Int
ern
a
ti
onal
Journ
al of Ele
ctrical
an
d
Co
mput
er
En
gin
eeri
ng
(IJ
E
C
E)
Vo
l.
10
,
No.
4
,
A
ugus
t
2020
,
pp.
4136
~
41
44
IS
S
N: 20
88
-
8708
,
DOI: 10
.11
591/
ijece
.
v10
i
4
.
pp
4136
-
41
44
4136
Journ
al h
om
e
page
:
http:
//
ij
ece.i
aesc
or
e.c
om/i
nd
ex
.ph
p/IJ
ECE
A steep
est des
cent al
gorith
m for th
e o
ptim
al cont
rol of
a cascad
ed hydro
power sy
stem
Olaleka
n O
gunbiyi
1
, C
ornel
ius T. Th
om
as
2
, O
lud
are
Y.
Og
u
ndep
o
3
, Is
aa
c
O. A
. Om
ei
z
a
4
,
Jimo
h Aka
nni
5
, B. J. Ol
uf
e
agb
a
6
1
El
e
ct
ri
ca
l
and
C
om
pute
r
Engi
n
e
eri
ng
Dep
art
m
en
t,
K
war
a
Sta
te Unive
rsit
y
,
Nig
eria
2
El
e
ct
ri
ca
l
and
I
nform
at
ion
Eng
i
nee
ring
Depa
r
tment, Achieve
rs U
nive
rsit
y
,
Nig
eria
3
El
e
ct
ri
ca
l
and
E
le
c
troni
c
Eng
ineeri
ng,
Feder
al Unive
rsit
y
of
Pet
r
o
le
um
Resourc
e
s,
Niger
i
a
4
El
e
ct
ri
ca
l
and
I
nform
at
ion
Eng
i
nee
ring
,
La
ndm
a
rk
Un
ive
rsit
y
,
Ni
ger
ia
5,6
El
e
ct
ri
ca
l
and
El
e
ct
roni
cs
Eng
i
nee
ring
Depa
r
tment, Unive
rsi
t
y
o
f
Ilori
n
,
Nig
eria
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
Ja
n 1
3
, 2
019
Re
vised
Feb 1
2
,
2020
Accepte
d
Fe
b 25
, 202
0
Optimal
power
gene
ra
ti
on
al
on
g
the
ca
sca
d
ed
Kainj
i
-
Jebba
h
y
droe
lectr
i
c
power
s
y
stem
h
ad
bee
n
v
er
y
di
ffic
ul
t
to
a
chiev
e.
Th
e
rese
r
voir
s
oper
at
ing
hea
ds
are
bei
ng
aff
ec
t
ed
b
y
po
ss
ibl
e
var
iation
in
impoundm
ent
s
upstrea
m
,
stocha
sti
c
fa
ct
or
s tha
t
ar
e
we
at
he
r
-
rel
a
te
d
,
av
ai
l
ab
il
ity
of
the t
urbo
-
al
t
ern
at
ors
and
power
generat
ed
a
t
an
y
ti
m
e.
Propos
ed
in
thi
s
pape
r,
is
an
al
gorit
hm
for
solving
the
opti
m
al
relea
s
e
of
wate
r
on
the
ca
sc
ade
d
h
y
dropower
s
y
stem
base
d
o
n
stee
pest
d
esc
e
nt
m
et
hod.
Th
e
unique
ness
of
t
his
work
is
the
conv
ersion
of
the
infi
ni
te
d
imensio
nal
con
t
rol
proble
m
to
a
finite
on
e,
the
in
troduc
t
ion
of
cleve
r
te
chn
i
ques
for
choosi
ng
the
st
ee
pest
desc
ent
st
ep
size
in
ea
ch
i
te
r
at
ion
and
the
no
nli
ne
ar
pen
alt
y
embedde
d
in
th
e
proc
edur
e.
The
con
trol
a
lgo
rit
hm
was
implemente
d
in
an
Ex
ce
l
VBA
®
envi
r
onm
en
t
to
solve
the
form
ula
t
ed
La
gr
ange
proble
m
withi
n
an
accura
c
y
of
0.
03%.
It
i
s
rec
om
m
ende
d
for
use
in
s
y
ste
m
studie
s
and
cont
rol
design
fo
r
the
opti
m
a
l
power
gen
era
t
io
n
in the
ca
sc
aded h
y
dropower
s
ystem.
Ke
yw
or
d
s
:
Hydro
powe
r
Inflo
ws
Op
e
rati
ng
head
Perfo
rm
a
nce
ind
ex
Stee
pest
desce
nt
Copyright
©
202
0
Instit
ut
e
o
f Ad
vanc
ed
Engi
n
ee
r
ing
and
S
cienc
e
.
Al
l
rights re
serv
ed
.
Corres
pond
in
g
Aut
h
or
:
Olal
ekan O
gunbiy
i,
Dep
a
rtm
ent o
f El
ect
rical
an
d
Com
pu
te
r
E
ng
i
neer
i
ng,
Kw
a
ra S
ta
te
U
niv
e
rsity
,
P.M.B
1530, Il
or
i
n,
Kwara
Stat
e, N
ige
ria.
Em
a
il
: biy
ikan@
gm
ai
l.co
m
1.
INTROD
U
CTION
Hydro
powe
r
ge
ner
at
io
n
in
N
igeria
is
cur
re
ntly
pr
ovide
d
at
three
m
ajo
r
locat
ion
s
[
1].
Tw
o
of
thes
e
sta
ti
on
s
are
l
oc
at
ed
on
the
Ri
ver
Nige
r,
ope
rated
in
casca
de
.
They
a
re
the
Kain
j
i
hy
droe
le
ct
ric
power
s
ta
ti
on
(KHEP
S)
a
nd
t
he
Jeb
ba hyd
roel
ect
ric p
ower
sta
ti
on
(J
HEPS
).
T
hey are th
e
Kain
j
i
hydr
oelect
ric p
ower st
at
ion
(
K
HEPS
)
a
nd
t
he
Jeb
ba
hy
dro
el
ect
ric
powe
r
sta
ti
on
(J
HEP
S
).
The
K
HEP
S
w
hich
is
locat
ed
at
0
9
0
5
1
′
4
5
′′
,
0
4
0
3
6
′
4
8
′′
with
a
n
i
ns
ta
ll
capaci
ty
of
760
M
W
form
eig
ht
unit
s
of
tur
bo
al
te
r
nato
rs.
The
JHE
PS
is
locat
e
d
103
km
downs
tream
of
the
K
HEP
S
on
0
9
0
0
8
′
0
8
′′
,
0
4
0
4
7
′
1
6
′′
.
It
w
as
com
m
issi
on
ed
on
A
pri
l
13,
1985,
with
a
rated
ca
pacit
y of
57
8.4
M
W
from
six
(6) fi
xe
d blade
[2,
3]
.
The
Je
bba
Re
s
ervoir
de
pends
on
discha
rg
e
and
s
pill
from
the
KHEPS
,
t
his
ar
ra
ng
em
ent
im
po
ses
the n
ee
d for
be
tt
er w
at
er m
anag
em
ent if the un
it
s at Je
bba a
re to
operate e
f
fici
ently
all
the year.
