Inequality Constraints in Optimization

l Practical problems contain inequality as well as equality

constraints
l (
Minimize the cost function f x1 , x2 , L xn )
u subject to the equality constraints
gi ( x1 , x2 ,L xn ) = 0 i = 1, 2,L, k
u and the inequality constraints
u j ( x1 , x2 , L xn ) ≤ 0 j = 1,2, L , m
l The Lagrange multiplier is extended to include the
inequality constraints by introducing the m-dimensional
vector µ of undetermined quantities

Power Systems I
Kuhn-Tucker Method
l The unconstrained cost function becomes
L = f + ∑ λi gi + ∑ µ j u j
i j

l The resulting necessary conditions for contrained local

minima of L are the following

∂L = 0 l = 1,L, n ∂L = gi = 0 i = 1,L, k
∂xl ∂λi
∂L = u j ≤ 0 j = 1,L, m µ j u j = 0 & µ j > 0 j = 1,L, m
∂µ j

Power Systems I
Example
l Use the Kuhn-Tucker method to determine the minimum
distance from the origin of the x-y plane to a circle
described by
(x − 8)2 + ( y − 6)2 = 25 or
g ( x, y ) = ( x − 8) + ( y − 6 ) − 25
2 2

constrained by
u ( x, y ) = 2 x + y ≥ 12
u The minimum distance is
obtained by minimizing the
distance squared
f ( x, y ) = x 2 + y 2
Power Systems I
Example
f ( x, y ) = x 2 + y 2 g ( x, y ) = ( x − 8) + ( y − 6 ) − 25 = 0
2 2

u ( x, y ) = 2 x + y ≥ 12
The cost function
L = f + λ ⋅ g + µ ⋅u
[ ]
= x 2 + y 2 + λ ( x − 8) + ( y − 6 ) − 25 + µ [2 x + y − 12]
2 2

The resulting necessary conditions for constrained local minima of L

∂L ∂L
= 2 x + 2λ ( x − 8) + 2 µ = 0 = 2 y + 2λ ( y − 6 ) + µ = 0
∂x ∂y
∂L ∂L
= ( x − 8) + ( y − 6) − 25 = 0 = 2 x + y − 12 = 0
2 2

∂λ ∂µ
Power Systems I
Example
u eliminating λ from the first two equations
16λ 12λ 3
= →y= x
2x 2y 4
u substituting for y in the third equation yields
2

(x − 8) +  x − 6  − 25 = 0
2 3
4 
25 2
x − 25 x + 75 = 0 → x = 4 & x = 12
16
extrema : (4,3), λ = 1 and (12,9 ), λ = −3
x=4
min →
y=3
Power Systems I
Economic Dispatch with Generator Limits
l The power output of any generator should not exceed its
rating nor be below the value for stable boiler operation
u Generators have a minimum and maximum real power output
limits
l The problem is to find the real power generation for each
plant such that cost are minimized, subject to:
u Meeting load demand - equality constraints
u Constrained by the generator limits - inequality constraints
l The Kuhn-Tucker conditions
dCi dPi = λ ← Pi (min) < Pi < Pi (max)
dCi dPi ≤ λ ← Pi = Pi (max)
dCi dPi ≥ λ ← Pi = Pi (min)
Power Systems I
Example
l Neglecting system losses, find the optimal dispatch and
the total cost in $/hr for the three generators and the
given load demand and generation limits
C1 = 500 + 5.3P1 + 0.004 P12 [$ / MWhr ]
C2 = 400 + 5.5P2 + 0.006 P22
C3 = 200 + 5.8P3 + 0.009 P32
200 ≤ P1 ≤ 450
150 ≤ P2 ≤ 350
100 ≤ P3 ≤ 225
PDemand = 975 MW
Power Systems I
Example
βi
n gen

PDemand + ∑
i =1 2γ i 800 + 0.5008 + 0.5012 + 05.018
.3 .5 .8
λ= = = $8.5 / MWhr
0.008 + 0.012 + 0.018
ngen 1 1 1
1
∑ i =1 2γ i

8.5 − 5.3
P1 = = 400 MW
2(0.004 )
λ − βi 8.5 − 5.5i
Pi = ⇒ P2 = = 250 MW
2γ i 2(0.006 )
8.5 − 5.8
P3 = = 150 MW
2(0.009 )
PDemand = 800 MW = 400 + 250 + 150 MW

