TELKOM
NIKA Indonesia
n
Journal of
Electrical En
gineering
Vol. 12, No. 11, Novembe
r
2014, pp. 76
1
3
~ 762
1
DOI: 10.115
9
1
/telkomni
ka.
v
12i11.64
10
7613
Re
cei
v
ed
Jun
e
21, 2014; Revi
sed Septe
m
ber
4, 2014
; Accepte
d
Septem
ber 25,
2014
A Clean Economic Dispatch of PV Energy Storage
Connected to Grid Based on LHS-SR-GAMS
Technology
Liu Jiang-T
a
o*
1
, Wang Hai-Yun
1
, Luo Jian-Chun
1
, Luo Qing
2
, Chen Xing
1
1
Colle
ge of Ele
c
trical Eng
i
ne
e
r
ing,
Xin
jia
ng U
n
iversit
y
, Urum
qi 83
00
47, Chi
na;
2
State Grid Pow
e
r Comp
an
y
of Xi
nji
ang E
l
e
c
tric Po
w
e
r Re
search Institute
,
Urumqi 83
00
00, Chi
n
a
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: xyer.1
227
@1
63.com
A
b
st
r
a
ct
PV outp
u
t unc
ertainty, the P
V
sched
uli
ng
prob
le
m h
a
s b
e
co
me
urg
ent. T
he co
mb
in
e
d
en
er
g
y
storage a
nd P
V
can solve e
ffectivel
y the prob
le
m. W
e
establ
ishe
d t
he joint en
ergy
-savin
g econ
o
m
i
c
sched
uli
ng
mo
del o
n
PV- Storage g
ener
atio
n, w
h
ich mai
n
l
y
includ
ed t
he
pen
alty mode
l abo
ut PV positi
v
e
devi
a
tion o
u
tp
ut and the pe
nalty mod
e
l a
bout PV neg
ative devi
a
tio
n
o
u
tput
and "
o
ve
rflow
i
ng ne
gati
v
e
reven
ue"
mod
e
l. Co
nsi
deri
n
g the
unc
ertai
n
ty of t
he
PV
output, this
pa
per a
n
a
l
y
z
e
d
the
pred
iction
erro
r
distrib
u
tion c
h
a
r
acteristics ab
o
u
t PV output b
y
the
prob
ab
ilit
y density esti
mate metho
d
. Based
on the
L
H
S
(Latin hy
percu
be sa
mp
lin
g) –
S
R (scenes re
duce) tech
no
lo
gy, PV uncertai
n
output is co
n
v
erted into a fi
nit
e
PV outp
u
t sce
nes u
n
d
e
r diffe
rent pro
b
a
b
il
ity cond
itio
ns
. F
i
nally, w
e
used
the PV o
u
tput
scenar
ios
as t
h
e
inp
u
t dat
a, a
n
d
w
e
so
lve
d
t
he
prop
ose
d
Mode
l
b
a
se
d
on GAMS (Ge
nera
l
Al
ge
brai
c Mod
e
l Syste
m
)
softw
are,
w
h
ich the opti
m
i
z
at
ion g
oal is the
join
t ex
pectati
o
n
max
i
mu
m po
w
e
r generati
o
n
benefit. Resu
lt
s
show
that rel
a
tive to the
e
nergy-s
avin
g
mo
de
l of
the
indiv
i
d
ual
use
s
ener
gy storage
pla
n
e
d
o
u
tput
conn
ected to
grid, b
a
se
d o
n
the tech
nol
og
y of LHS-
S
R
-
G
AMS, the co
mb
in
ed e
ner
gy
storage
an
d
PV
mo
de
l pla
n
n
e
d
output incre
a
s
ed by 8%, a
nd to a ce
rtai
n extent, improved t
he PV
output pre
d
icti
on
accuracy.
Ke
y
w
ords
:
P
V
an
d stora
ge,
GAMS, positi
v
e a
nd
ne
gativ
e d
e
viati
on,
ov
erflow
ing
ne
ga
tive b
enefit, cl
e
a
n
econ
o
m
ic dis
p
atch
Copy
right
©
2014 In
stitu
t
e o
f
Ad
van
ced
En
g
i
n
eerin
g and
Scien
ce. All
rig
h
t
s reser
ve
d
.
1. Introduc
tion
As the traditional e
nergy
sho
r
tage,
en
vir
onme
n
t po
llution is i
n
creasi
ngly seri
ously, a
new
kind of clean en
ergy
become
s
very important.
The solar
re
source
s are ri
ch in ou
r cou
n
try,
and PV
po
we
r ge
ne
ration
has many
ch
ara
c
teri
stic
s,
su
ch
as sust
ainabl
e, no
p
o
llution[1-3].The
state co
un
cil issued ‘Sola
r
powe
r
tech
n
o
logy dev
elo
p
ment "twelfth five-year" speci
a
l planni
ng’,
and will ma
ke PV capa
city more and
more bi
g. Bu
t PV domestic mainly used
in the form of a
large,
ce
ntral
i
zed
gri
d
, PV sho
r
t-te
rm
output p
r
edi
ction p
r
e
c
i
s
io
n is l
o
w, o
u
tput un
ce
rtain
t
y
(clo
ud
s) m
a
kes the g
r
id P
V
powe
r
gen
eration
sc
he
d
u
ling is the v
e
ry difficult problem
s urgen
tly
need
ed to sol
v
e.