T
he op
e
r
at
or
s
of
the
J
HE
PS
face
serio
us
chall
enges
tha
t
inv
ol
ve
bala
ncin
g
co
nf
li
ct
ing
nee
ds
in
volving
t
he
oper
at
ion
al
safety
of
the
s
ta
ti
on
s
a
nd
t
he
dem
and
re
quirem
ents
fr
om
an
e
ner
gy
-
sta
r
ved
el
ect
rici
ty
gr
i
d
[2,
4].
Pra
ct
ic
al
ob
s
er
vation
of
operati
ons
re
ve
al
s
the
s
erio
us
c
halle
nges
confro
nting
th
e
JHE
PS
op
e
r
at
or
s
as
they
t
ry
to
perform
the
f
un
ct
io
ns
of
a
re
gu
la
to
r
t
ha
t
was
om
itted
in
the
fi
xed
van
e
de
sig
ns
of
th
e
J
HE
PS
tur
bo
al
te
rn
at
ors
[
5,
6].
T
hese
op
e
r
at
ion
al
pr
ob
le
m
s
are
pr
esent
in
each
HE
PS
in
one
for
m
or
the
oth
e
r,
a
nd
it
is
on
ly
because
of the
robust
nat
ur
e
of the
tu
rbo
al
te
r
nato
rs
th
at
m
ajo
r
cata
strop
hes have
no
t
ye
t occurre
d
[
7].
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
A steepest
desc
ent a
l
gorit
hm for
t
he op
ti
m
al
con
tr
ol
…
(
Ol
al
ekan
O
gunbiy
i
)
4137
Nonetheless
,
there
is
evide
nc
e
of
su
c
h
pro
bl
e
m
s
lurk
i
ng
i
n
the
possibil
it
ie
s
wh
en
on
e
ob
s
er
ves
that
on
e
t
urbo
al
te
r
nato
r
cat
ast
rop
hical
ly
fail
ed
at
JHEP
S
a
nd
a
no
t
her
at
K
HE
PS.
Bot
h
cases
sh
are
the f
eat
ure
that
no
local
regula
tor
was
i
nclu
de
d
in
the
i
niti
al
desig
n.
F
or
e
xam
ple,
the
use
of
F
ran
ci
s
T
urbines
at
a
ve
ry
lo
w
head
sc
hem
e
as
t
he
KHE
PS
posed
es
pe
ci
al
prob
le
m
s
that
m
eant
po
or
pe
rform
ance
an
d
ul
tim
a
te
l
y
cat
ast
ro
phic
fa
il
ur
e.
I
n
the
J
HEP
S
case
,
th
e
nar
r
ow
oper
at
ing
hea
d
de
m
and
ed
by
the
fixed
van
e
Kap
la
n
tur
bin
es
pose
d a ve
ry seri
ou
s
chall
enge to
th
e operat
or
s
[
8].
The
la
tt
er
pr
oble
m
,
ho
wev
e
r
,
le
nd
s
it
sel
f
to
the
ap
plica
tio
n
of
o
ptim
a
l
con
tr
ol
m
et
ho
ds.
I
nd
ee
d,
this
is
the
pro
blem
add
resse
d
in
t
his
rese
arch.
It
is
pos
ed
as
a
n
op
ti
m
al
con
trol
pr
ob
le
m
to
dete
rm
ine
the
infl
ow
int
o
the
JHE
PS
reserv
oir
s
o
t
hat
the
op
e
rat
ing
hea
d
fall
s
within
a
sp
e
ci
fied
ra
nge
[
9,
10]
.
This
w
ork
c
onsidere
d
the
determ
inati
on
of
op
ti
m
al
con
tro
l
la
w
for
the
r
el
ease
of
water
from
KH
EP
S
suc
h
that
the
reservo
i
r
head
at
JHEPS
rem
ai
ns
relat
ively
const
ant.
T
he
optim
al
con
t
ro
l
al
gorithm
is
to
be
inco
rpor
at
e
d
i
nt
o
a
real
-
ti
m
e e
m
bed
ded
c
ontr
oller.
Unfortu
n
at
el
y,
the
desig
n
of
su
c
h
a
syst
e
m
that
ensures
th
e
op
ti
m
al
us
e
of
hydro
powe
r
resour
ces
t
o
m
axi
m
iz
e
po
w
er
ge
ne
rati
on
within
a
ca
sca
ded
syst
em
is
chall
eng
i
ng.
T
he
pro
blem
m
us
t
be
pr
op
e
rly
pose
d
in
a
sta
nd
a
r
d
f
orm
bef
ore
it
can
be
s
olv
e
d
[
11,
12
]
.
So
l
utions
of
optim
al
con
tr
ol
pro
blem
s
are
oft
en
a
nal
yt
ic
al
l
y
intract
able
a
nd co
m
pu
ta
ti
on
al
ly
co
m
plex
[9,
13
]
.
Du
e
to
t
he
c
om
plexit
y
of
sy
stem
s
and
a
ppli
cat
ion
s,
a
naly
ti
cal
m
e
tho
ds
are
ra
rely
us
e
d
to
c
om
pu
te
the
so
luti
ons
to
optim
al
con
trol
pro
blem
s.
Nu
m
erical
so
lut
ion
s
are
m
os
tl
y
us
ed
in
de
te
rm
ining
optim
al
con
t
ro
l
[14,
15]
.
I
n
m
os
t
cases,
the
num
erical
m
et
ho
ds
are
highly
so
phist
ic
at
ed,
an
d
t
hey
do
ta
sk
com
pu
te
rs
[
16,
17
]
.
T
her
e
hav
e
bee
n
nu
m
ero
us
nu
m
erical
pr
oce
dure
s
dev
el
op
e
d
over
the
ye
ars
,
these
proce
dures
ca
n
be
cl
assifi
ed
into
tw
o
cat
e
gories,
t
he
di
r
ect
m
et
ho
ds
a
nd
i
ndirect
m
e
thods
[18].
T
he
direct
m
et
ho
d
of
co
m
pu
ti
ng
optim
al
con
tr
ol
invo
lves
the
discr
e
ti
zat
ion
of
the
sta
te
and
the
c
on
t
ro
l
in
s
uch
a
way
that
the
pro
blem
is
conve
rted
to
a
no
nlinear
optim
iz
at
ion
pro
blem
or
no
nlinear
p
r
ogra
m
m
ing
pro
blem
[1
7,
19,
20
]
.
T
he
in
direct
m
et
ho
d app
li
es
cal
c
ulus
of v
ariat
io
n
t
o
set
up n
ecess
ary
co
ndit
ion
s
that
m
us
t
be
sa
ti
sfied
by the
opti
m
a
l con
t
ro
l.
2.
STE
EPE
ST D
ESCENT
SO
LUTION
OF
AN OPTI
M
A
L CO
NTR
OL
PROB
LE
M
The
ste
epest
d
escent
al
gorith
m
is
gen
erall
y
us
ed
f
or
dete
r
m
ining
the
m
i
nim
u
m
of
a
diff
e
ren
ti
able
functi
on
an
d
he
nce,
em
plo
ye
d
as
a
direct
m
et
ho
d
of
so
l
ving
opti
m
al
c
o
nt
ro
l
prob
le
m
in
this
w
ork
[
21
-
23
]
.
