TELKOM
NIKA Indonesia
n
Journal of
Electrical En
gineering
Vol.12, No.5, May 2014, pp
. 3357 ~ 33
6
5
DOI: http://dx.doi.org/10.11591/telkomni
ka.v12i5.4937
3357
Re
cei
v
ed O
c
t
ober 2
0
, 201
3; Revi
se
d Decem
b
e
r
1, 2013; Accepte
d
De
cem
ber
20, 2013
Unit Commitment with Battery Energy Storage
Considering Wind Forecast E
rror
Cai Zhi*
1
, Zeng Lili
1
, Zha
o Kun
1
, Men De
y
u
e
1
, Xu
Dan
1
, Dai Sai
1
, Zhao Xu
2
1
Chin
a Electric
Po
w
e
r R
e
sear
ch Institute, Beijin
g, Chi
n
a
2
Heilo
ng
jia
ng E
l
ectric Po
w
e
r C
o
mpa
n
y
Limite
d, Heil
ong
jia
ng
, China
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: caizhi@
epr
i.sgcc.com.cn
A
b
st
r
a
ct
T
he integr
atio
n of w
i
nd reso
urce into th
e electric
gr
id br
ings si
gnific
ant
chall
eng
es du
e to th
e
varia
b
le
n
a
ture
an
d a
n
ti-p
eak
-regul
atio
n c
h
a
r
acteristic
of w
i
nd
pow
er. Bas
ed
on
le
ast sq
uare
metho
d
,
a
n
improve
d
nor
mal distrib
u
tio
n
mo
de
l is prop
osed to fi
t the actual w
i
nd p
o
w
e
r forcast error. F
u
rthermore
,
consi
deri
ng w
i
nd p
o
w
e
r fore
cast error
and
the gr
eat
p
o
t
entia
l of b
a
tter
y
ener
gy stora
ge syste
m
(BE
SS)
techno
lo
gy to miti
gate the i
m
pact of volatil
e
w
i
nd pow
er, a unit co
mmit
me
nt (UC) mod
e
l
w
i
th large cap
a
c
ity
BESS has
been estanbis
hed in this st
udy. Case studies wi
t
h
modified IEEE 39-bus system
ar
e employ
ed
to vali
date th
e
prop
osed
meth
od. The ro
le
of BESS on
ec
o
n
o
m
ics, p
eak l
oad s
h
iftin
g
an
d acco
mmodati
n
g
w
i
nd pow
er is discuss
ed.
Ke
y
w
ords
: uni
t commit
ment, w
i
nd pow
er
, forecast error, bat
tery energy sto
r
age syste
m
(B
ESS)
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
Wind p
o
wer
has g
r
o
w
n si
gnifica
ntly in Ch
ina in re
cent years f
o
r environme
n
tal and
su
staina
ble p
u
rpo
s
e. Mainl
and Chi
na h
a
s add
ed
win
d
en
ergy
ca
p
a
city 12
960M
W in
20
12, u
p
20.8%, and t
he total win
d
energy ca
pa
city has
re
a
c
hed 7
5324.2
M
W. In contrast to the ra
pid
developm
ent
of win
d
en
e
r
gy capa
city, the a
c
co
mm
odation
of wi
nd po
we
r i
s
relatively limi
t
ed.
Due to the u
n
ce
rtainty ch
ara
c
teri
stic of
wind po
w
e
r
,
th
e
p
o
w
er
gr
id
fa
c
e
s
gr
ea
t c
h
a
lle
ng
es i
f
large
-
scale
wind g
ene
rat
i
ons
are i
n
te
grated. Be
si
des,
wind p
o
we
r ha
s a
n
t
i-pea
k-reg
u
la
tion
cha
r
a
c
teri
stic, esp
e
ci
ally in wi
nter,
win
d
po
we
r h
a
s
to be
cu
rtaile
d when
co
nventional
thermal
power in
crea
se
s for heatin
g system.
Energy
storage sy
stem (E
SS)
is
considered to
be a
good
option t
o
undertake t
he tasks
of peak loa
d
shifting an
d suppo
rt the wi
nd po
we
r p
e
netration. Co
nsid
erin
g that energy sto
r
a
ge
techn
o
logie
s
can help
th
e power syste
m
to
ac
co
mm
odate m
o
re
wind po
we
r, th
ey have
com
e
to
the attention
of all the wo
rld [1]. Energ
y
st
orag
e
technolo
g
y
inclu
des pump
ed hydro storag
e
system, batt
e
ry energy
stor
age
system (BESS), compress
ed ai
r
storage, flywheel
,
sup
e
rcap
acit
or an
d so
on.