With the
appl
ication
of en
e
r
gy sto
r
a
ge t
e
ch
nolo
g
y, maturity and
lo
wer cost, PV
stora
ge
power
gen
era
t
ion be
com
e
t
he effe
ctive
way to
so
lve
the
a
bove pro
b
lems,
su
ch
as dome
s
tic has
impleme
n
ted
Zhang B
e
i
PV and wi
nd
stora
ge
key
demon
stration proje
c
t. Whe
n
PV actual
output value
s
gre
a
ter tha
n
plann
ed o
u
tp
ut value,
the
energy sto
r
ag
e devic
es
ca
n sto
r
e the
extra
power;
whe
n
the actu
al o
u
tput value i
s
l
o
we
r
tha
n
the
plann
ed
outp
u
t value, the
energy sto
r
a
ge
device
s
ca
n relea
s
e
th
e p
o
we
r,
avoi
de
d
sho
r
t
of out
put
po
we
r pl
anne
d
valu
e and re
ceived
the
puni
shme
nt o
f
the powe
r
sector [4
-7], a
nd it can
imp
r
ove the effici
ency of the p
hotoele
c
tri
c
. No
w
the main ene
rgy storage i
n
clu
d
ing fly whe
e
l ene
rg
y
storag
e, pu
mped sto
r
a
g
e, comp
re
sse
d
air
energy storag
e, ba
ttery, etc. [10].
In the literature [11]
p
r
op
osed wind
farm
the
con
c
ept of
"negative
e
ffect" ope
ratio
n
, they
establi
s
h
ed t
he mod
e
l a
bout a la
rge
-
scale
wi
n
d
power g
r
id i
n
terconn
ectio
n
optimizatio
n
sched
uling, b
u
t they did
n
o
t co
nsi
der the o
u
tput
p
r
edictio
n e
rro
r, and
a la
ck of practi
calit
y.
Literatu
re [1
2] put fo
rward th
e g
r
id
-co
nne
cted
static
sched
uli
ng mo
del
ab
out PV po
wer
gene
ration
wi
th
generating
co
st
minimu
m a
s
the
obj
ective.
Lite
rat
u
re
[13]
c
o
nsid
e
r
ing
gr
id th
e
bigge
st acce
pt PV powe
r
gene
ration
ca
pacity, it
put forward aba
n
don PV puni
shment cost, a
nd
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 11, Novem
ber 20
14: 76
13 – 762
1
7614
they establi
s
h
ed the stati
c
sched
uling m
odel. Literature
[14] con
s
id
ering th
e PV predi
ction
error,
it put forward the dyna
mic
sc
heduli
ng mod
e
l b
a
se
d on PV
pre
d
iction
e
rro
r. But it didn’t
analysi
s
di
stri
bution
cha
r
a
c
terist
ics of
the
predi
ction error, wh
i
c
h
woul
d have
a
gre
a
t influen
ce
on sche
dulin
g results.
Most of the a
bove do
cum
e
nts did
not co
nsid
er
p
hotov
oltaic o
u
tput
predi
ction
error when
arrangi
ng schedul
e, thus
ignori
ng
the
photovoltai
c
output un
ce
rtai
nty. which
woul
d affect
the
safety of p
o
wer g
r
id, m
a
ke
s the
po
we
r
sector in
crea
sed
system
re
serve
c
apa
cit
y
, leading
to t
h
e
increa
sed
co
st of powe
r
gene
ration;
Bas
ed o
n
co
mbined p
o
wer gen
eratio
n
and PV storage
sched
uling
p
r
oble
m
, integ
r
ated
natio
na
l ne
w
e
n
e
r
gy
gen
eration su
ch as
PV encourage
me
nt
policy and th
e influen
ce o
f
PV power gene
ration o
n
power g
r
id
safe ope
rati
on. Con
s
id
eri
ng
encourage
PV powe
r
gene
ration a
bout co
mbin
ed PV storage outp
u
t, we e
s
tabli
s
he
d
respe
c
tively positive d
e
viation revenu
e mod
e
l a
n
d
neg
ative de
viation puni
shment
reven
u
e
model, and
we put forwa
r
d
the con
s
ide
r
ation grid
of
PV maximum given cap
a
cit
y
of “overflowing
negative reve
nue”
mod
e
l. Then ba
sed on
several
ki
nds
of reven
ue mo
del p
u
t forward
by t
he
pape
r, we e
s
tablish
ed co
mbined PV storage p
o
wer energy
-saving economi
c
dispat
ch mo
del.
Whi
c
h con
s
id
ering the
un
certainty of PV output co
m
b
i
ned with the
actual
situati
on of PV outpu
t
in a
reven
u
e
model, i
n
o
r
der to
solve
the propo
se
d
model, thi
s
pape
r a
nalyzed di
strib
u
tio
n
characteri
stics of the PV
output predi
ction er
ror based on the probability density estim
a
te;
Based on
the
LHS
(Latin h
y
percube sa
mpling
)
–SR (scen
e
s red
u
ce)
te
ch
nolog
y,
PV
uncertain
output is
con
v
erted into a
finite PV output scene
s u
nder
different
prob
ability condition
s, wh
ich
provide
d
the
basi
s
cal
c
ula
t
ing data fo
r
PV scheduli
n
g mod
e
l. Fin
a
lly, we
cho
s
e a PV p
o
wer
station i
n
Xi
njiang
a
s
a
n
exampl
e,
and
we
solved the
p
r
op
o
s
ed
Mo
del
b
a
se
d o
n
GAMS
(Gen
eral Alg
ebrai
c Mod
e
l
System)
soft
ware,
whi
c
h t
he o
p
timization g
oal i
s
the
joint
expecta
tion
maximum po
wer
gen
erati
on ben
efit. Result
s sh
ow
t
hat the pro
p
o
se
d metho
d
is effective, and
had go
od en
g
i
neeri
ng ap
pli
c
ation valu
e.