Give
n
a
pe
rfor
m
ance
ind
e
x
(
ℎ
,
)
that
is
differentia
ble,
t
he
ste
epest
des
cent
directi
on
is
the
path
opposit
e
∇
(
)
.
The
search
sta
rts
at
a
diff
e
re
ntiable
po
i
nt
=
0
(
)
an
d
de
creases
afte
r
e
ach
it
erati
on
unti
l
it
reaches
the m
inim
u
m
p
oin
t w
it
h
∗
(
)
, such
that:
∗
=
∂
∗
∗
(
)
≈
0
(1)
Stee
pest
desce
nt
al
gorithm
is
ver
y
fast
i
n
m
ov
i
ng
a
s
ol
utio
n
f
ro
m
any
local
po
int
within
the
feasibl
e
reg
i
on
to
t
he
vi
ci
nity
of
the
point
of
co
nver
gen
ce
[
2
1
,
2
4
,
2
5
]
.
I
f
a
s
uitab
le
m
et
ho
d
of
s
olu
ti
on
is
unkn
own
,
ste
epest
desce
nt
is
guara
nteed
to
it
erate
to
wards
the
m
inim
u
m
po
int
but
char
act
erise
d
by
slow
c
onve
rg
e
nce
.
Nu
m
erous
rese
arch
es
on
this
m
et
ho
d
ha
ve
be
en
on
the
determ
inati
on
of
appr
opriat
e
ste
p
siz
e
and
m
e
ans
of
sp
ee
ding it
up [2
6].
2.1
.
St
ateme
nt
of the pr
ob
le
m
T
he
optim
al
con
tr
ol
so
l
ve
d
in
this
w
ork
is
th
e
determ
inati
on
of
the
co
ntr
ol
sign
al
(
)
(
ge
nerat
ed
by
the
c
on
t
ro
l
Sys
tem
)
to
be
rele
ased
f
ro
m
K
H
EPS
(actuat
or)
that
will
f
orce
the
ope
rati
ng
head
ℎ
(
)
of
JH
EP
S
(Contr
olled Pla
nt)
t
o
f
ollo
w
a
pr
e
def
i
ned tra
je
ct
or
y
within a
g
ive
n
ti
m
e
(
0
→
)
.
T
he
syst
em
is
the
dy
nam
ic
a
l
m
od
el
f
or
t
he
J
HEP
S
op
erati
ng
hea
d
de
scribe
d
by
th
e
nonlinea
r
di
ff
e
re
ntial
h
ea
d 3.1
[8,
27
]
:
ℎ
(
)
=
−
2
1
√
2
ℎ
1
2
⁄
(
)
+
1
1
(
(
)
−
(
)
−
(
)
)
(2)
=
+
+
(3)
(
)
=
(
)
−
(
)
−
(
)
(4)
Hen
ce
, th
e
syst
e
m
m
od
el
can
be writt
en
i
n d st
and
a
rd form
as;
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
10
, No
.
4
,
A
ugus
t
2020
:
4136
-
4144
4138
ℎ
̇
(
)
=
(
ℎ
(
)
,
(
)
,
)
;
0
≤
≤
(5)
wh
e
re:
: t
i
m
e
ℎ
: Op
e
rati
ng
he
ad
a
nd the
sta
te
v
a
riable
:
Con
tr
ol si
gn
a
l
: Nu
m
ber
of op
erati
ng unit
s (
i
ntege
r nu
m
ber
1
to
6)
1
: Eff
ect
ive
S
urface are
a
of
t
he
r
ese
rvoir
2
: Eff
ect
ive
area
of the
scroll
c
asi
ng
: Accel
erati
on
du
e
to g
rav
it
y
: Inflo
w
int
o
J
HEP
S
: Eva
porati
on l
os
s
on J
HEPS
: Spill
way d
isc
harge
from
JH
EPS
: Total
d
isc
harge
from
K
HEPS ta
il
race
: Spill
way d
isc
harge
from
K
H
EPS
: Inflo
w fr
om
c
at
chm
ent area b
et
wee
n KHE
PS
a
nd JHEPS
: Nonli
nea
r
f
un
ct
ion
2.2.
P
erfo
r
ma
nce i
ndex
‘
J’
Perfo
rm
a
nce in
dices can be se
le
ct
ed
to r
eflec
t t
he
aspect of
t
he
syst
em
’s
beh
avi
our
that i
s co
nsi
der
e
d
as
vital
.
As
a
r
esult,
a
pe
rform
ance
ind
e
x
wh
ic
h
accom
m
od
at
es
and
a
ppr
opriat
el
y
pen
al
iz
es
de
viati
on
from
a
sp
eci
fie
d
hea
d
a
nd
e
nsure
s
that
the
c
on
tr
ol
do
es
not
re
quire
values
out
side
the
ca
pa
bili
ty
of
K
HEP
S
was
sel
ect
ed.
A
quadr
at
ic
perfor
m
ance
in
dex
was
sel
ect
e
d
c
on
sist
in
g
of
th
e
integ
ral
of
t
he
s
quare
er
ror
f
r
om
the
desire
d
operati
onal
hea
d
an
d
the
s
quare
dev
ia
ti
on
fr
om
the
m
a
xim
u
m
dischar
ge
possible.
Give
n
a
pe
rfor
m
ance
ind
e
x
(
ℎ
,
,
)
that i
s d
i
ff
e
ren
ti
able:
=
min
=
∫
{
ℎ
(
ℎ
(
)
−
ℎ
(
)
)
2
}
0
(6)
Subj
ect
t
o
t
he
s
yst
e
m
co
ns
trai
nts:
ℎ
̇
(
)
=
(
ℎ
(
)
,
(
)
,
)
;
0
≤
≤
ℎ
(
0
)
=
ℎ
0
(7)
ℎ
(
)
=
ℎ
(
)
(8)
Nonlinea
r pena
lt
ie
s o
n
:
[
(
)
,
(
)
]
,
wh
e
re
ℎ
(
)
re
pr
e
se
nts the
d
esi
red
final
value f
or
the stat
e an
d
is a
po
sit
ive
we
igh
in
g sca
la
r
c
on
sta
nt.
2.3.
S
olut
i
on o
f op
timal c
ontrol
using steep
est desc
ent
a
l
go
ri
th
m
A
s
olu
tion
of
t
he
op
ti
m
al
con
trol
usi
ng
ste
e
pest
de
scent
a
ppr
oac
h
wa
s
ear
li
er
rev
ie
wed.
The
ste
e
pest
desce
nt
di
recti
on
is
the
pa
th
opposit
e
∇
(
)
.