Among
all feasibl
e
ene
rgy
stora
ge tech
nologi
es, batt
e
ry syste
m
s
are
the mos
t
widely us
ed energy
s
t
orage devic
e
[2,
3]. BESS
t
e
c
h
nologies
aim to trans
f
orm
electri
c
ity int
o
chemi
c
al f
o
rm
of en
ergy, whi
c
h i
s
sto
r
ed
and
afterwa
r
d
s
converted
ba
ck to
electri
c
ity, su
ch
as conve
n
tional b
a
tteries
(Li
-
ion, P
b
-Aci
d), hi
gh
-tempe
ratu
re
batterie
s
(NaS,
ZEBRA)
and
flow batteries (VRB, PSB, ZnBr).
Comparing
with pump hydro s
t
orage, BESS is
more exp
e
n
s
ive. Howeve
r, in some pl
ace
s
where don’t have water co
nditio
n
to build pu
mp
hydro sy
stem
, large-scale
BESS is a u
nnegli
g
ible
al
ternative choi
ce. There are already some
s
u
c
c
ess
f
ul applications
of
BESS in different
c
o
untries
s
u
c
h
as
Cas
t
le Valley
Americ
a, King
Island Au
stral
i
a and Shan
g
hai Chi
na.
Some researches on power
system techonol
ogies
wi
th wi
nd power and
ESS have been
carrie
d out. A SCUC formation emp
h
a
si
zing o
n
wind po
wer a
nd CAES is pre
s
ente
d
in
[4].
Garcia
-Go
n
zalez et al.
[5] investig
ate t
he im
pact
of pumpe
d-storage on sy
ste
m
with
hig
h
win
d
penetration.
Rodi
ca
Loise
l [6] propo
se
s a te
chni
cal
-
econo
mic
a
s
sessme
nt o
f
a large
-
sca
l
e
storage facility. In curre
nt power grid, BESS is usuall
y
utilized
in small scale and com
b
ined wit
h
wind
gene
rati
ons. In thi
s
case, thi
s
sm
al
l-scale
ene
rg
y storag
e is
consi
dered d
e
pend
ent on
wind
unit and
not
modele
d
in
u
n
it com
m
itme
nt. This
pap
e
r
focuses on
indep
end
ent
larg
e
capa
ci
ty
BESS which is
s
u
itable for
places where are not po
ss
ible to build pump
hydro
s
t
orage
s
y
s
t
ems
.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 5, May 2014: 3357 – 33
65
3358
This paper presents
a uni
t commi
tment model considering larg
e capacity BESS as it might
become a de
velopment tre
nd in the future.
The
remai
n
d
e
r of thi
s
pa
per i
s
org
ani
zed
a
s
follo
ws: An imp
r
ov
ed
wind fo
re
ca
st erro
r
model i
s
proposed in Section 2.
In Section 3, characteri
stics of
BESS are analyzed and
the
establi
s
hm
en
t of unit
com
m
itment mo
d
e
l con
s
ide
r
in
g wi
nd
po
we
r forecast
error i
s
i
n
trod
uced.
Ca
se
s with
1
0
units
and 1
00 unit
s
are
studie
d
an
d analyzed in
Section 4. Se
ction 5
dra
w
s the
con
c
lu
sio
n
s.
2. Impro
v
ed
Wind Fore
c
a
s
t Error Mod
e
l
In win
d
fo
re
cast e
r
ror mo
deling
re
se
arch fiel
d, the
norm
a
l di
strib
u
tion fun
c
tion
is
mo
st
comm
only ap
plied [7, 8]. The pro
bability
density functi
on ca
n be ex
pre
s
sed a
s
:
2
2
()
2
1
()
2
x
fx
e
(1)
Whe
r
e
is the expected val
ue of wind foreca
st error
x
,
is the stand
ard dev
iation, it shows
the degree of
deviation fro
m
the expect
ed value.
Figure 1
is a
diagram
of EI
RG
RID
201
0.2~2
010.1
1
wi
nd p
o
wer fo
reca
st e
r
ror fit
t
ed by
norm
a
l di
strib
u
tion. The va
lues
of erro
rs are
ex
pressed a
s
p
e
rcen
tages
of wi
nd
cap
a
city. Th
e
model
gen
erally suits the
actual
value
s
, however,
wi
thin 0%~5%
actual
value
s
are
hig
her th
an
norm
a
l den
si
ty function value, whil
e actual de
nsity
values are lowe
r within
-10%
~0% an
d
5%~20%. To
a certai
n extent, this norma
l functi
on exa
ggerates the
wind p
o
wer p
r
edi
ction e
rro
r.