2.
PV Storage
Conn
ected-g
rid Energ
y
-sav
i
ng Economic Optimization Sche
d
u
ling Mode
l
and Solv
ing
Acco
rdi
ng to
the engi
nee
ri
ng a
c
tual
situ
ati
on, the mo
del p
r
op
ose
d
by the pa
per
need
ed
to con
s
ide
r
the photovolt
a
ic po
we
r output unc
e
r
t
a
inty. So this pape
r firstl
y analyzed the
distribution
of PV output predi
ct
ion error
based
on the probability densi
ty esti
mation methods,
whi
c
h
combi
n
ed the PV pre
d
ict outp
u
t value to PV fo
re
ca
st distri
buti
on char
acte
ri
stics, ba
sed
on
the L
H
S -
RS techn
o
logy
sam
p
ling
an
d re
du
cing, f
o
rmin
g a limi
t
ed outp
u
t. Seco
ndly, in t
he
pro
c
e
ss
of sche
duling m
odelin
g integ
r
ated
state for PV power gene
ration
encourage
me
nt
polici
e
s a
nd
PV powe
r
ge
neratio
n for t
he influen
ce
of the safe
o
peratio
n of p
o
we
r g
r
id an
d PV
output un
certainty, we establi
s
he
d PV store
joint
powe
r
ene
rgy-savin
g
economi
c
disp
atch
model. Finall
y
using GA
MS softwa
r
e
written
en
ergy-savin
g economi
c
disp
atch mo
del
and
solving the P
V
output scen
es num
be
r. T
he gen
eral id
ea is sho
w
n i
n
Figure 1.
Figure 1. The
Structure ab
out the Econ
omic Di
sp
atch of PV Energy Storage Base
d on L
H
S-SR-
GAMS
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
A Clean Econ
om
ic Disp
atch of PV Energy St
ora
ge
Conne
cted
to Grid… (Liu Ji
ang-Tao
)
7615
3.
Determine di
ffer
e
nt s
cen
arios outp
u
t
of the PV
The p
r
opo
se
d reven
ue m
odel con
s
ide
r
ing the un
ce
rtainty of PV
output, so
we
neede
d
to input th
e P
V
output
sce
nario
of
all d
a
y
whe
n
mo
de
l in the
proce
s
s of
cal
c
ulati
on (se
e
se
ction
3).
The
r
efo
r
e
this sectio
n
first intro
d
u
c
ed
ho
w to
cal
c
ulate PV
output sce
n
a
rio. PV po
wer
station
u
s
ed
predi
ction
m
e
thod i
s
quo
ted in thi
s
p
aper is b
a
se
d on
the i
m
proved
fee
d
b
a
ck
neural net
wo
rk
about PV
sho
r
t-te
rm p
r
edictio
n
outp
u
t, but too m
any and
com
p
licate
d
facto
r
s
influen
cing th
e PV output, resulting in t
he PV had a
certai
n erro
r. So
we co
ul
dn’t conve
r
t the
uncertainty o
f
the PV out
put into
a fini
te numb
e
r of
scen
ario
s
u
n
til study th
e
pre
d
ictio
n
e
r
ror
distrib
u
tion,
whi
c
h p
r
ovid
e ba
sic data
for the
next
step
a
r
rang
e sch
edule.
The d
e
finitio
n
of
predi
ction e
r
ror as follo
ws:
..
.m
a
x
(
%
)
100%
real
t
f
ore
t
t
W
PP
erro
r
P
(
1
)
Whi
c
h
.
re
a
l
t
P
is the a
c
tual outp
u
t of PV under the t time,
.
f
ore
t
P
is the predi
ct outp
u
t of PV under
the t time,
.m
a
x
P
P
is the ca
pa
city of PV powe
r
station.
2.1. PV Predicted O
u
tpu
t
Error Distrib
u
tion
We n
eed to
study PV output predi
ction
distrib
u
tion to
find out different scen
ario
s output.
This a
r
ticle
use
s
the PV output predi
ction erro
r
p
r
oba
bility density function
erro
rx (x, t) to
r
e
pr
es
e
n
t
the p
r
ed
ic
tion
err
o
r e
r
r
o
r
(
x
, t)
d
i
s
t
r
i
b
u
tion
cha
r
a
c
teri
stics, an
d
solvin
g by e
s
timati
ng
the prob
abilit
y density function, the solving pro
c
e
s
s is as follows:
12
()
,
(
)
.
.
(
)
.
.
.
()
cy
X
tX
t
X
t
X
t
PV sampl
e
predi
ction
error pr
obability density functi
on of
(,
)
X
Px
t
estimate
s
can b
e
obtai
n
ed by
den
sity evol
ution.
We
ca
n p
u
t the
sampl
e
e
r
ror a
s
a
rep
r
e
s
entativ
e pro
c
e
ss [6,
7] beca
u
se it is inde
pend
e
n
t, and its pro
bability is:
1
(,
)
c
px
t
y
(2)
Obviou
sly
1
(,
)
1
m
c
px
t
, for c
(
1
cy
)
rep
r
e
s
en
tative process, t
he den
sity solving evol
u
t
ion
equatio
n is:
:
(,
)
(
,
)
()
0
Xc
c
q
px
t
p
x
t
Xt
tx
(
3
)
The co
rrespo
nding initial condition
s a
s
follows:
:0
:
(,
)
(
)
Xc
c
c
Px
t
x
x
P
(
4
)
The initial value of
0:
c
x
is
the firs
t c
s
a
mple:
0:
0
()
cc
x
Xt
(5)
To solve th
e 3-4 type,
we get
:
(,
)
Xc
Px
t
, and we
can get th
e pro
bability
den
sity function estimate o
f
()
X
t
:
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 11, Novem
ber 20
14: 76
13 – 762
1
7616
::
1
(,
)
(
,
)
m
Xc
Xc
c
p
xt
p
x
t
(
6
)
2.2. The De
termination o
f
PV Outpu
t
Scenario
We throug
h the form
ula 1,
every predict
ion erro
r corresp
ondi
ng to
a scen
ario fo
r
i
p
, the
output for the scenari
o
in the probability
of
i
p
as
follows
:
..