Power Systems I
Example
l Solve using iterative methods
u initial guessλ[1] = 6.0
[1] 6.0 − 5.3 [1] 6 .0 − 5 .5
P1 = = 87.5 P2 = = 41.7
2(0.004 ) 2(0.006)
[1] 6.0 − 5.8
P3 = = 11.1
2(0.009 )
∆P [1] = 975 − (87.5 + 41.7 + 11.1) = 834.7
[1]
∆P (834.7)
∆λ[1] = [1] = = 3.1632
(∂P ∂λ ) 2 (0.004 ) + 2 (0.006 ) + 2 (0.009 )
1 1 1

λ[2 ] = 6.0 + 3.1632 = 9.1632

Power Systems I
Example
[2 ]9.16 − 5.3
λ[2 ] = 9.1632 P1 = = 483
2(0.004 )
[2 ] 9.16 − 5.5 [2 ] 9.16 − 5.8
P2 = = 305 P3 = = 187
2(0.006) 2(0.009 )
∆P [2 ] = 975 − ( 483 + 305 + 187) = 0
[2 ]
P1 = 483 > 450max → P1 = 450max
[ FIXED ]

∴ ∆P [2 ] = 975 − ( 450 + 305 + 187) = 33

[2 ]
∆P (33)
∆λ[2 ] = [2 ] = = 0.2368
(∂P ∂λ ) 2 (0.006 ) + 2 (0.009 )
1 1

λ[3] = 9.1632 + 0.2368 = 9.4

Power Systems I
Example
[3]
λ = 9.4 P 1
[ FIXED ]
= 450
[3] 9.4 − 5.5 [3] 9.4 − 5.8
P2 = = 325 P3 = = 200
2(0.006) 2(0.009 )
∆P [3] = 975 − ( 450 + 325 + 200) = 0

( [3]
)
150 ≤ P2 = 325 ≤ 350 ( [3]
100 ≤ P3 = 200 ≤ 225 )
Ctotal = 500 + 5.3(450) + 0.004(450)
2

( ) (
+ 400 + 5.5 325 + 0.006 325 )2

+ 200 + 5.8(200) + 0.009(200) = 8,236.25 $ / hr

2

Power Systems I
Economic Dispatch including Losses
l For large interconnected system where power is
transmitted over long distances with low load density
areas
u transmission line losses are a major factor
u losses affect the optimum dispatch of generation
l One common practice for including the effect of
transmission losses is to express the total transmission
loss as a quadratic function of the generator power
outputs n gen n gen

u simplest form: PL = ∑∑ Pi Bij Pj

i =1 j =1
n gen n gen n gen

u Kron’s loss formula: PL = ∑∑ Pi Bij Pj + ∑ B0 j Pj + B00

i =1 j =1 j =1
Power Systems I
Economic Dispatch including Losses
l Bij are called the loss coefficients
u they are assumed to be constant
u reasonable accuracy is expected when actual operating
conditions are close to the base case conditions used to compute
the coefficients
l The economic dispatch problem is to minimize the overall
generation cost, C, which is a function of plant output
l Constraints:
u the generation equals the total load demand plus transmission
losses
u each plant output is within the upper and lower generation limits -
inequality constraints

Power Systems I
Economic Dispatch including Losses
n gen n gen
f : Ctotal = ∑ Ci = ∑αi +β i Pi + γ i Pi
2

i =1 i =1
n gen
g: ∑P = P
i =1
i demand + Plosses

u : Pi (min ) ≤ Pi ≤ Pi ( max ) i = 1,L, ngen

The resulting optimization equation
 n gen
 ngen
L = Ctotal + λ  Pdemand + Plosses − ∑ Pi  + ∑ µi ( max ) (Pi ( max ) − Pi )
 i =1  i =1
n gen
+ ∑ µi (min ) (Pi − Pi (min ) )
i =1

Pi < Pi (max ) : µi (max) = 0 Pi > Pi (min ) : µi (min) = 0

Power Systems I
Economic Dispatch including Losses
l The minimum of the unconstrained function is found
when:
∂L
=0
∂Pi
∂L
=0
∂λ
∂L
= Pi − Pi (max) = 0
∂µi (max)
∂L
= Pi − Pi (min) = 0
∂µi (min)

Power Systems I
Economic Dispatch including Losses
l When generator limits are not violated:
∂L ∂Ctotal  ∂PL 
=0= + λ  0 + − 1
∂Pi ∂Pi  ∂Pi 
∂Ctotal
∂Pi
=
∂
∂Pi
(C1 + C2 + L + Cngen =) dCi
dPi
dCi ∂PL  1  dCi dCi
∴ λ= +λ =   = Li
dPi ∂Pi  1 − ∂PL ∂Pi  dPi dPi
∂L
n gen n gen
= 0 = PD + PL − ∑ Pi ∴ ∑ Pi = PD + PL
∂λ i =1 i =1