Figure 1. Dia
g
ram of Fo
re
ca
st Erro
r Fi
tted by Normal
Distrib
u
tion (EIRGRI
D
)
In order to i
m
prove
the
accuracy
of
nor
m
a
l di
stri
bution
model
, an im
prove
d
de
nsit
y
function i
s
propo
sed:
2
2
()
2(
/
)
1
()
(
0
)
2(
/
)
x
a
gx
e
a
a
(2)
We
can
obta
i
n the valu
e
of variabl
e
a
to get a
more
suitabl
e
sta
ndard d
e
viation by
followin
g
next steps:
Step 1:
Win
d
po
we
r fo
reca
st e
r
ror
stand
ard
d
e
viation
an
d
expectatio
n
are
cal
c
ulate
d
ba
sed o
n
histo
r
i
c
data
s
.
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
0
2
4
6
8
10
12
Wind p
o
w
e
r f
o
re
ca
st er
ror
(p
e
r
centag
e of wind capa
city)
Probability densit
y
Actual proba
blity
den
sity values
No
rmal de
nsi
t
y
funciton valu
es
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Unit Com
m
i
tm
ent with Battery Ene
r
g
y
Storage
Co
n
s
i
derin
g Win
d
Fore
ca
st Erro
r (Cai Zhi)
3359
Step 2: The
amount of
actual errors
b
e
y
ond a
cert
ain bou
nda
ry (su
c
h a
s
±
30% in this
ca
se) is
sig
n
ificantly small,
so
we
ca
n se
t bound
ary in
this mo
del. A
s
suming
that
maximum e
r
ror
is
Ub
and lo
we
r minim
u
m e
rro
r i
s
Lb
, and symmetrically
Ub
L
b
, then
we
ca
n divide th
e
erros
into
Tz
o
n
e
intervals:
[,
]
,
(
,
2
]
,
,
(
,
]
Ub
L
b
U
b
L
b
Ub
L
b
Ub
L
b
L
bL
b
L
b
L
b
U
b
U
b
T
z
on
e
T
zo
n
e
T
z
on
e
T
zo
ne
Each
erro
r b
e
l
ong
s to
one
of the i
n
terval
s o
b
tain
a
ne
w valu
e. Th
e
values in th
e
Tz
o
n
e
internal
s a
r
e:
13
(
)
1
,,
,
22
2
Ub
L
b
Ub
L
b
Ub
L
b
Lb
Lb
U
b
Tzone
Tzo
n
e
T
zone
Step 3: The
logarith
m
ic form of the Equation (2
)
:
2
2
2
ln
(
)
l
n
(
)
(
)
2
2
aa
gx
x
(3)
A
ssu
ming
ln
(
)
Yg
x
,
2
()
Xx
,
ln
(
)
2
a
B
,
2
2
2
a
A
, then
YB
A
X
.
Lea
st sq
ua
re
method
ca
n
be u
s
ed to
o
b
tain the valu
e of
A
and
B
. We use
A
to identify the
value of
a
as
A
has hig
her
reli
ability than
B
.
Step 4: Each
Tz
o
n
e
ha
s a
corresp
ondi
ng
()
g
x
,
we
can
s
e
le
ct
t
he b
e
st
Tz
o
n
e
by
cal
c
ulatin
g the expre
ssi
on
as follo
ws:
()
1
21
()
(
)
2
Tz
o
n
e
i
gx
i
T
z
one
P
iU
b
L
b
E
r
ro
r
T
zo
n
e
g
L
b
U
b
L
b
T
z
one
(4)
()
1
21
()
2
Tz
o
n
e
i
fx
i
Tz
o
n
e
P
iU
b
L
b
E
rro
r
f
L
b
Ub
Lb
Tz
o
n
e
(5)
()
()
()
(
)
10
0%
fx
g
x
fx
E
rro
r
E
rror
ERR
Tz
one
Er
ror
(6)
W
h
er
e
i
P
indi
cate
s the
p
r
obability in
i
n
terval
((
1
)
,
]
Ub
Lb
U
b
Lb
Lb
i
L
b
i
T
z
one
T
z
on
e
. The
maximum
()
E
RR
Tz
one
indicate
s the b
e
st
Tzon
e
.
In EIRGRID
ca
se,
0.
02
03
,
0.06
67
. Followin
g
the above step
s, we obtain
80
Tzo
n
e
,
1.
2
6
5
a
,
()
(
)
20.
