.
Lalitha,
“
An
optimal
ener
gy
management
system
for
pv/battery
standalone
system,
”
International
J
ournal
of
Electrical
and
Computer
Engineering
,
v
ol.
6,
no.
6,
pp.
2538–2544,
2016,
cited
By
1.
[16]
B.
Allaah
and
L.
Djamel,
“Control
of
po
wer
and
v
oltage
of
solar
grid
connected,
”
International
J
ournal
of
Electrical
and
Computer
Engineering
,
v
ol.
6,
no.
1,
pp.
26–33,
2016,
cited
By
5.
[17]
ASU,
“Pv
system
asu09:
F
aculty
of
engineering,
”
2017,
last
access
ed:
2017-10-14.
[Online].
A
v
ailable:
https://goo.gl/cxGYVb
[18]
U.
ASU,
“W
eather
station,
”
2017,
last
accessed:
2017-10-14.
[Online].
A
v
ailable:
http://web
.asu.edu.jo/wsp
[19]
N.
J.
Nilsson,
“Introduction
to
machine
learning.
an
early
draft
of
a
proposed
te
xtbook,
”
1996.
[20]
T
.
M.
Mitchell,
Mac
hine
Learning
,
1st
ed.
Ne
w
Y
ork,
NY
,
USA:
McGra
w-Hill,
Inc.,
1997.
[21]
S.
A.
Kalogirou,
“
Applications
of
artificial
neural-netw
orks
for
ener
gy
systems,
”
Applied
Ener
gy
,
v
ol.
67,
no.
1,
pp.
17
–
35,
2000.
[22]
A.
Mellit,
S.
Kalogirou,
L.
Hontoria,
and
S.
Shaari,
“
Artificial
intelligence
techniques
for
sizing
photo
v
oltaic
systems:
A
re
vie
w
,
”
Rene
wable
and
Sustainable
Ener
gy
Re
vie
ws
,
v
ol.
13,
no.
2,
pp.
406
–
419,
2009.
[23]
R.
Qahw
aji,
M.
Al-Omari,
T
.
Colak,
and
S.
Ipson,
“Using
the
real,
gentle
and
modest
adaboost
learning
al-
gorithms
to
in
v
estig
ate
the
computerised
associations
between
coronal
mass
ejections
and
filaments,
”
in
2008
Moshar
aka
International
Confer
ence
on
Communications,
Computer
s
and
Applications
,
Aug
2008,
pp.
37–42.
[24]
M.
AL-Omari,
R.
Qahw
aji,
T
.
Colak,
S.
Ipson,
and
C.
Balch,
“Ne
xt-day
prediction
of
sunspots
area
and
mcintosh
classifications
us
ing
hidden
mark
o
v
models,
”
in
2009
International
Confer
ence
on
CyberW
orlds
,
Sept
2009,
pp.
253–256.
[25]
M.
Al-Omari,
R.
Qahw
aji,
T
.
Colak,
and
S.
Ipson,
“Machine
leaning-based
in
v
estig
ation
of
the
associations
between
cmes
and
filaments,
”
Solar
Physics
,
v
ol.
262,
no.
2,
pp.
511–539,
Apr
2010.
[26]
M.
Alomari,
E.
A
w
ada,
and
O.
Y
ounis,
“Subject-independent
ee
g-based
discrimination
between
imagined
and
e
x
ecuted,
right
and
left
fists
mo
v
ements,
”
Eur
opean
J
ournal
of
Scientific
Resear
c
h
,
v
ol.
118,
no.
3,
pp.
364–373,
Solar
PV
P
ower
F
or
ecasting
in
J
or
dan
using
Artificial
...
(Alomari)
Evaluation Warning : The document was created with Spire.PDF for Python.
504
ISSN:
2088-8708
02
2014.
[27]
R.
Qahw
aji,
T
.
Colak,
M.
Al-Omari,
and
S.
Ipson,
“
Automated
prediction
of
cmes
using
machine
learning
of
cme
–
flare
associations,
”
Solar
Physics
,
v
ol.
248,
no.
2,
pp.
471–483,
Apr
2008.
[28]
S.
E.
F
ahlman
and
C.
Lebiere,
“The
cascade-correlation
learning
architecture,
”
in
Pr
oceedings
of
the
2Nd
Inter
-
national
Confer
ence
on
Neur
al
Information
Pr
ocessing
Systems
,
s
er
.
NIPS’89.
Cambridge,
MA,
USA:
MIT
Press,
1989,
pp.
524–532.
BIOGRAPHIES
OF
A
UTHORS
Mohammad
Alomari
is
currently
an
Associate
Professor
of
Electrical
Engineering
(Solar
Systems)
at
Applied
Science
Pri
v
ate
Uni
v
ersity
,
Jordan.
He
recei
v
ed
his
B.Sc.
and
M.S.
de
grees
in
Electrical
Engineering
(Communications
and
Electronics)
from
Jordan
Uni
v
ersity
of
Science
and
T
echnology
,
Irbid,
Jordan,
in
2005
and
2006,
respecti
v
ely
and
the
PhD
de
gree
from
the
Uni
v
ersity
of
Bradford
in
2009.
His
research
interests
include
smart
and
green
b
uildings,
solar
PV
applications,
space
weather
and
solar
ener
gy
,
computer
vision,
brain
computer
interf
ace
and
digital
image
processing.