T
he
search
sta
rts
at
a
diff
e
ren
ti
a
ble
point
(
)
an
d
de
crease
s
after eac
h
it
era
ti
on
un
ti
l i
t rea
ches t
he
m
ini
m
um
p
oin
t
∗
(
)
. Whe
re:
∗
=
∂
∗
∗
(
)
≈
0
(9)
i
f
(
)
is a u
nit v
e
ct
or
al
ong
t
he
i
ncr
easi
ng gra
dient,
(
(
)
)
=
[
(
)
]
‖
(
)
‖
(10)
t
hen
,
m
ov
ing
i
n
the
d
i
recti
on
−
(
)
,
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
A steepest
desc
ent a
l
gorit
hm for
t
he op
ti
m
al
con
tr
ol
…
(
Ol
al
ekan
O
gunbiy
i
)
4139
+
1
(
)
=
(
)
−
(
)
(11)
is
the
op
ti
m
u
m
ste
epest
desce
nt
ste
p
siz
e.
A
n
opti
m
u
m
val
ue
of
m
us
t
be
determ
ined
be
ca
us
e
a
l
arg
er
value
of
resu
lt
s
in
t
he
local
m
ini
m
u
m
of
bein
g
ov
e
rs
ho
t.
A
s
m
al
le
r
value
f
or
will
req
ui
r
e
m
uch
tim
e
an
d
it
erati
ons
f
or
the
sea
rc
h.
He
nce
,
m
us
t
be
determ
i
ned
s
uc
h
a
s
t
o
m
ov
e
to
wa
rd
s
the
m
ini
m
u
m
at
the
sm
al
le
s
t
com
pu
ta
ti
onal
tim
e,
this
is
an
oth
e
r
op
ti
m
iz
at
ion
.
As
sh
ow
n
in
(
12)
giv
e
s
an
a
ppr
ox
im
ate
en
d of t
he
se
arch i
n
t
he pres
ent d
i
recti
on
:
+
1
(
)
=
(
)
−
(
)
(
)
‖
(
)
‖
(12)
An in
novative
appr
oach to
fin
ding the
opti
m
um
is
by u
si
ng the
qu
a
dr
at
ic
ap
pr
oach
.
(
)
=
+
+
2
(13)
(
)
=
2
+
=
0
(14)
(
)
=
−
2
(15)
Hen
ce
,
[
+
1
(
)
]
=
[
(
)
]
−
(
)
[
1
(
)
,
2
(
)
,
3
(
)
,
4
(
)
]
√
(
1
(
)
)
2
+
(
2
(
)
)
2
+
(
3
(
)
)
2
+
(
4
(
)
)
2
(16)
The
com
pu
ta
ti
on
was
car
rie
d
out
num
eric
al
ly
in
an
EX
CEL
VBA
®
pro
gr
am
m
ing
env
i
ronm
ent.
The
A
dam
s
-
m
ou
lt
on
num
erical
integrato
r
with
ste
epes
t
descen
t
te
chn
i
qu
e
was
e
m
plo
ye
d
in
so
lvin
g
the syst
em
m
o
del and t
he
c
om
pu
ta
ti
on
of c
on
t
ro
ls.
2.4.
A
lg
orithm
f
or
t
he
nu
meri
c
al
so
luti
on
of th
e o
p
timal c
ont
rol usin
g
th
e
s
teepes
t
desce
n
t alg
orithm
Set u
p: D
ecl
a
r
e a contr
ol
vector
(
)
(
)
by u
si
ng a
f
in
it
e p
a
rtit
ion
of the ti
m
e inte
rv
al
[
0
,
]
.
[
0
,
]
=
0
<
1
<
2
<
3
<
(
)
(
)
=
[
1
(
)
(
)
,
2
(
)
(
)
,
3
(
)
(
)
,
4
(
)
(
)
]
Step
1:
Let
=
0
Set t
he
init
i
al
c
onditi
on
ℎ
(
=
0
)
(
)
=
ℎ
0
Gu
es
s
values
for
(
=
0
)
(
)
from
0
→
.
Step
2:
Nu
m
erical
ly
in
te
gr
at
e
(
ℎ
(
)
,
(
)
,
)
fr
om
0
→
to
obta
in
ℎ
(
)
(
)
.
The
nu
m
erical
integrati
on
is
c
a
r
r
i
e
d
o
u
t
u
s
i
n
g
a
n
A
d
a
m
s
–
M
o
u
l
t
o
n
t
e
c
h
n
i
q
u
e
w
i
t
h
A
d
a
m
s
–
B
a
s
h
f
o
r
t
h
a
s
p
r
e
d
i
c
t
o
r
a
n
d
R
u
n
g
e
-
K
u
t
t
a
for
sta
rtin
g.
Step
3:
Com
pu
te
the
pe
rfor
m
ance in
de
x
(
)
(
)
=
(
1
(
)
(
)
,
2
(
)
(
)
,
3
(
)
(
)
,
4
(
)
(
)
)
The
T
ra
pezo
i
da
l ru
le
was
em
plo
ye
d i
n
t
his c
om
pu
ta
ti
on
S
te
p 4:
Sele
ct
a p
e
rturbati
on
value
∆
and co
m
pu
te
(
)
;
=
1
,
2
,
3
,
4
1
(
)
=
(
1
(
)
(
)
+
∆
,
2
(
)
(
)
,
3
(
)
(
)
,
4
(
)
(
)
)
2
(
)
=
(
1
(
)
(
)
,
2
(
)
(
)
+
∆
,
3
(
)
(
)
,
4
(
)
(
)
)
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
10
, No
.
4
,
A
ugus
t
2020
:
4136
-
4144
4140
3
(
)
=
(
1
(
)
(
)
,
2
(
)
(
)
,
3
(
)
(
)
+
∆
,
4
(
)
(
)
)
4
(
)
=
(
1
(
)
(
)
,
2
(
)
(
)
,
3
(
)
(
)
,
4
(
)
(
)
+
∆
)
Step
5:
Com
pute
the app
roxi
m
at
e g
rad
ie
nt
vecto
r
.
=
≈
−
∆
;
=
1
,
2
,
3
,
4
Step
6:
Com
pute
the no
rm
o
f t
he
gra
dient
‖
‖
.