.
.
.
m
a
x
(
)
,
1
,
..,
P
i
t
f
or
e
t
i
f
or
e
t
P
PP
e
P
P
i
N
(8)
Whi
c
h
..
P
it
P
is the i
t
h scen
ario
PV output u
n
d
e
r the
time t,
i
e
is
th
e co
rr
es
p
o
n
d
i
ng
PV
output pre
d
ict
i
on error in th
e ith sce
nari
o
, N is
the sum
of all the prediction e
r
ror scen
ario
s.
3. PV Output Scenario Ge
neratio
n
and
Reduc
tion
3.1. Latin H
y
percub
e Sa
mpling of th
e Outpu
t
of
PV
By 1.1 we o
b
tained th
e d
i
stributio
n ch
ara
c
teri
stics
of PV output pre
d
iction
error, from
the type 8 we co
uld b
e
known sce
ne
distrib
u
tion o
f
PV output. Since th
e L
a
tin hypercu
be
sampli
ng is a
improved
sa
mpling metho
d
of Monte Carlo samplin
g
,
which extra
c
ts the sampl
e
more
rep
r
e
s
e
n
tative of the entire sampl
e
interv
al, an
d the any si
ze of
the num
ber of sampl
e
s
coul
d easily
produ
ce [7]. So this article us
ed th
e Latin hype
rcu
be sampli
ng for effect
ive
sampli
ng of PV output scenari
o
, Latin hypercub
e
sa
mpling of PV output scen
ario pro
c
ed
ure a
s
follows
:
1)
It is con
c
lud
e
d
that pro
bab
ility distribut
io
n of the PV output scene i
s
divided i
n
to
m
equal p
r
ob
abi
lity interval.
2)
Any equal probability interval: from
(1
)
/
,
/
1
me
m
e
m
e
m
ran
dom extra
c
te
d a
numbe
r
m
p
,
m
p
ca
n be expre
s
sed
as:
1
m
ri
p
mm
(9)
In which r is t
he random variables of
equal probability distribution in [0,1] .
3)
We u
s
e
the i
n
verse tra
n
sf
ormatio
n
of
PV forecast
output di
strib
u
tion fun
c
tion
, and
got the probability interval
(1
)
/
,
/
em
e
m
of the PV output sampl
e
s,
namely:
1
..
()
P
it
m
PF
p
(10)
3.2. PV Scenes Output
Based on the
Technolo
g
y
of Scene
Re
duce
By Latin hypercub
e
sa
mpling, with
t
he co
rrespondi
ng t a
t
a ce
rtain
moment,
photovoltai
c
output scen
a
r
ios
we
re ma
ny, forming
n
u
merou
s
sce
ne tree; if we
did not pro
c
ess
the scen
e, a
nd the
comp
uter would fa
ce h
uge a
m
o
unt of cal
c
ula
t
ion. So this
pape
r u
s
ing t
he
scene
red
u
ci
ng technol
og
y to redu
ce
scen
e, we
use
d
the sce
ne
whi
c
h h
ad h
a
d
re
du
ced i
n
stead
of multiple scenari
o
s,
thu
s
forming
a finite numb
e
r
of PV output col
l
ection [7]. Hy
pothe
sized th
e
output
scena
rio thro
ugh
La
tin sq
uare
sa
mpling
wa
s
m, red
u
ced it
s
scena
rio fo
r n. Scene
re
d
u
ce
at a certain m
o
ment sp
ecifi
c
step
s were
as follo
ws:
a) Assu
ming
lm
,
l
i
s
the
num
be
r of sce
nari
o
s
whi
c
h
need
t
o
re
du
ce.
Cal
c
ulate
any
time the Kan
t
orovich
dista
n
ce
of
..
.
.
,
P
jt
P
k
t
PP
of th
e two
j
,
k
scenari
o
,
,
j
lk
l
.This
article u
s
e
d
the Kantorovich distan
ce a
s
follows:
..
.
.
..
.
.
(,
)
k
P
jt
Pk
t
P
jt
Pk
t
dP
P
P
P
(
1
1
)
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
A Clean Econ
om
ic Disp
atch of PV Energy St
ora
ge
Conne
cted
to Grid… (Liu Ji
ang-Tao
)
7617
b)
For every scene j, we we
re found
the
output scena
rio
..
P
kt
P
which is the sho
r
te
st
distan
ce
fro
m
the
outp
u
t scen
ario
..
P
jt
P
, and
namely
..
.
.
mi
n
(
,
)
,
kp
j
t
P
k
t
dP
P
j
k
,
s
e
t:
mi
n
.
.
.
.
mi
n
(
,
)
,
jk
p
j
t
P
k
t
dP
P
j
k
(12
)
c) Cal
c
ulate
.
mi
n
.
.
it
K
Dj
t
j
t
Pp
, wh
ere
.
jt
p
is the probability of
..
P
jt
P
.
d)
Rep
eat
step
s c, in
that all
the
i
K
D
P
, look
ing for the s
m
alles
t
..
K
Dit
P
. Mark
ed for
.