17%
gx
E
rro
r
T
zo
n
e
. The a
c
tu
al values, i
m
poved a
n
d
origin
al
curve
s
a
r
e shown in Figu
re 2. The improved
wind
power fore
ca
st error curv
e is clo
s
e
r
to the
actual p
r
ob
ab
ility distributio
n than origi
n
a
l
model
, whi
c
h verifies the
validity of proposed metho
d
.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 5, May 2014: 3357 – 33
65
3360
Figure 2. Foreca
st Error Fi
tted with Initial Norm
al Di
stribution an
d Improve
d
No
rmal Dist
ributi
o
n
3. UC Forma
tion
The m
a
in
UC mo
del i
s
formulate
d
a
s
a
n
optimi
z
ation p
r
obl
e
m
that mini
mize
s the
obje
c
tive function con
s
trai
n
ed by system
requi
reme
nt
con
s
id
erin
g wind po
wer fo
reca
st error.
(1) O
b
je
ctive Functio
n
The obje
c
tive
function is ex
pre
s
sed a
s
follows:
,
,
,
,
,,
,,
1
1
11
11
mi
n
[
(
)
]
(
)
NG
H
W
H
W
H
f
o
rec
a
st
c
i
ih
ih
ih
m
h
f
m
h
f
m
h
i
h
mh
mh
FP
I
S
U
M
q
N
P
P
(7)
Whe
r
e
,
()
ci
i
h
FP
indica
tes therm
a
l unit
i
operating co
st at time
h
;
,
ih
I
indicat
e
s the st
atu
s
thermal
unit
i
;
,
ih
P
indicates the a
c
tive po
wer of the
r
m
a
l unit
i
at time
h
;
,
ih
SU
indicat
e
s
startup
cost
o
f
therm
a
l uni
t
i
at time
h
;
,,
f
ore
c
ast
fm
h
P
indicate
s th
e fo
recast val
ue
of win
d
unit
m
at
time
h
;
,,
f
mh
P
indicate
s the
sche
dul
e value
of wi
nd unit
m
at tim
e
h
;
M
indi
cate
s
the weight of
wind po
we
r fore
ca
st deviation penalty
function;
N
indicate
s the
weig
ht of wind cu
rtailme
n
t
penalty functi
on;
,
mh
q
is wind p
o
we
r deviatio
n
indicator ex
pre
s
sed a
s
follows:
,,
,,
,
,
,
,
,,
f
o
rec
a
s
t
ac
t
ual
f
mh
f
m
h
w
i
n
d
m
h
m
C
a
p
mh
wi
n
d
m
h
PP
P
q
(8)
Whe
r
e
,,
actua
l
f
mh
P
indi
cate
s the
a
c
tual
po
wer of wi
nd u
n
it
m
at tim
e
h
;
,,
wi
nd
m
h
and
,,
(/
)
wi
nd
m
h
a
indi
cate
exp
e
ctation
an
d i
m
prove
d
stan
dard
d
e
viation of
wind
u
n
it
m
forecast
relativ
e
er
ro
r at time
h
, res
p
ec
tively;
,
mC
a
p
P
indi
cate
s the ca
p
a
city of wind
unit
m
.
If actual po
wer of
wind
uni
t is greater th
an fore
ca
st v
a
lue,
,
mh
q
is written as
,
,
up
a
c
tu
al
mh
q
.
If
actual p
o
wer
of wind unit is smalle
r than
forecast valu
e,
,
mh
q
is written
as
,
,
do
wn
a
c
t
u
al
mh
q
.
(2) System
Constraints
Therm
a
l unit cap
a
city co
nstraints:
,m
i
n
,
,
,
,
,m
a
x
,
(
1
...
;
1
.
.
.
)
i
i
h
i
hi
h
i
hi
i
h
PI
P
L
P
P
U
P
I
i
N
G
h
H
(9)
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0
2
4
6
8
10
Wind p
o
w
e
r f
o
re
ca
st er
ror
(p
e
r
centag
e of wind capa
city))
Probability densit
y
Actual value
s
Improved m
o
del
Initial model
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Unit Com
m
i
tm
ent with Battery Ene
r
g
y
Storage
Co
n
s
i
derin
g Win
d
Fore
ca
st Erro
r (Cai Zhi)
3361
Dema
nd bal
a
n
ce
con
s
trai
n
t
s:
,,
,
,
,
,
,
11
1
,
,,
,
,
,
,
,
,
,
,
11
1
,
,,
,
,
,
,
,
,
,
,
11
1
()
()
(1
,
NG
W
S
ih
i
h
f
m
h
s
t
o
r
s
h
D
h
im
s
NG
W
S
down
a
c
t
ual
i
h
i
h
f
m
h
m
h
w
in
d
m
h
s
to
r
s
h
D
h
im
s
NG
W
S
up
ac
t
ual
i
h
i
h
f
m
h
m
h
w
in
d
m
h
s
to
r
s
h
D
h
im
s
PI
P
P
P
PU
I
P
q
P
P
PL
I
P
q
P
P
h
2
...