‖
(
)
(
)
(
)
)
‖
=
√
(
(
)
1
(
)
(
)
)
2
+
(
(
)
2
(
)
(
)
)
2
+
(
(
)
3
(
)
(
)
)
2
+
(
(
)
4
(
)
(
)
)
2
Step
7:
Com
pute
the unit
vector
(
(
)
(
)
)
z
(
(
)
(
)
)
=
[
(
)
(
)
(
)
)
]
‖
(
)
(
)
(
)
)
‖
Step
8:
Dete
rm
ine the
opti
m
u
m
step
(
)
Sele
ct
a set of t
hr
ee
Fib
onacci
num
ber
s
1
,
2
3
su
c
h
as
to o
btain
ψ
1
(
+
1
)
(
)
=
(
)
(
)
−
ψ
1
(
)
z
(
(
)
(
)
)
ψ
1
(
)
ψ
2
(
+
1
)
(
)
=
(
)
(
)
−
ψ
2
(
)
z
(
(
)
(
)
)
ψ
2
(
)
ψ
3
(
+
1
)
(
)
=
(
)
(
)
−
ψ
3
(
)
z
(
(
)
(
)
)
ψ
3
(
)
Step
9:
S
olv
e
f
or the c
onsta
nt
s
a
nd
[
]
=
[
(
ψ
1
−
ψ
2
)
(
ψ
1
2
−
ψ
2
2
)
(
ψ
2
−
ψ
3
)
(
ψ
2
2
−
ψ
3
2
)
]
−
1
[
(
(
ψ
1
)
−
(
ψ
2
)
)
(
(
ψ
2
)
−
(
ψ
3
)
)
]
Step
10
:
C
om
pu
te
(
)
ψ
(
)
(
)
=
−
2
Step
11
: C
om
pu
te
the c
ontr
ol
vecto
r
[
(
+
1
)
(
)
]
=
[
(
)
(
)
]
−
ψ
(
)
(
)
[
(
)
1
(
)
(
)
,
(
)
2
(
)
(
)
,
(
)
3
(
)
(
)
,
(
)
4
(
)
(
)
]
√
(
(
)
1
(
)
(
)
)
2
+
(
(
)
2
(
)
(
)
)
2
+
(
(
)
3
(
)
(
)
)
2
+
(
(
)
4
(
)
(
)
)
2
Step
12
:
C
hec
k i
f
‖
∇
(
(
+
1
)
‖
−
‖
∇
(
(
)
‖
≤
10
−
an
d
∗
∗
≈
0
,
w
her
e
n is a
posit
ive consta
nt.
If
t
his is tr
ue, t
hen
∗
(
)
=
(
+
1
)
(
)
an
d ou
t
put
ℎ
∗
(
)
el
se
le
t
=
+
1
,
(
)
(
)
=
(
+
1
)
(
)
an
d retu
r
n t
o
ste
p 2
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
A steepest
desc
ent a
l
gorit
hm for
t
he op
ti
m
al
con
tr
ol
…
(
Ol
al
ekan
O
gunbiy
i
)
4141
3.
RESU
LT
S
A
ND AN
ALYSIS
The
res
ults
obta
ined
f
r
om
the
com
pu
ta
ti
on
of
op
ti
m
al
con
trol
us
i
ng
the
ste
epest
desce
nt
al
gorithm
are
pr
es
ente
d
in
case
1
to
case
5.
A
nota
ti
on
is
us
ed
f
or
s
pecifyi
ng
oper
at
ing
co
nd
it
io
ns
under
a
case
bein
g
consi
der
e
d,
the
for
m
at
is as f
ol
lows
: (
num
ber
of m
achines,
s
ta
rting
hea
d
(m
),
num
ber
of
da
ys
, p
e
nalty
).
-
Ca
se 1
:
(
5
,
25
.
8
,
1
,
)
Applyi
ng
t
he
sta
te
d
op
e
rat
ing
c
onditi
ons
into
the
ste
epest
desce
nt
al
go
rit
hm
and
ass
um
in
g
a
co
nver
gen
ce
crit
erion
of
10
−
5
for
the
gra
dient
l
ed
t
o
c
onve
rg
e
nce
a
fter
15
it
e
rati
on
s
with
ℎ
(
)
tr
ajecto
ries
as
show
n
i
n
F
i
gure
1.
T
he
tr
ajecto
ries
ex
hi
bit
so
m
e
ov
er
s
hoot
becau
se
the
inter
val
of
6
hrs
us
e
d
to
s
pecify
the contr
ols
preven
t t
he
a
dju
s
t
m
ent o
f
th
e co
ntr
ol w
it
h
t
he p
recisi
on that
w
ou
l
d gu
a
ra
ntee the
desire
d
te
r
m
inal
value
near
ly
preci
sel
y.
It
is
evide
nt
that
ope
ra
tor
s
us
i
ng
pr
oto
c
ols
ba
sed
on
t
his
res
ult
will
be
able
t
o
m
or
e
pr
eci
sel
y
con
t
r
ol
and
m
anag
e
their
plants.
T
he
res
ults
dem
on
st
rate
the
use
of
the
optim
al
con
tr
ol
appr
oach
and
prov
i
de
de
penda
ble
m
et
ho
ds
f
or
operati
on
s
with
a
ne
gl
igible
de
viati
on
of
0
.
03%
of
the
optim
u
m
head
ℎ
∗
(
)
from
the set va
lue
ℎ
(
)
.
Figure
2
prese
nts
the
co
ntr
ol
that
is
requir
ed
to
produce
the
nee
ded
he
ad
tra
j
ect
ory
of
Fig
ur
e
1.
The
fiftee
nth
i
te
rati
on
s
produced
th
e
desir
ed
opti
m
a
l
con
tr
ol.
The
c
ontrol
la
w
sta
rts
with
a
high
in
flo
w
in
the
fir
st
6
ho
urs
of
the
day
(
1
)
a
nd
re
duces
to
the
m
ini
m
u
m
value
in
the
seco
nd
6
hours
of
the
day
(
2
).
The
c
on
tr
ol
increase
s
li
gh
tl
y
in
the
rem
ai
nin
g
12
ho
ur
s
t
o
m
ov
e
the
hea
d
to
t
he
opti
m
u
m
l
evel.
The
i
nput
pa
ra
m
et
ers
are:
ℎ
(
0
)
=
25
.
8
,
ℎ
(
)
=
26
.
1
,
1
(
0
)
=
2
(
0
)
=
3
(
0
)
=
4
(
0
)
=
1000
3
⁄
.
This
ge
ner
at
e
d
an
out
pu
t
with:
=
15
,
ℎ
∗
(
)
=
26
.
10869
,
1
∗
(
)
=
7127
.
522
3
⁄
,
2
∗
(
)
=
470
.
5617
3
⁄
,
3
∗
(
)
=
1626
.
928
3
⁄
,
4
∗
(
)
=
1714
.
475
3
⁄
,
(
)
=
501
.
4565
,
‖
‖
=
0
.
042747
(
)
=
0
.
0
0
5
3
6
28
.
Figure
1. Re
ser
vo
i
r head
v
e
rs
us
ti
m
e (case 1)
Figure
2. O
ptim
u
m
co
ntro
l
(
case 1
)
Plots
showi
ng
the
perf
or
m
ance
an
d
c
har
act
erist
ic
s
of
the
ste
epest
desce
nt
al
gorithm
a
re
show
n
i
n
Figures
3
to
6.
The
plo
t
in
Fi
gure
3
prese
nts
the
su
cc
essiv
e
it
erati
on
s
determ
ined
by
us
i
ng
a
unidirect
i
on
a
l
search
al
on
g
th
e
desce
nt on
th
e
gradie
nt v
ect
or
to
t
he
l
ocal m
ini
m
u
m
.
A
pe
culia
r
f
eat
ure of
this
al
gorith
m
can
be
obser
ve
d
by
stu
dying
1
(
)
2
(
)
.
W
hile
1
(
)
ke
pt
i
ncrea
sing
afte
r
e
ver
y
it
erati
on,
2
(
)
only
increases
for t
he
f
irst t
wo it
era
ti
on
s a
nd it
d
ec
reases ti
ll
the a
ll
ow
able t
olera
nce level
.