Power Systems I
The Penalty Factor
l The incremental transmission loss equation becomes the
penalty factor
1
Li =
∂PL
1−
∂Pi

u The effect of transmission losses introduces a penalty factor that

depends on the location of the plant
u The minimum cost is obtained when the incremental cost of each
plant multiplied by its penalty factor is the same for all plants

Power Systems I
The Penalty Factor
n gen n gen n gen
PL = ∑∑ Pi Bij Pj + ∑ B0 j Pj + B00
i =1 j =1 j =1

∂PL
n
gen
dCi
= 2∑ Bij Pj + B0i = β i + 2γ i P
∂Pi j =1 dPi
∂PL
n gen
dCi
λ= +λ = β i + 2γ i Pi + 2λ ∑ Bij Pj + λB0i
dPi ∂Pi j =1

Rearrange the equation

γi βi 
n gen
 1
 + Bii  Pi + ∑ Bij Pj = 2 1 − B0i − 
λ  j =1  λ
j ≠i
Power Systems I
The Penalty Factor
l Extending the equation to all plants results in the
following linear equations (in matrix form)

γ 1   β1 
1 − B01 − 
 λ + B11 B12 L B1n 
 P1   λ
 γ2    β 
 B21 + B22 L B2 n   P2  = 1 1 − B02 − 2 
 λ  M  2  λ
 M M O M    
γn  βn 
 B + Bnn   n 
P
Bn 2 L
1 − B0 n − 
 n1
λ 
  λ
or
EP=D

Power Systems I
Economic Dispatch including Losses
l To find the optimal dispatch for an estimated value of λ[1]
u The simultaneous linear equation is solved, E P = D
u Then the iterative process is continued using the gradient method
u If an approximate loss formula is used
n gen
PL = ∑ Bii Pi 2
i =1
the solution of the simultaneous equations reduces to

λ − βi
[k ] n gen
 ∂Pi 
n gen
 γ i + Bii β i 
Pi [k ]
=
2(γ i + λ Bii )
∑   =∑  2
i =1  ∂λ   2(γ i + λ Bii ) 
2
i =1 
[k ] [ k ]

Power Systems I
Example
l Find the optimal dispatch and the total cost in $/hr
u fuel costs and plant output limits
C1 = 200 + 7.0 P1 + 0.008P12 [$ / hr ] 10 ≤ P1 ≤ 85 MW
C2 = 180 + 6.3P2 + 0.009 P22 10 ≤ P2 ≤ 80
C3 = 140 + 6.8P3 + 0.007 P32 10 ≤ P3 ≤ 70
u real power loss and total load demand
Ploss = 0.000218 P12 + 0.000228 P22 + 0.000179 P32
PDemand = 150 MW

Power Systems I
Example
First iteration with initial guess at 8.0
λ[ k ] − βi
λ = 8.0
[0]
Pi =
[k ]

2(γ i + λ[ k ] Bii )
2

8.0 − 7.0
P1 =
[1]
= 51.3
2(0.008 + 8.0 ⋅ 0.000218)
2

8.0 − 6.3
P2[1] = = 78.5
2(0.009 + 8.0 ⋅ 0.000228)
2

8.0 − 6.8
P3 =
[1]
= 71.2
2(0.007 + 8.0 ⋅ 0.000179 )
2

Ploss = 0.000218(51.3) + 0.000228(78.5) + 0.000179(71.2 )

2 2 2

Ploss = 2.886
Power Systems I
Example
∆P[ k ] = PD + PL[ k ] − (P1[ k ] + P2[ k ] + P3[ k ] ) Newton-Raphson
∆P[1] = 150 + 2.9 − (51.3 + 78.5 + 71.2 ) = −48.1
 ∂Pi  γ i + Bii β i
3 [k ] 3
J [k ]
= ∑ =∑
i =1 2 (γ i + λ Bii )
the Jacobian
i =1  ∂λ 
[k ] 2

0.008 + 0.000218 × 7.0 0.009 + 0.000228 × 6.3

J [1] = + +
2(0.008 + 8.0 × 0.000218) 2(0.009 + 8.0 × 0.000228)
2 2

0.007 + 0.000179 × 6.8

= 152.5
2(0.007 + 8.0 × 0.000179 )
2

− 48.1
∆λ[1] = = −0.316 the improvement
152.5
λ[ 2 ] = λ[1] + ∆λ[1] = 8.0 − 0.316 = 7.684
Power Systems I
Example
Second iteration
λ [k ]
− βi
λ[ 2 ] = 7.684 Pi [ k ] =
2(γ i + λ Bii )
[k ] 2