)
H
(10)
Rampi
ng con
s
traint
s:
,,
1
,
,
1
,
,
1
,
m
i
n
,1
,
,
1
,
,1
,
,
m
i
n
[1
(1
)
]
(1
)
[1
(1
)
]
(1
)
(
1
...
;
1
...
)
ih
ih
ih
ih
i
i
h
i
h
i
ih
ih
ih
i
h
i
i
h
i
h
i
P
U
P
L
II
U
R
II
P
PU
PL
I
I
DR
I
I
P
iN
G
h
H
(11)
Reg
u
lation ca
pacity
co
nstraints:
,m
a
x
,
,
,
,
,
,
11
,,
m
i
n
,
,
,
,
,
11
(
)
(
1
,
2
...
)
(
)
(
1
,
2
...
)
NG
S
ii
h
i
h
s
t
o
r
u
p
s
h
u
p
h
is
NG
S
i
h
i
i
h
s
tor
d
o
w
n
s
h
dow
n
h
is
PP
U
I
R
R
h
H
PL
P
I
R
R
h
H
(12)
Line flow
con
s
traint
s:
,,
m
i
n
,
,
,
,
m
a
x
(
1
...
;
1
...
)
L
i
n
e
l
L
ine
l
h
L
i
ne
l
PP
P
l
L
h
H
(13)
Wind p
o
wer constraints:
,,
,,
,
,,
,
,
,
,
,,
,
,
,
,
0
0
(
1
...
;
1
...
)
f
o
re
cast
fm
h
f
m
h
do
w
n
ac
tu
al
fm
h
m
h
w
i
n
d
m
h
up
act
ual
f
m
h
mh
w
i
n
d
mh
m
c
a
p
PP
Pq
Pq
P
mW
h
H
(14)
Wind p
o
wer d
e
viation co
nst
r
aints:
,
,
,
,
0
0
(1
.
.
.
;
1
.
.
.
)
up
a
c
tual
m
h
given
dow
n
a
ctual
m
h
g
i
ven
qq
qq
mW
h
H
(15)
Whe
r
e
,m
i
n
i
P
and
,m
a
x
i
P
rep
r
e
s
ent th
e minimum/
maximum a
c
t
i
ve power of
unit
i
;
,
ih
PU
and
,
ih
PL
re
pre
s
e
n
t th
e a
c
tive po
wer
of unit
i
wh
en
wind
po
wer i
s
gre
a
ter
or
small
e
r th
an fo
re
ca
st
value, respectively;
,
Dh
P
in
dic
a
t
e
s t
h
e
sy
st
e
m
loa
d
at
t
i
me
h
;
,,
s
tor
s
h
P
indi
cate
s active
po
we
r of
energy sto
r
a
ge unit
s
at time
h
;
i
UR
and
i
D
R
indicate th
e ra
mping u
p
/do
w
n limit of u
n
it
i
;
,
up
h
R
and
,
down
h
R
indicat
e
the reg
u
lati
on up/do
wn
capa
city at time
h
;
,,
,
s
to
r
u
p
s
h
R
and
,,
,
s
t
o
r
dow
n
s
h
R
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 5, May 2014: 3357 – 33
65
3362
indicate the
regulatio
n up/
down re
serve
ca
pacity of
storage
unit
s
at time
h
;
,,
Line
l
h
P
indi
cates
the active po
wer flo
w
of line
l
at time
h
;
,,
m
i
n
L
ine
l
P
and
,,
m
a
x
Li
ne
l
P
indica
te the maximum and
minimum a
c
tive powe
r
flo
w
of line
l
;
g
iv
en
q
is a given limit according to
the reliabity of powe
r
grid.
(3) BESS Formation
The followi
ng modes for BESS are considered:
a. BESS c
an be used as
either generator or load.
b. Maximum cha
r
ge a
nd d
i
scharge po
wer are
not co
nstant
s and
chang
e in accorda
n
ce
with the stat
e of cha
r
ge
(SOC) whi
c
h is t
he pe
rcenta
ge of stored el
ectri
c
energy. Their
relation
shi
p
can be re
presented by
a pi
ece
w
i
s
e linea
r function.
c. In o
r
d
e
r to
extend th
e
service
life, th
e de
gre
e
of
store
d
ene
rg
y
sho
u
ld be kept
i
n
a
certai
n ra
nge
r.
d. Power
ram
p
ing spee
d is much fa
ster
than
therm
a
l
so that the re
spo
n
ce time can b
e
negle
c
ted.
e. After the schedule, BESS should have more
than a given amount of electri
c
energy.