K
DS
t
P
.The ne
w
scene p
r
ob
abili
ty for
..
.
kt
j
t
kt
p
pp
, which
the output
scena
rio
s
s
j
P
need to con
c
entrate redu
ce.
e)
After a
scene
wa
s
red
u
ced
,
again i
n
a
step, wh
en
the
output
scena
rio
re
du
ced t
o
n
.We can b
e
con
c
lu
ded th
at the re
duce output
sce
nario
at t time whi
c
h
ha
s the
numbe
r of n.
4.
Ba
sed
on
the LHS - SR - GAMS P
V
Storage Connec
t
ed
-gri
d Energ
y
-sav
ing Economic
Optimiza
tion
Scheduling Model
4.1. Establis
h the Objec
t
i
v
e
Function
Due to
un
certainty of PV
output, the o
u
tput
may be
have many
scene
s, there
f
ore we
can
not u
s
e
the obj
ective
function
of a
singl
e
co
nfirmed
to
optim
ize
the
PV store hybrid po
wer
gene
ration
efficien
cy. This pape
r u
s
e
s
the expe
ct
ed obje
c
tive
fun
c
tion whi
c
h contain
s
ra
nd
om
variable
s
to descri
be the
probl
em more rea
s
on
able
and pra
c
tica
l. Combine
d
with the natio
nal
related
poli
c
y
in n
e
w ene
rgy po
wer ge
neratio
n to
e
n
co
ura
ge th
e
gen
eratio
n o
f
electri
c
ity, this
pape
r con
s
id
ers
ea
rning
s
of sell ele
c
tri
c
ity, benef
its o
f
PV storage j
o
int output d
e
v
iation po
sitive
and n
egative
deviation p
u
n
ishm
ent, fin
a
lly we al
so
con
s
id
ere
d
th
e ‘overflo
w n
egative ea
rni
ngs’
of the power grid to PV maximum given abilities.
.1
2
3
4
(,
)
PB
t
M
ax
E
i
P
R
R
R
R
(
1
4
)
12
1.
1
tP
B
t
R
MP
P
(15)
12
2.
.
.
.
.
1
(1
)
(
)
up
ti
t
i
j
o
t
P
B
t
i
t
i
R
MP
b
P
P
p
(16
)
12
3.
.
.
.
.
1
.(
)
dow
n
ti
t
P
B
t
i
j
o
t
i
t
i
R
MP
b
P
P
p
(17
)
4.
.
.
dr
o
i
d
r
o
t
RP
P
(18)
This a
r
ticle select
s the PV storag
e joint
powe
r
time is 8:00-17:59,
a total of 10
hours.
t
M
P
is the sell el
ectri
c
ity price
s
at t momen
t,
.
P
Bt
P
is the PV storage joi
n
t plan output;
up
t
M
P
positive devi
a
tion sell el
e
c
tri
c
ity price
s
, which
rep
r
ese
n
ts the b
enefits of PV stora
ge out
put
deviation;
down
t
MP
is negative d
e
viation pu
nish
price, wh
i
c
h
repre
s
e
n
ts the
benefit of P
V
output
negative d
e
viation pe
nalty. The
state o
f
PV output p
o
sitive an
d n
egative devia
tion is
.
it
b
; when
.
1
it
b
, whi
c
h
rep
r
e
s
ent
s the
stat
e of ne
gative
dev
iation. Th
e PV sto
r
ag
e
overflo
w
p
r
ice is
dro
P
,
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 11, Novem
ber 20
14: 76
13 – 762
1
7618
whi
c
h rep
r
e
s
ents
PV store
overflo
w
b
enefits.
Th
e i
t
h scen
e PV store joi
n
t p
o
we
r o
u
tput
at t
moment is
..
ij
o
t
P
. The ith scene
PV store outp
u
t overflow n
u
mbe
r
at t moment is
..
id
r
o
t
P
.
4.2. Establis
h the Co
nstr
aint Con
d
itio
n
Firstly, we n
e
eded
to the
p
o
we
r b
a
lan
c
e
co
nstraint b
e
twee
n PV a
nd
stora
ge
when PV
and sto
r
a
ge
power g
ene
ration sche
dul
ing. We al
so
need to
con
s
i
der the
re
stri
ction
s
of the PV
maximum gi
ven, whi
c
h
contai
ns th
e
ene
rgy
s
t
or
a
g
e
d
e
v
ic
e c
h
ar
ge
-
d
isch
a
r
ge
c
a
pa
city
con
s
trai
nts,
PV output constrai
nts, e
nergy storag
e device
ch
arge
-di
s
cha
r
g
e
sto
r
ed
en
ergy
balan
ce
con
s
traints in a cycle. And we
neede
d to consi
der the
combine
d
po
wer gen
eration
repo
rt
ca
pa
cit
y
const
r
aint
s
cau
s
e
d
by
it
s cap
a
cit
y
.
The output b
a
lan
c
e con
s
traint of PV storage joi
n
t po
wer g
ene
ratio
n
:
..
.
.
.
.
i
j
ot
P
i
t
d
i
s
t
c
ht
PP
P
P
(
1
9
)
The re
stri
ctio
ns of the gr
id bigge
st
co
nsumption:
..
c
o
n
.
t
id
r
o
P
B
t
PP
P
(20)
1) The
con
s
traints of t
he re
port output
ca
pacity:
.
.
ma
x
.
ma
x
0
PB
t
P
d
i
s
PP
P
(
2
1
)
2) The power
capa
city con
s
trai
nts of en
ergy storag
e de
vices b
e
twe
e
n
the moment
:
1.
.