7.684 − 7.0
P1 =
[2]
= 35.4
2(0.008 + 7.684 ⋅ 0.000218)
2

7.684 − 6.3
P2 =
[2]
= 64.4
2(0.009 + 7.684 ⋅ 0.000228)
2

7.684 − 6.8
P3 =
[2]
= 52.8
2(0.007 + 7.684 ⋅ 0.000179 )
2

( ) ( ) (
Ploss = 0.000218 35.4 + 0.000228 64.4 + 0.000179 52.8
2 2
)2

Ploss = 1.717
Power Systems I
Example
∆P[ k ] = PD + PL[ k ] − (P1[ k ] + P2[ k ] + P3[ k ] ) Newton-Raphson
∆P[ 2 ] = 150 + 1.7 − (35.4 + 64.4 + 52.8) = −0.84
 ∂Pi  γ i + Bii βi
3 [k ] 3
J [k ]
= ∑  =∑
i =1 2 (γ i + λ Bii )
the Jacobian
i =1  ∂λ 
[k ] 2

0.008 + 0.000218 × 7.0 0.009 + 0.000228 × 6.3

J [2] = + +
2(0.008 + 7.68 × 0.000218) 2(0.009 + 7.68 × 0.000228)
2 2

0.007 + 0.000179 × 6.8

= 154.6
2(0.007 + 7.68 × 0.000179 )
2

− 0.84
∆λ[ 2 ] = = −0.00543 the improvement
154.6
λ[ 3] = λ[ 2 ] + ∆λ[ 2 ] = 7.684 − 0.0.00543 = 7.679
Power Systems I
Example
Third iteration
λ [k ]
− βi
λ[ 3] = 7.679 Pi [ k ] =
2(γ i + λ Bii )
[k ] 2

7.679 − 7.0
P1[ 3] = = 35.09
2(0.008 + 7.679 ⋅ 0.000218)
2

7.679 − 6.3
P2 =
[ 3]
= 64.14
2(0.009 + 7.679 ⋅ 0.000228)
2

7.679 − 6.8
P3 =
[ 3]
= 52.48
2(0.007 + 7.679 ⋅ 0.000179 )
2

Ploss = 0.000218(35.09 ) + 0.000228(64.14 ) + 0.000179(52.48)

2 2 2

Ploss = 1.699
Power Systems I
Example
∆P[ k ] = PD + PL[ k ] − (P1[ k ] + P2[ k ] + P3[ k ] ) Newton-Raphson
∆P[ 3] = 150 + 1.697 − (35.09 + 64.14 + 52.48) = −0.0174
 ∂Pi  γ i + Bii βi
3 [k ] 3
J [k ]
= ∑  =∑
i =1 2 (γ i + λ Bii )
the Jacobian
i =1  ∂λ 
[k ] 2

0.008 + 0.000218 × 7.0 0.009 + 0.000228 × 6.3

J [ 3] = + +
2(0.008 + 7.679 × 0.000218) 2(0.009 + 7.679 × 0.000228)
2 2

0.007 + 0.000179 × 6.8

= 154.6
2(0.007 + 7.679 × 0.000179 )
2

− 0.0174
∆λ[ 3] = = −0.000113 the improvement
154.6
λ[ 4 ] = λ[ 3] + ∆λ[ 3] = 7.684 − 0.0.000113 = 7.6789
Power Systems I
Example
Forth iteration
λ [k ]
− βi
λ[ 4 ] = 7.6789 Pi [ k ] =
2(γ i + λ Bii )
[k ] 2

7.6789 − 7.0
P1[ 3] = = 35.09
2(0.008 + 7.6789 ⋅ 0.000218)
2

7.6789 − 6.3
P2 =
[ 3]
= 64.13
2(0.009 + 7.6789 ⋅ 0.000228)
2

7.6789 − 6.8
P3 =
[ 3]
= 52.47
2(0.007 + 7.6789 ⋅ 0.000179 )
2

Ploss = 0.000218(35.09 ) + 0.000228(64.13) + 0.000179(52.47 )

2 2 2

Ploss = 1.699
Power Systems I
Example

∆P[ k ] = PD + PL[ k ] − (P1[ k ] + P2[ k ] + P3[ k ] )

∆P[ 2 ] = 150 + 1.699 − (35.09 + 64.13 + 52.47 ) = 0.0

Ctotal = 200 + 7.0(35.09 ) + 0.008(35.09 )

2

+ 180 + 6.3(64.13) + 0.009(64.13)

2

+ 140 + 6.8(52.47 ) + 0.007(52.47 )

2

= 1592.65 $ / hr

Power Systems I