Accordi
ng to the specifi
c
characteri
stics, the BESS model
is establi
s
hed as follows:
Cha
r
ge a
nd d
i
scharge po
wer limit con
s
traints:
,m
a
x
,m
a
x
,,
,,
,,
(
1
...
;
1
...
)
ch
a
d
i
s
ch
a
stor
s
h
st
or
s
h
st
or
s
h
P
PP
s
S
h
H
(16)
Stored en
ery limit const
r
ain
t
s:
,,
m
i
n
,
,
m
a
x
,,
,,
m
a
x
,
,
m
a
x
(
1
...
;
1
.
.
.
)
st
or
s
s
t
o
r
s
s
tor
s
h
st
or
s
s
tor
s
CC
C
s
S
h
H
(17)
Energy con
s
traints:
,,
,
,
,,
1
,
,
,,
,,
1(
0
)
(
1
...
;
1
.
.
.
)
1(
0
)
dis
c
ha
ss
t
o
r
s
h
H
s
t
o
r
s
h
sto
r
s
h
sto
r
s
h
cha
ss
t
o
r
s
h
H
s
t
o
r
s
h
PP
CC
s
S
h
H
PP
(18)
Reg
u
lation re
serve
cap
a
cit
y
const
r
aints:
,,
,
,
m
i
n
,
,
m
a
x
,m
a
x
,,
,
,
,
,
,
,,
m
a
x
,
,
m
a
x
,,
,m
a
x
,,
,
,
,
,
,
mi
n
{
,
}
1
mi
n
{
,
}
1
(
1
..
.
;
1
.
..
)
st
or
s
h
st
or
s
s
tor
s
di
scha
st
or
up
s
h
st
or
s
h
st
or
s
h
H
s
to
r
s
s
t
o
r
s
s
to
r
s
h
ch
a
st
or
down
s
h
s
t
o
r
s
h
s
tor
s
h
H
CC
RP
P
CC
RP
P
sS
hH
(19)
Amount of stored e
n
e
r
gy in the end of sche
dule:
,,
,,
m
a
x
,
,
m
i
n
(
1
...
;
)
cap
st
or
s
h
st
or
s
s
to
r
s
CC
s
S
h
H
(20)
Whe
r
e
,m
a
x
,,
di
scha
st
or
s
h
P
and
,m
a
x
,,
cha
s
to
r
s
h
P
indicates maxi
mum disch
a
rge and cha
r
g
e
powe
r
of st
orag
e unit
s
at time
h
;
,,
m
i
n
st
or
s
and
,,
m
a
x
s
t
or
s
indicate the
maximum/mi
nimum p
r
op
o
r
tions
of elect
r
ic e
nergy;
,,
s
tor
s
h
C
indicate
s the
store
d
ene
rg
y of unit
s
at time
h
;
,,
m
a
x
st
o
r
s
C
indicate
s the ene
rgy
cap
a
city
of storage
uni
t
s
;
di
scha
s
and
cha
s
indi
cate the
ch
arg
e
an
d di
scha
rge effici
en
cie
s
of
unit
s
;
1
H
is
one ho
ur;
,,
m
i
n
cap
st
o
r
s
ind
i
cate
s the req
u
ired mi
nimu
m prop
ortion
of energy in the end of sch
edule.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Unit Com
m
i
tm
ent with Battery Ene
r
g
y
Storage
Co
n
s
i
derin
g Win
d
Fore
ca
st Erro
r (Cai Zhi)
3363
4. Results a
nd Analy
s
is
In order to testify the feas
ibility of proposed method, a
modified 10-unit IEEE 39-bus
system i
s
em
ployed a
s
sh
own i
n
Figu
re
3. A wi
nd g
e
n
erato
r
i
s
lo
cated at Bus 5
who
s
e
ca
pa
city
is
s
e
t 400MW. An BESS
is
loc
a
ted at
Bus
6 whose c
a
pac
ity is
set 100MW.
As sh
own in Figure 4, Ca
se
1 represe
n
ts the num
b
er of
ope
rati
ng therm
a
l u
n
its with
BESS and Case 2 represents result without BESS.
During t
i
me 2~5,
system has a
high
prop
ortio
n
of
win
d
p
o
wer
owin
g to l
o
w load.
Time
11~14 i
s
on
-pea
k lo
ad
ho
urs.