/
t
t
c
h
t
c
h
d
is
t
d
is
EE
P
P
(
2
2
)
3) The power
capa
city con
s
traint
s of en
ergy storag
e de
vices:
mi
n
m
a
x
t
EE
E
(23)
4) Powe
r re
stri
ctions of Ene
r
g
y
storage
cha
r
ge an
d disch
a
rge:
..
m
a
x
0
dis
t
dis
PP
(
2
4
)
..
m
a
x
0
ch
t
c
h
PP
(
2
5
)
5) Powe
r bala
n
ce con
s
trai
nts
in a cycle:
0
T
EE
(30)
In whic
h
.m
a
x
dis
P
is t
he maxim
u
m
ch
arg
e
p
o
wer of th
e e
n
e
rgy
storage
device; and
.m
a
x
ch
P
is the minim
u
m discha
rg
e
powe
r
of the energy storag
e device.
5. Numerical
Examples Validate
This a
r
ticle
u
s
ed
a PV po
wer
station
of the Xinjiang
regio
n
, whi
c
h
PV powe
r
ge
neratio
n
cap
a
city is 5
0
mw. PV power
station b
a
se
d on
feed
back neu
ral
netwo
rk p
r
edi
ction sy
stem, and
its fore
ca
st in
terval time is
10 minute
s
.
PV powe
r
pla
n
t put into op
eration fo
r 4
years, a
nd it
has
the rich lo
cal
meteorologi
cal data.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
A Clean Econ
om
ic Disp
atch of PV Energy St
ora
ge
Conne
cted
to Grid… (Liu Ji
ang-Tao
)
7619
5.1.1. Analy
s
is Prediction
Error Distri
bution Char
acte
r
istics
We colle
cted actual data
a
nd
p
r
edi
cted data
for
3 ye
ars of a
ce
rtai
n PV power
station in
Xinjiang, its f
o
re
ca
st time
wa
s 10 mi
nut
es, a total
sa
mples
we
re 3
*365 * 1
0
*6,
then we written
the pro
g
ram
about the
pro
bability den
sit
y
estimation
method b
a
se
d on m
a
tlab,
by whi
c
h
we
can
solve the prediction error. The proba
bility distribution
solved as shown.
Figure 2. PV
Predi
ction Error Di
strib
u
tio
n
Diag
ram
From th
e fig
u
re
we
co
uld
kno
w
parag
raph
s
pre
d
ict
i
on e
rro
r di
st
ribution
app
roximate
symmetri
c
al, and
the outsi
de
line had n
o
rmally
di
st
ri
buted
cha
r
a
c
teristics. Th
u
s
we con
c
lud
ed
that the pre
d
iction
erro
r had
norm
a
l
l
y distrib
u
te
d
cha
r
a
c
teri
sti
cs. T
h
ro
ugh
cal
c
ulatio
n, the
predi
ction e
r
ror of a ce
rtain
PV power
stati
on in Xinjia
ng of this pap
er ado
pted was.
()
(
0
,
0
.
1
4
)
e
rro
r
t
5.1.2. PV Sc
ene Ou
tput
Sampling and Redu
ce
We con
c
lud
e
d
the distri
but
ion of PV output pre
d
ictio
n
erro
r from th
e front se
ctio
n, then
we
sa
mpled
t
o
p
r
edi
ction
error sce
ne
b
a
se
d o
n
L
H
S
tech
nolo
g
y.
The
sam
p
ling
num
ber of PV
predi
ction
error sce
ne i
s
2
000, an
d co
mbining
wi
th
the PV forecast value
of the PV fore
ca
st
system, acco
rding
to
the
type
eight, we
c
ould
get PV
output scene
s.
Repe
a
t
ed
the abov
e
operation, we
could g
e
t PV output scena
rio of 10 hou
rs.
In orde
r to m
a
ke th
e cal
c
u
l
ation efficien
cy
, we
writte
n the progra
m
of PV output scene
redu
ce
ba
se
d on matla
b
. Unlimited
sce
ne re
du
ce
would lea
d
to the fitting deg
ree i
s
not hi
g
h
,
and we ca
nn
ot blindly pursue the fit of the output e
rror, this will m
a
ke the
cal
c
ul
ation efficien
cy is
not high.
The
ratio of the area
pre
d
icti
on erro
r dist
ri
bution curve
and ab
scissa
when
red
u
ci
ng
before
and af
ter wa
s fitting
. So we first
cho
s
e
the
re
duce num
ber, which i
s
: 15
, 25, 35, 45, 55,
65.Fitting in the followi
ng
Table 1.
Table 1. Storage Capa
city Config
uratio
n
under
Differe
nt Numbe
r
ab
out Red
u
ced
PV Output
Scene
s
Scenes
number
15 25
35 45
65
Fitting
0.68 0.74
0.85 0.87
0.9
Table 2. PV Output Sce
n
e
s
at 14:00 on
July 25 20
13
scenes PV
output
(
MW
)
probabilit
y
1 3.3
0.0003
2 6.5
0.0027
3 14.5
0.0365
4 22.7
0.0523
5 26.1
0.1264
6 27.8
0.1153
7 31.5
0.1032
8 35.3
0.1026
9 37.6
0.1186
10 40.7
0.0733
11 42.8
0.1212
12 45.3
0.0727
13 47.8
0.0726
14 49.3
0.0021
15 50
0.0002
-1
.
5
-1
-0
.
5
0
0.
5
1
1.
5
0
0.
