You are on page 1of 55

2/16/2014

TỐI ƯU HÓA VẬN HÀNH


HỆ THỐNG ĐIỆN

Chapter 2:
INTRODUCTION TO OPTIMIZATION

Dr. Vo Ngoc Dieu


Department of Power Systems
Ho Chi Minh City University of Technology
Email: vndieu@gmail.com

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

10

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

11

12

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

13

14

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

Thm (Sufficient Condition)


∂f 1 ∂ 2f
f(x + ∆x) = f(x) + ⋅ ∆x + ⋅ ∆x 2 + H.O.T.
∂x 2 ∂x 2
∂f ∂ 2f
= 0, >0
∂x ∂x 2
Given a twice continuously differentiable function
f : IR n → IR1
x = x̂ is a local (relative) minimum
it is necessary that ∇f(x̂) = 0, ∇ 2 f(x̂) ≥ 0

it is sufficient that ∇f(x̂) = 0 and H = ∇ 2 f(x̂) ≥ 0 (positive definite)

 ∂ 2f ∂ 2f ∂ 2f 
 L 
 ∂x1 ∂x1∂x 2 ∂x1∂x n 
2


 M O M 
H= 
M O M 
 2 
 ∂ f L L
∂ 2f 
 ∂x n ∂x1 ∂x n 
2

n n
A > 0 ⇔ X T AX = ∑∑ x x a
i =1 j=1
i j ij > 0

pos. def. ∀X ≠ 0

15

1
Proof : f(X̂ + ∆X) = f(X̂) + ∇f(X̂) ⋅ ∆X + ∆X T ∇ 2 f(X̂)
2

If the solution of gradient set ∇f(X) is unique and if H is p.d., the solution is the global minimum.

EX : Determine the minimum of the solution


f(x) = 1 - x + 2x 2
and show that it is a global minimum

∂f
Solution = −1 + 4x = 0 ⇒ x̂ = 0.25
∂x
∂ 2f
= 4 > 0 ⇒ Global minimum
∂x 2
EX : Show that the minimum of the function
f(x 1 , x 2 ) = 10 + x1 - x 2 + 2x12 + x1x 2 + 4x 22
is the global minimum
 ∂f 
 ∂x   1 + 4x1 + x 2 
Solution ∇f(x 1 , x 2 ) =  1  =  =0
 ∂f  − 1 + x1 + 8x 2 
 ∂x 2 

16

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

17

18

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

19

20

10

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

21

22

11

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014



23

Minimization of function f(x1 , x 2 )


∂f 
∆∂x1 
∇f(x1 , x 2 ) = =0 Necessary condition
∂f 
∂x 2 
By Taylor' s expansion
∆x1
f(x̂1 + ∆x1 , x̂ 2 + ∆x 2 ) ≈ f(x̂1 , x̂ 2 ) + ∇f(x1 , x 2 ) ∆x 2 + H.O.T

∂f ∂f > 0
Suppose ≠ 0, Say
∂x1 ∂x1 < 0
± 0.1
Choose ∆x =
0 
∂f
f(x̂1 - 0.1, x̂ 2 + ∆x 2 ) ≈ f(x̂ 1 , x̂ 2 ) − ⋅ 0.1 < f(x̂1 , x̂ 2 )
∂x1
(This is not minimum)
Thus, necessary condition for a minimum or maximum points.

∇f(x) = 0

24

12

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

Thm (Sufficient Condition)


∂f 1 ∂ 2f
f(x + ∆x) = f(x) + ⋅ ∆x + ⋅ ∆x 2 + H.O.T.
∂x 2 ∂x 2
∂f ∂ 2f
= 0, >0
∂x ∂x 2
Given a twice continuously differentiable function
f : IR n → IR1
x = x̂ is a local (relative) minimum
it is necessary that ∇f(x̂) = 0, ∇ 2 f(x̂) ≥ 0

it is sufficient that ∇f(x̂) = 0 and H = ∇ 2 f(x̂) ≥ 0 (positive definite)

 ∂ 2f ∂ 2f ∂ 2f 
 L 
 ∂x1 ∂x1∂x 2 ∂x1∂x n 
2


 M O M 
H= 
M O M 
 2 
 ∂ f L L
∂ 2f 
 ∂x n ∂x1 ∂x n 
2

n n
A > 0 ⇔ X T AX = ∑∑ x x a
i =1 j=1
i j ij > 0

pos. def. ∀X ≠ 0

25

1
Proof : f(X̂ + ∆X) = f(X̂) + ∇f(X̂) ⋅ ∆X + ∆X T∇ 2f(X̂)
2

If the solution of gradient set ∇f(X) is unique and if H is p.d., the solution is the global minimum.

EX : Determine the minimum of the solution


f(x) = 1 - x + 2x 2
and show that it is a global minimum

∂f
Solution = −1 + 4x = 0 ⇒ x̂ = 0.25
∂x
∂ 2f
= 4 > 0 ⇒ Global minimum
∂x 2
EX : Show that the minimum of the function
f(x 1 , x 2 ) = 10 + x1 - x 2 + 2x12 + x1x 2 + 4x 22
is the global minimum

 ∂f 
 ∂x   1 + 4x1 + x 2 
Solution ∇f(x1 , x 2 ) =  1  =  =0
 ∂f   − 1 + x1 + 8x 2 
 ∂x 2 

26

13

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

4x 1 + x 2 = −1
x 1 + 8x 2 = 1
 − 18 
 x 1   62 
=
x   5 
 2  
 31 
 ∂ f 2
∂ 2f 
 
∂ x 2
∂x 1 ∂x 2  4 1
H =  21 =
 ∂ f ∂ f  1 9
2
 2 
 ∂x 2 ∂x 1 ∂x 2 

A matrix A ∈ IR n×n is positive definite if X T AX > 0 for all nonzero X ∈ IR 2 . Consider the 2 × 2 symmetical case.
 a 11 a 12 
A= 
a 21 a 22 
X = (1,0) T ⇒ X T AX = a 11 > 0
(a 11 + a 12 )
a 12 ≤ X = (0,1) T ⇒ X T AX = a 22 > 0
2
X = (1,1) T ⇒ X T AX = a 11 + 2a 22 + a 22 > 0

X = (1,-1) T ⇒ X T AX = a 11 − 2a 22 + a 22 > 0

27

28

14

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

29

30

15

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

H=

31

32

16

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

33

34

17

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

35

36

18

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

37

38

19

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

λ
-2-λ λ2 - 4λ
λ - 16 = 0
λ
6-λ

39

40

20

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

41

42

21

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

43

44

22

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

45

46

23

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

47

48

24

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

49

50

25

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

51

52

26

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

53

54

27

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

55

56

28

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

57

58

29

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

59

60

30

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

61

62

31

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

63

64

32

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

65

66

33

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

67

68

34

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

69

70

35

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

71

72

36

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

73

74

37

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

75

76

38

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

77

78

39

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

79

80

40

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

81

82

41

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

83

84

42

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

85

86

43

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

87

88

44

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

89

90

45

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

91

92

46

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

93

94

47

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

95

96

48

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

97

98

49

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

99

100

50

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

101

102

51

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

103

104

52

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

105

106

53

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

107

108

54

CuuDuongThanCong.com https://fb.com/tailieudientucntt
2/16/2014

109

110

55

CuuDuongThanCong.com https://fb.com/tailieudientucntt

You might also like