Durin
g
ti
me
15~21, the
wind p
o
wer i
n
crea
se
s. In th
ese
pe
riod
s,
more
op
erati
on the
r
mal
u
n
its a
r
e
re
qui
re
d
due to the un
certai
nty of wind po
wer
an
d pea
k l
oad.
But with the cha
r
ge
and d
i
scharge po
wer
provided by BESS, the n
u
mber of operating ther
mal units
can be reduced
s
i
gnific
a
ntly. P
e
ak-
load shifting attribute
of e
nergy sto
r
ag
e
sy
stem
i
s
evidently re
prese
n
ted th
ro
ugh thi
s
re
su
lt. In
Figure 5, results of total t
herm
a
l regul
ation rese
rve
cap
a
city
with/without BE
SS are
sho
w
n. In
most time p
e
r
iod
s
, therm
a
l
regul
ation u
p
re
se
rve
ca
pacity de
crea
se
s with th
e
cap
a
city brou
ght
by BESS, but thermal regulati
on down reserve capaci
ty doesn’t change much.
The average
o
f
Therm
a
l reg
u
l
ation
u
p
re
serve ca
paciti
e
s
i
s
431.7
M
W whi
c
h ca
n
be re
du
ced
to
32
4.3M
W with
the hel
p of B
ESS. In Figure 6,
we can observe
the acceptable
boundaries of wi
nd powe
r error
are s
l
ightly broad
ened wit
h
the help of
BESS.
This term mos
t
ly
related to weight
M
. If
M
becomes larger, m
o
re thermal
unit
s
will be turned on to tolera
te more
wind
power fo
recast
deviation. F
o
r in
stan
ce, if
M
is triple
d in
t
h
is
ca
se,
the
acce
ptable
boun
dari
e
s o
f
wind
p
o
we
r
error will
rise up 7.2%.
G
G
G
G
G
G
G
G
G
G
30
39
1
2
25
37
29
17
26
9
3
38
16
5
4
18
27
28
36
24
35
22
21
20
34
23
19
33
10
11
13
14
15
8
31
12
6
32
7
W
S
Figure 3. Structure of Po
we
r System wi
th
10 Thermal Units a
nd 1
Wind
Unit
Figure 4. Operating Therm
a
l Units’
Nu
m
ber of UC Results
with/wit
hout BESS
Number
of
operating
thermal
units
Hour
Case
1
Case
2
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
0
46
TELKOM
NI
KA
Vol. 12, No. 5, May 2014: 3357 – 33
65
3364
Figure 5. The
r
mal Units Re
gulation
Re
se
rve
Capacity of UC Result
s wit
h
/without BESS
Figure 6. Acceptable Bou
n
d
arie
s of Win
d
Power Error
with/without BESS
Table 1 is th
e results of
UC a
nd wi
nd
powe
r
st
o
c
h
a
st
ic
simulat
i
on ch
ec
k.
Ob
ey
ing t
h
e
norm
a
l dist
rib
u
tion, wind
p
o
we
r sto
c
h
a
stic simul
a
tion
che
c
k data
s
are g
e
ne
rate
d from MATL
AB
whi
c
h have
mean
and st
anda
rd devia
tion
/
a
. As sho
w
n in Ta
ble
1, the prop
osed model
pass
e
s
all
s
i
mulation c
hec
ks
. Calc
ulation
time
has
an inc
r
eas
e
with BESS, while t
he
s
ystem's
total cost becomes $ 869,6
61, down 2.35%, comparing to the case without BESS.
Table 1. Re
sults of UC a
n
d
Wind Po
we
r Stocha
stic
Simulation Check
Stochastic Simulation Check
Total cost($)
1
2 3 4
5
Without
BESS Pass
Pass
Pass
Pass
Pass
890604
With
BESS
Pass
Pass
Pass
Pass
Pass
869661
A 100
-unit
system
whi
c
h
consi
s
ts of ten
abov
e
10
-un
i
t system
s i
s
employed
ex
cludi
ng
line flow co
n
s
traint
s. The
numbe
rs of operatin
g t
hermal units in
every time slice mo
re or l
e
ss
decrease
wit
h
the help
of BESS.
The costs of unit commitmen
t are $7172815 and $7213122
with/without
BESS, res
pec
tively. In this
c
a
s
e
,
we obs
e
rve the ec
onomic impac
t
of
BESS on
large
r
scal
e p
o
we
r gri
d
.