5
1
P
V
f
o
r
e
ca
st
i
n
g
o
u
t
p
u
t
e
r
r
o
r
PD
F
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 11, Novem
ber 20
14: 76
13 – 762
1
7620
From the a
b
o
v
e table we could kno
w
when
the num
ber of re
du
ce
sce
ne
s wa
s
65, the
fitting was th
e highe
st, bu
t it would affect t
he comp
utational efficiency. When
the numb
e
r
of
redu
ce
scen
es
wa
s 15,
a
nd the
fit tin
g
was 0.6
8
. If the n
u
mbe
r
of
redu
ce
scen
es was
1
5
,
it
woul
d greatly improve th
e
comp
utationa
l efficien
cy
. So this p
ape
r
selecte
d
15
as the num
ber
of
redu
ce
scen
es. We
sele
cted fore
ca
st
numbe
r of 3
5
.3 MW
of a
PV powe
r
station in Xinji
ang
in14:00,
July
25, 20
13,
which
produ
ce
d the n
u
mb
e
r
of 2
000
out
put sce
n
e
s
b
y
LHS, then
we
use
d
the
SR
techn
o
logy to
re
du
ce
outp
u
t scen
e. Th
e nu
mbe
r
of
15 p
r
o
d
u
c
ed
scena
rio
s
out
put
sho
w
n in the
followin
g
table.
5.2. The Implementa
tion
Bas
e
d on G
A
MS Combi
n
ed PV and Store Energ
y
-sa
v
ing and
Economy
Dispatching
In order to i
m
pleme
n
t th
e meth
od
of
co
mbin
ed
PV and
sto
r
e en
ergy-saving a
n
d
eco
nomy
system propo
se
d by thi
s
pap
er,
we
esta
bli
s
he
d the
mo
del a
nd
written p
r
og
ram
ba
se
on GAMS.
We cho
s
e
DICOPT alg
o
rith
m a
s
the
la
st algo
rithm.
T
he whol
e system
ru
nnin
g
t
i
me
wa
s 30 seco
nds. At 12:00-15:5
9
,
t
M
P
wa
s 0.8 yuan/kW h, at
9:00-11:5
9
and
16:00-17:59
electri
c
ity pri
c
e wa
s 0.55 y
uan/kW. H.
up
t
M
P
and
down
t
MP
were
0.2 times of
t
M
P
. The pri
c
e o
f
dro
P
was 0.0
5
yuan/kW. J h, PV storag
e rel
a
ted paramet
ers a
s
sho
w
n
in Table 3.
Table 3. Rel
e
vant Paramet
e
rs a
bout PV and Stora
g
e
Parameter name
Parameter
v
a
l
u
e
PV capacity
50MW
Capacit
y
of ene
r
g
y
storage device
35MW.h
.m
a
x
P
P
50MW
.m
a
x
di
s
P
30MW
.m
a
x
ch
P
29MW
ch
0.7
di
s
0.69
mi
n
E
0
ma
x
E
50MW.h
We inp
u
t the above data, and colle
cted
the PV
forecast data, and
we co
uld en
ter the
disp
atchi
ng p
r
og
ram. Thi
s
pape
r ch
ose a PV power
station in Xinj
iang on
July 26, 2013 a
s
the
sched
uled ti
me. PV output predi
ctive value as
s
h
o
w
n in Fig
u
re 3. Combi
ned
PV and storage
output dispat
ch value
s
a
s
sho
w
n in Fig
u
re 4.
Figure 3. PV
Output Forecasting Valu
e
Diag
ram
Figure 4. PV
and Stora
ge
Planned O
u
tp
ut
Diag
ram
6. Conclusio
n
1)
This pa
per
co
nclu
ded that PV output predictio
n er
ror had cha
r
a
c
te
ristic of no
rm
al distrib
u
tion
.
0
10
20
30
40
50
60
0
10
20
30
40
50
60
t /d
is
p
a
tc
h
i
n
g
p
o
i
n
t
s
PV pr
ed
i
c
t
out
pu
t
/
M
W
0
10
20
30
40
50
60
-3
0
-2
0
-1
0
0
10
20
30
40
50
60
t /d
i
s
p
a
tch
i
n
g
po
in
ts
P
V
-
S
hy
br
i
d
sy
st
em
out
put
/
M
W
E
n
er
g
y
s
t
or
a
g
e
pl
a
n
n
e
d
ou
t
p
u
t
P
V
s
t
o
r
ag
e
hyb
r
i
d
p
l
a
nne
d
o
u
t
p
ut
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
A Clean Econ
om
ic Disp
atch of PV Energy St
ora
ge
Conne
cted
to Grid… (Liu Ji
ang-Tao
)
7621
This p
r
ovide
d
the basi
s
fo
r further
re
sea
r
ch PV effect
on po
we
r gri
d
and related
studies. Th
e PV
output u
n
ce
rtainty probl
e
m
is
co
nverted into
effe
ctive PV out
put sce
ne
with LHS
-
RS
techn
o
logy.
This provide
d
an
effecti
v
e sol
u
tion
to
solve
the
problem
of
multi-scen
ari
o
uncertainty.
2)
Relative to t
he en
ergy
-sa
v
ing model
o
f
the individu
al uses
ene
rgy storage pl
aned
output
con
n
e
c
ted to
g
r
id,
based
on th
e
tech
nolo
g
y
of LHS
-
SR-G
AMS, the co
mbined
en
erg
y
stora
ge an
d PV model planned outp
u
t increa
sed by
8%, and to a certain exte
nt, improved the
PV output pre
d
iction a
c
curacy.
Referen
ces
[1]
Research on
several critic
a
l
prob
lem
of PV grid-c
on
ne
cted
ge
ner
atio
n s
y
stem Po
w
e
r S
y
ste
m
Predicti
on cont
rol. 201
0; 38(2
1
): 209-2
14.