A
neces
sary
co
nd
it
io
n
f
or
a
c
on
t
ro
l
to
be
op
tim
u
m
is
that
t
he
pe
rfor
m
anc
e
ind
e
x
at
the
la
st
it
erati
on
m
us
t
be
m
ini
m
um
.
This
ca
n
be
obser
ve
d
i
n
Figure
4
t
hat
the
perform
ance
ind
e
x
kee
ps
decr
easi
ng
ti
ll
the
la
st
it
erati
on
.
If
the
c
urve
dif
f
ers
f
r
om
this
ex
pected
be
hav
i
our,
the
n
the
c
on
t
ro
l
is
no
t
opti
m
u
m
,
and
the
tra
j
ect
or
ie
s
m
ay
no
t
be
a
s
seen
in
Fig
ure
1.
It
ca
n
be
obser
ved
th
at
the
first
th
r
ee
it
erati
on
s
r
edu
c
e
the p
e
rfo
rm
ance ind
e
x g
reatl
y, this is a
uniq
ue
cha
racteri
sti
c
of stee
pe
st des
cent m
et
ho
d.
Figures
5
an
d
6
sho
w
the
va
riat
ion
of
the
norm
of
gr
a
di
ent
and
t
he
ste
epest
desce
nt
ste
ps
siz
e.
The
st
opping
crit
eria
in
the
al
gorithm
are
that
the
nor
m
of
gr
a
dient
m
us
t
be
ap
pro
xim
at
e
ly
zero
a
nd
the
ste
epest
de
scent
ste
p
siz
e
sh
oul
d
al
so
be
ver
y
sm
al
l
a
nd
insig
nificant.
It
is
on
ly
the
n
that
the
co
ntr
ol
can
be
ass
um
ed
to
be
op
ti
m
u
m
.
A
s
co
uld
be
see
n
duri
ng
the
im
plem
entat
ion
,
the
gr
a
dient
at
t
ai
ns
the
lo
west
value
and
t
he
ste
p
siz
e
becam
e
s
o
sm
all
that
t
he
cha
nges
c
om
pu
te
d
we
re
within
the
er
ror
of
com
pu
t
at
ion
.
This m
eans th
a
t t
he
hea
d
i
n Fi
gure
1
a
nd c
ontrol in
Fig
ur
e
2 are
optim
u
m
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
10
, No
.
4
,
A
ugus
t
2020
:
4136
-
4144
4142
Figure
3
.
Co
ntr
ol in
pu
t
ve
rsus
it
erati
on
(case
1)
Figure
4
.
Per
f
orm
ance ind
e
x versus it
erati
on (
case
1)
Figure
5
.
N
orm
o
f
gr
a
dient
versu
s
it
erati
on
Figure
6
.
St
eep
est
d
esce
nt ste
ps
ver
s
us i
te
rati
on
-
Ca
se 2
:
(
5
,
25
.
8
,
1
,
>
0
=
3000
3
⁄
)
It
is
possi
ble
to
m
od
ify
the
pro
blem
su
ch
that
the
c
on
t
rol
is
pe
naliz
ed.
The
pe
nalty
is
to
im
po
se
a
m
axi
m
u
m
and
m
ini
m
u
m
va
lue
on
the
co
nt
ro
l.
The
c
ontr
ol
is
no
t
al
lo
wing
to
be
negat
iv
e
or
e
xcee
d
a
va
lue.
Diff
e
re
nt
sim
ulati
on
f
or
pen
a
li
zed
opti
m
a
l
con
t
ro
ls
is
pre
sented
in
ca
se
s
2
an
d
3
.
Fig
ur
es
7
an
d
8
presents
the
res
ults
of
t
he
case
(
5,
25.
8,
1,
>
0
=
3000
3
⁄
.
The
e
f
fect
is
that
the
ove
rsho
ot
on
the
sta
te
trajector
y
is
rem
ov
ed,
an
d
it
ta
kes
m
or
e
it
erati
ons.
The
opti
m
u
m
con
trol
lo
oks
m
or
e
ap
propriat
e
than
t
he results
prod
uced with
cases
wh
e
re t
he
contr
ol is Un
pen
al
iz
e
d.
Figure
7
.
Re
ser
vo
i
r head
v
e
rs
u
s ti
m
e (case
2
)
Figure
8.
O
ptim
u
m
co
ntro
l
(
case
2
)
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
A steepest
desc
ent a
l
gorit
hm for
t
he op
ti
m
al
con
tr
ol
…
(
Ol
al
ekan
O
gunbiy
i
)
4143
T
h
e
i
n
p
u
t
p
a
r
a
m
e
t
e
r
s
a
r
e
:
(
0
)
=
25
.
8
,
ℎ
(
)
=
26
.
1
,
1
(
0
)
=
2
(
0
)
=
3
(
0
)
=
4
(
0
)
=
1000
3
⁄
.
T
h
i
s
g
e
n
e
r
a
t
e
d
a
n
o
u
t
p
u
t
w
i
t
h
:
=
54
,
ℎ
∗
(
)
=
26
.
0905
,
1
∗
(
)
=
30
0
0
3
⁄
,
2
∗
(
)
=
30
0
0
3
⁄
,
3
∗
(
)
=
3000
3
⁄
4
∗
(
)
=
11
4
8
.
7
4767
3
⁄
.
Ca
se 3
:
(
3
,
25
.
8
,
1
,
>
0
=
3000
3
⁄
)
The
res
ults
in
Fig
ur
es
9
and
10
are
f
or
th
e
case
with
three
operati
ng
m
ac
hin
es
w
hile
>
0
=
3000
3
⁄
.
T
he
al
gorith
m
co
nv
e
r
ges
af
te
r
one it
erati
on and the o
ptim
al
co
ntro
l
reduces
gr
a
dual
ly
after
eve
ry
quarter
of
ti
m
e.
The
i
np
ut
p
a
ram
et
ers
are:
ℎ
(
0
)
=
25
.
8
,
ℎ
(
)
=
26
.
1
,
1
(
0
)
=
2
(
0
)
=
3
(
0
)
=
4
(
0
)
=
1000
3
⁄
.
This
ge
ne
rated
a
n
outp
ut
with
:
=
2
,
ℎ
∗
(
)
=
26
.
1
774
1
,
1
∗
(
)
=
3000
3
⁄
,
2
∗
(
)
=
26
9
7
.
7157
3
⁄
,
3
∗
(
)
=
1996
.
9178
3
⁄
4
∗
(
)
=
13
1
2
.
0
8281
6
3
⁄
.
Figure
9
.
Re
ser
vo
i
r head
v
e
rs
us
ti
m
e (case
3
)
Figure
10
.
O
ptim
u
m
co
ntro
l
(
case 3
)
4.
CONCL
US
I
O
N
The
wor
k
has
consi
der
e
d
t
he
dev
el
op
m
ent
of
an
opti
m
a
l
con
tr
ol
proce
dur
e
for
the
casca
ded
K
HEPS
and
J
HEPS
.
T
her
e
ca
n
be
num
ero
us
opti
m
al
con
trol
pr
ob
le
m
,
bu
t
th
e
prob
le
m
so
lved
is
the
La
gr
a
nge
pro
blem
fo
r
the
opti
m
a
l
release
for
in
flo
ws
into
JHE
P
s
su
c
h
that
it
s
op
e
rati
ng
hea
d
rem
ai
ns
relat
ively
const
ant
at
26
.