5. Conclusio
n
In response t
o
the insecuri
ty brought by
wi
nd energy uncertainty,
large
capacity BESS
is p
r
op
osed t
o
solve this p
r
oble
m
. An i
m
prove
d
wi
n
d
fore
ca
st e
r
ror m
o
del i
s
propo
sed
to sui
t
the
actual datas.
Based
on this, a unit comm
itment
model
with wind powe
r
and large capacity BESS
is analyzed a
nd esta
blish
e
d
. Cases st
u
d
y with 10 un
its and 100 u
n
its are em
pl
oyed to validate
the model. The effec
t
of the BESS on
power
s
y
s
t
em c
a
n be summariz
e
d as follows
: (1)
help
pea
k loa
d
shi
fting; (2)
de
crease
the
nu
mber of op
erating the
r
ma
l
units;
(3) re
duce the
syst
em
operating co
sts; (4
) help
powe
r
sy
ste
m
accomm
o
date win
d
po
we
r. In furth
e
r re
se
arch,
the
energy loss in cha
r
gin
g
an
d discha
rgin
g
processe
s should b
e
anal
yzed.
Referen
ces
[1]
Alec Bro
o
ks, Ed Lu, D
an
R
e
ich
e
r, Char
le
s Sp
irakis, B
ill
W
e
ihl. D
e
ma
nd Dis
patch-
U
sing R
e
a
l
-T
ime
Contro
l of Dem
and to He
lp Ba
lanc
e G
ener
ati
on an
d Lo
ad.
IEEE pow
er & ener
gy mag
a
z
i
ne
. 201
0; (5):
21-2
9
.
[2]
KC Div
y
a, Jac
ob O
s
terga
a
rd
. Batter
y
e
ner
g
y
stor
ag
e tec
hno
log
y
for
po
w
e
r s
y
stems-A
n
overvi
e
w
.
Electric Power System
s Res
e
arch.
200
9; 79(
4): 511-5
20.
[3]
Daneshi
H, Srivastava AK.
Im
pact
of battery energy st
orag
e on pow
er system
wit
h
high wind
pen
etratio
n
.
T
r
ansmissi
on a
n
d
Distrib
u
tion
Confer
ence
an
d Expositi
on (T
&D). O
r
lando.
201
2: 1-8.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Unit Com
m
i
tm
ent with Battery Ene
r
g
y
Storage
Co
n
s
i
derin
g Win
d
Fore
ca
st Erro
r (Cai Zhi)
3365
[4]
Dan
e
shi H, Da
nesh
i
A,
T
abari NM, Jahromi
AN.
Security-constrai
ned u
n
i
t commit
me
nt in a syste
m
w
i
th w
i
nd ge
ne
ration
an
d co
mpresse
d air
en
ergy stora
ge.
Proc. Si
xth Int. Conf. on
Euro
pea
n En
erg
y
Market, EEM.
Leuv
en. 20
09: 1-6.
[5]
Garcia-Gonz
al
ez J, d
e
l
a
M
uel
a RMR,
Sa
ntos
LM, Go
n
z
alez
AM. Sto
c
hasti
c Joint
Optimi
za
ti
on
of
W
i
nd Gener
ati
on an
d Pumpe
d
-Storag
e
Unit
s in an Electri
c
it
y
Market.
IEEE Transactions on Power
System
s
. 20
08
; 23(2): 460-4
6
8
.
[6]
Rodic
a
Lois
e
l.
Po
w
e
r
s
y
ste
m
fle
x
ib
ilit
y
w
i
th el
ectricit
y storag
e tech
no
l
ogi
es: A tech
n
i
cal-ec
on
omic
assessme
n
t of a larg
e-scal
e
storage fac
ilit
y.
Electrical P
o
w
e
r and E
ner
gy Systems
. 20
12
;42(1): 54
2-
552.
[7]
Jian
hui W
a
ng,
Mohamm
ad
Shah
ide
h
p
our,
Z
u
y
i
Li. Sec
u
rit
y
-
C
onstr
ain
ed U
n
it Com
m
itment W
i
th
Volatile Wind Po
w
e
r Generation.
IEEE Transactions on Power System
s.
20
08; 23(3):
13
19
-1
32
7
.
[8
]
Ai
d
a
n
T
u
o
h
y
,
Pe
te
r Me
i
b
o
m
, El
e
a
n
o
r
D
enny
, Ma
rk O’
Ma
l
l
e
y
.
Un
i
t
C
o
mmi
tme
n
t
fo
r Sy
ste
m
s Wi
th
Sign
ificant W
i
n
d
Penetrati
on.
IEEE Transactions on Power
System
s.
20
09
; 24(2): 592-6
0
1
.
Evaluation Warning : The document was created with Spire.PDF for Python.