[2]
Yuan T
i
ej
ian
g
, Cha
o
Qin, T
ong F
e
i. An
envir
onme
n
tal/
econ
omic
disp
atch mod
e
l f
o
r po
w
e
r gr
i
d
contai
nin
g
w
i
n
d
po
w
e
r g
e
n
e
r
a
tion
units
an
d its
simu
latio
n
in electric
it
y
market
env
iron
ment.
Power
System
Technology
. 200
9; 33(6): 67-7
1
(in
Chin
ese).
[3]
Yuan T
i
eji
a
n
g
, Cha
o
Qin.
Optimi
z
e
d
Econ
omic an
d Env
i
ron
m
e
n
t-friend
l
y
Dispatc
h
in
g
Mode
lin
g for
Larg
e
-scal
e
W
i
nd Pow
e
r Integ
r
ation
. Proce
e
d
i
ngs of the CS
EE. 2010; 30(
3
1
): 7-13.
[4]
W
U
Xi
on
g, W
A
NG Xi
ul
i.
A Joint Oper
atio
n Mod
e
l
and
Soluti
on for
H
y
brid W
i
nd En
ergy Stora
g
e
System
s.
Proc
eed
ings
of the CSEE. 2013; 3
3
(13): 10-
17.
[5]
Hag
h
ig
hatH,
SeiH, K
i
anA
R
.On the s
e
lf-
sched
ul
i
n
g
of
a
po
w
e
r
pr
oduc
er i
n
un
certain
tradi
ng
envir
onme
n
t.
Electric Power SystRes (EPSR).
2008; 78: 31
1
-
317.
[6]
Usaola J, Angarita J.
Bi
ddi
ng
w
i
nd e
ner
gy u
nder
unc
ertaint
y
.
In Presente
d
at Internati
o
n
a
l
Confer
enc
e
in Cle
an El
ectri
c
al Po
w
e
r, ICCEP: 2007; 75
4-
9.
[7]
Matevos
y
a
n
J, Sder L. Min
i
mizatio
n
of im
bal
ance c
o
st tradi
ng
w
i
nd
p
o
w
e
r th
e shor
t-term po
w
e
r
market.
IEEE Trans PowerSys
t.
2006; 21: 13
96-4
04.
[8]
CHEN
Can, W
U
Wench
u
a
n
.
An Active
Dist
r
ibuti
on Syste
m
Rel
i
ab
ility E
v
alu
a
tion
Meth
od B
a
sed
o
n
Mu
l
t
i
p
le
Sce
nari
o
s Te
ch
ni
q
ue.
Proceedi
ngs
of the C
SEE. 2012; 32(
34): 67
-73.
[9]
Morales JM, Pineda S, Co
n
e
j
o
AJ. Sce
nari
o
red
u
ction
for f
u
tures m
a
rket t
r
adi
ng
in
el
ectricit
y markets.
IEEE Transactions on Power
System
s
. 20
09
; 24(2): 878-8
8
8
.
[10]
Grow
e-Kusk
aN, Heitsch H, Romisch W.
Sce
nari
o
red
u
ction
and sce
nar
io t
r
ee co
nstructio
n
for pow
er
ma
na
ge
me
nt p
r
obl
ems.
Proc
e
edi
ngs of the IEEE Po
w
e
r T
e
ch C
onfer
enc
e. Bolog
na, Ital
y
,
IEEE. 2003:
23-2
6
.
[11]
Bathurst GN,
Strbac G. Valu
e of com
b
in
i
n
g
ener
g
y
stora
g
e
an
d
w
i
n
d
in
short term e
n
e
r
g
y
.
El
e
c
tr
i
c
Power SystRes (EPSR)
. 200
3; 67: 1-8.
[12]
Badru
l
H. Cho
w
d
h
u
r
y
, Sri
n
iv
as Che
lla
pil
l
a
.
Doub
le-fe
d
i
nducti
on g
e
n
e
rator control for
varia
b
le sp
ee
d
w
i
nd po
w
e
r
ge
nerati
on.
Electr
ic Pow
e
r Systems Res
earc
h
.
200
6: 76(1
2
): 786-8
0
0
.
[13]
Sloot
w
e
g JG,
Klin
g W
L
. M
ode
lin
g
w
i
n
d
t
u
rbi
nes for
p
o
w
e
r
s
y
stem
d
y
n
a
mics s
i
mul
a
tions.
Win
d
Engi
neer
in
g
. 2004; 28(
1): 7-2
6
.
[14]
Sun T
,
Chen Z
,
Blaab
jer
g
F
.
T
r
ans
i
ent stabi
lit
y of DF
IG W
i
nd T
u
rbines at a
n
e
x
terna
l
shor
t-circuit fault.
W
i
nd Ener
gy
. 200
5; (8): 345-
360
.
[15]
Holds
w
o
r
th
L,
Char
alam
bous
I, Ekana
yak
e
JB, et al
. Po
wer s
y
stem fau
l
t ride
throu
gh c
apa
bil
i
ties
o
f
ind
u
ction
gen
e
r
ator base
d
w
i
nd turbi
nes.
W
i
nd Eng
i
n
eeri
n
g
. 2004: 28(
4): 399–
40
9
.
[16]
A T
apia, G T
a
pia, J
X
Ostol
a
za, et al.
M
o
deli
n
g
an
d co
ntrol of
a
w
i
nd
turbin
e dr
ive
n
DF
IG.
IEEE
T
r
ans. Energy
Conv
ers.
200
3; 18(2): 194
–2
0
4
.
Evaluation Warning : The document was created with Spire.PDF for Python.