1
m
.
The
c
on
t
ro
l
al
gorithm
is
base
d
on
t
he
ste
epest
de
scent,
a
m
et
ho
d
that
ens
ur
e
the
determ
inati
on
of
a
l
oca
l
op
ti
m
u
m
fo
r
a
co
nvex
pro
blem
.
The
qu
a
drat
ic
li
ne
was
em
plo
ye
d
f
or
the
determ
inatio
n
of
optim
um
ste
epest
descen
t
ste
p
siz
e.
W
he
n
the
con
t
ro
l
is
penal
iz
ed,
the
al
go
rithm
conve
rg
es
fast
e
r.
T
he
co
ntr
ol
al
go
rithm
was
i
m
ple
m
ented
in
an
E
xcel
VBA
®
en
vir
onm
ent
to
ensure
that
the
opti
m
u
m
h
ead
fall
s
within
0
.
03%
.
It
is
rec
omm
end
ed
f
or
use
in
syst
em
st
ud
ie
s
,
decisi
on
m
aking
a
nd
con
t
ro
l
desig
n for the
opti
m
al
powe
r ge
ner
at
ion
i
n
the
casc
aded hy
dr
opower
syst
em
ACKN
OWLE
DGE
MENTS
We
ack
olwe
dg
e
the
Tran
sm
i
ssion
C
om
pan
y
of
Nige
ria
(
TCN)
Nati
on
a
l
con
tr
ol
Ce
ntr
e
Osho
gbo,
for
pro
vid
in
g
r
el
evan
t
data
use
d
i
n
the
c
ouses
of
t
his
re
s
earch
.
We
a
re
aso
gr
at
e
fu
l
t
o
the
m
anag
em
ent
of
Ma
instream
En
erg
y s
olu
ti
on
f
or grantin
g
ac
c
ess the t
wo h
y
dro
power sta
ti
on
s
.
REFERE
NCE
S
[1]
A.
S.
Sam
bo,
B.
Garba
,
I.
H.
Za
rm
a,
and
M
.
M.
Gaji
,
“
Elec
t
ric
ity
Gen
erati
o
n
and
the
Prese
nt
Chal
l
eng
es
i
n
the
Nig
eri
an
Po
wer
Sect
or
,
”
J
ou
rnal
Ene
rgy
Po
wer
Engi
n
ee
ring
,
vol
.
6
,
no
.
7
,
pp
.
1
–
17
,
2012
.
[2]
M.
A.
Am
inu
and
U.
G.
Kangi
wa,
“
Perform
anc
e
Ev
al
ua
ti
on
a
nd
Eff
iciency
I
m
prove
m
ent
for
Jebba
,
Kain
ji
a
nd
Shiroro
H
y
dro
P
ower
Schemes,”
3rd
IEE
E
In
te
r
nati
onal
Conf
er
enc
e
on
Adaptiv
e
Scienc
e
and
T
ec
hnolog
y
(
ICAST
2011)
,
Abuja, pp
.
115
-
118
,
2011
.
[3]
C.
T.
Thomas,
M.
F.
Akorede
,
O.
Ogunbi
y
i
,
B.
J.
Olufe
agba
,
a
nd
J.
S.
Sam
uel
,
“
A
study
of
en
erg
y
conv
ersion
at
the
Jebb
a
H
y
dr
oel
e
ct
ri
c
power
stat
ion
,
”
2017
IEEE
3rd
Int
e
rnational
Conf
e
renc
e
on
Elec
tro
-
Technol
ogy
fo
r
Nati
onal
D
ev
e
lo
pment
(
NIGERCON)
,
Ow
err
i,
pp.
82
8
-
833
,
2017.
[4]
O.
M.
Bamigbola
and
Y.
O.
Ade
rint
o,
“
On Opti
m
al
Control
Model
of
Elec
tr
ic
Pow
er
Gene
rating
Sy
st
ems
,
”
Jour
nal
Re
s. Phys. Sc
i.
,
vol.
5
,
no
.
1
,
pp
.
59
–
70,
2015
.
[5]
T.
S.
Abdulkadi
r
,
A.
W
.
Sala
m
i,
A.
R.
Anw
ar,
and
A.
G.
Kare
em,
“
Modell
ing
of
H
y
dropow
er
Re
servoir
Vari
able
s
for
Ene
rg
y
Gen
e
rat
ion:
Neura
l
Network
Approac
h,
”
E
thi
opian
Jo
urnal
of
Environm
ent
al
Stud
ie
s
a
nd
Manage
ment
,
vol.
6
,
no
.
3
,
pp
.
310
–
316,
Ma
y
2
013.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
10
, No
.
4
,
A
ugus
t
2020
:
4136
-
4144
4144
[6]
C.
T.
Thomas,
O
.
Ogunbi
y
i,
M.
F.
Akorede
,
and
J.
B.
Olufe
agb
a,
“
Assess
m
ent
of
Fail
ure
a
nd
R
ep
ai
r
Beha
v
iours
of
the
Jebba
H
y
dr
o
el
e
ct
ri
c
Pow
er
Stat
ion
,
”
E
lektri
ka
Journal
of
E
le
c
tric
al
Engi
ne
ering
,
vol
.
17,
n
o.
3,
pp
.
13
–
19,
Dec
.
2018
.
[7]
C.
T.
Thomas,
M.
F.
Akorede
,
O.
Ogunbi
y
i
,
B.
J.
Olufe
agba
,
a
nd
J.
S.
Sam
uel
,
“
A
study
of
en
erg
y
conv
ersion
at
the
Jebb
a
H
y
dr
oel
e
ct
ri
c
power
stat
ion
,
”
2017
IEEE
3rd
Int
e
rnational
Conf
e
renc
e
on
Elec
tro
-
Technol
ogy
fo
r
Nati
onal
D
ev
e
lo
pment
(
NIGERCON)
,
Ow
err
i,
pp.
828
-
833,
2017.
[8]
O.
Ogunbi
y
i
,
C.
T.
Thomas,
I.
A.
O.
Om
eiza
,
J.
Akanni
,
and
B.
J.
Olufe
agb
a
,
“
D
y
n
amica
l
C
o
ntrol
Mode
l
of
the
C
asc
ade
d
Ka
inj
i
-
Jebb
a
H
y
dro
power
Opera
ti
ng
Hea
d,
”
FUO
Y
E
Journal
o
f
Eng
i
nee
ring
and
Tec
hnol
ogy
,
vol
.
4,
no.
1
,
pp
.
140
-
1
44,
Ma
y
2019.
[9]
S.
Guo,
X.
Li,
P.
Li
u
,
and
F.
Guo,
“
Optimal
Opera
ti
on
of
Ca
sca
de
H
y
dropo
wer
Plant
s,
”
in
2009
Asia
-
Pa
ci
f
ic
Powe
r and Ener
gy
Eng
ine
ering
Confe
renc
e
,
W
uhan,
pp
.
1
-
4
,
200
9.
[10]
A.
Aja
y
i
,
e
t
al
.
,
“
Major
Fact
or
s
Affec
ti
ng
El
e
ct
ri
ci
t
y
Gene
r
ati
on,
Tra
nsm
ission
and
Distribut
i
on
in
Niger
ia
,
”
Et
hiop
.
J
.
Environ.
Stud
.
Manag
.
,
vol. 1, no. 3
,
2
013.
[11]
S.
A.
A.
Moos
avi
an,
A.
Gh
af
ari,
A.
Sali
m
i,
and
N.
Abdi,
“
Non
-
li
ne
ar
Multi
obj
e
ct
iv
e
Optimizati
on
for
Control
of
H
y
dropower
Pla
nts
Network,
”
2
008
IEE
E/ASME
Inte
rnational
C
onfe
renc
e
on
Ad
vanc
ed
Int
elligent
Me
chat
roni
cs
,
Xian,
pp
.
1278
-
1
283,
2008
.
[12]
G.
Robert
and
F.
Micha
ud
,
“
A
Sim
ple
M
ult
i
-
Objec
t
ive
Con
tr
ol
for
C
asc
ade
d
H
y
dro
Pow
er
Plant
s,”
in
I
FAC
Proce
ed
ings V
ol
umes
,
vol
.
44
,
n
o.
1
,
pp
.
4960
–
4
963,
Jan
.
2011
.
[13]
A.
F.
Ribe
iro
,
M.
C.
M.
Gued
es,
G.
V.
Sm
irnov,
and
S.
Vil
el
a
,
“
On
the
Optimal
Control
of
a
Casca
d
e
o
f
H
y
dro
-
e
le
c
tric
P
owe
r
Stat
ions,”
El
e
ct
ric
Pow
er
S
yste
ms
Re
sear
ch
,
vol
.
88
,
pp
.
121
–
129,
Jul.
2012.
[14]
F.
R.
Paiva,
“
Num
eri
ca
l
Me
tho
ds
for
Optimal
Control
and
Mo
del
Predictiv
e
Control
,
”
Program
a
Doutoral
em
Mate
máti
ca
Ap
licada
,
2014
.
[15]
R.
Steng
el a
nd
O.
Contro
,
“
Princi
pl
es
for
Opt
imal
Co
n
trol
of
D
ynamic
S
y
s
te
m
s!
Tra
j
ec
tor
y
of the
S
y
stem!
,
”
2015
.
[16]
J.
L.
Schu
lt
z
,
D.
G.,
and
Me
lsa,
“
Stat
e
Funct
ions
and
Li
n
ea
r
Con
trol
S
y
st
ems
,
”
Mc
Gr
aw
Hill
Te
xt
;
F
irst
Edi
t
ion
,
Jan.
1967.
[17]
V.
M.
B
ecerra,
“
Optimal
Contro
l
,
”
S
chol
arped
ia
,
vol.
3
,
no
.
1
,
200
8.
[18]
O.
Vo
n
Str
y
k,
“
Nu
m
eri
ca
l
Sol
uti
on
of
Optimal
Control
Prob
le
m
s
b
y
Dire
ct
Coll
ocation
,
”
Optimal
Control
,
vol.
111
,
pp
.
129
–
143,
1993
.
[19]
S.
S.
R
ao,
“
Enginee
ring
Op
ti
m
izati
on:
Th
eor
y
an
d
Prac
t
ic
e
,
”
Wi
ley
4
edition
,
Jul.
2009.
[20]
J.
T.
Bet
ts
,
“
Prac
tical
Methods
f
or
Opti
m
al
Cont
rol
Us
ing
Nonline
ar
Program
m
ing,
”
Soc
ie
t
y
fo
r
Industrial
and
Appl
ie
d
Math
emati
cs
,
2001.
[21]
D.
E
.
Kirk
,
“
Opt
imal
Contro
l
Th
e
or
y
:
An
Introdu
ct
ion
,
”
Dove
r
Pu
bli
cation
,
New
York,
1998.
[22]
C.
Y.
T
.
Ma,
D
.
K.
Y.
Yau,
N.
K.
Yip,
N.
S.
V.
Rao,
and
J.
Ch
en,
“
Stoc
hastic
Stee
pest
Desc
en
t
Optimization
o
f
Multi
ple
-
Obj
ecti
ve
Mobil
e
Senso
r
Covera
g
e,”
20
10
IEEE
30t
h
In
te
rnational
Conf
ere
nce
on
Distri
bute
d
Comput
in
g
Syste
ms
,
Genov
a
,
pp
.
96
-
105
,
20
10.
[23]
H.
Schät
tler
an
d
T.
J.
Ta
rn,
“
Desce
nt
Algorithm
s
for
Optim
al
Periodi
c
Ou
t
put
Feedba
ck
C
ontrol
,
”
in
IEEE
Tr
ansacti
ons on Automati
c
Contr
ol
,
vo
l. 37, no. 1
2,
pp
.
1893
-
190
4,
De
c. 1992.
[24]
K.
S.
Sriniva
s
a
nd
K.
Vaisa
kh,
“
Applic
at
ion
of
Stee
pest
Desc
en
t
and
Diffe
r
ent
i
a
l
Evol
ut
ion
Met
hods
for
Optimal
Control
of
Pow
e
r
S
y
stems
,
”
Imp
erial
Journal
o
f
Inte
rdiscip
li
nary
Res
earc
h
(
IJI
R)
,
vol.
3,
no
.
1
,
pp.
2257
–
2263
,
2017.
[25]
H.
Zha
o,
Y.
Zh
ang,
J.
Hu,
and
Z.
Chen,
“
Esti
m
at
ion
of
Init
i
al
Guess
of
Stee
pe
st
Desce
nt
Method
for
Nea
r
Fiel
d
Phase
Ret
r
ie
va
l,”
2017
Asia
-
Pa
c
if
ic
Int
ernati
ona
l
Symposium
on
El
e
ct
rom
agnet
i
c
Compatibi
l
it
y
(
AP
EMC)
,
Seoul
,
pp.
306
-
308
,
20
17.
[26]
W
.
J.
Grant
ham,
A.
O,
and
Chin
gcua
nco
,
“
L
y
ap
unov
Stee
pest
D
esc
ent
Contro
l
o
f
Constrai
ned
L
i
nea
r
S
y
st
ems
,
”
IEE
E
Tr
ansacti
o
ns on
Aut
omat
ic
Control
,
vo
l
.
29,
no.
8,
pp.
740
-
7
43,
Aug.
1984.
[27]
O.
Og
unbi
y
i,
C.
T.
Thomas,
I
.
O.
A.
Om
ei
z
a,
J.
Akanni
,
and
B
.
J.
Oluf
ea
gb
a,
“
A
Nonline
ar
Co
ntrol
Model
and
Opera
ti
on
al
Support
S
y
stem
for
the
Kainji
H
y
droe
lectr
i
c
Pow
er
S
y
stem,
”
Ni
gere
rian
Journal
of
Technol
og
y
,
vol.
38
,
no
.
2
,
pp
.
449
–
455
,
2019
.
Evaluation Warning : The document was created with Spire.PDF for Python.