B.tsoniff Analytical Approach To Calculation of Eigenvectors and Eigenvalues For 5x5 Impedance Matrix

Download as docx, pdf, or txt
Download as docx, pdf, or txt
You are on page 1of 23

111Equation Chapter 1 Section 1

Dynamic evaluation of eigenvalues depending on the


parameter
Analytical Approach to Calculation of Eigenvectors
and
Eigenvalues for 5x5 Impedance Matrix
Mathematical background.
(by B.Tsoniff)

Introduction
The three phase line with two grounded shield wires algebraically described as

VA Z AA I A Z AB I B Z AC I C Z AN1 I N1 Z AN 2 I N 2
VB Z BA I A Z BB I B Z BC I C Z BN1 I N1 Z BN 2 I N 2
VC Z CA I A Z CB I B Z CC I C Z CN1 I N1 Z CN 2 I N 2
0 Z N1 A I A Z N1B I B Z N1C I C Z N1N1 I N1 Z N1N 2 I N 2
0 Z N 2 A I A Z N 2 B I B Z N 2C I C Z N 2 N1 I N1 Z N 2 N 2 I N 2
212\*
MERGEFORMAT (.)
or, in matrix form

Z AA
VA
Z
VB BA
VC Z CA

0 Z N1 A
0

Z N2 A

Z AB Z AC Z AN1 Z AN 2

IA


Z BB Z BC Z BN1 Z BN 2

Z CB Z CC Z CN1 Z CN 2

Z N1B Z N1C Z N1N1 Z N1N 2


Z N 2 B Z N 2C Z N2 N1 Z N2 N 2

IB
IC

I N1

I N 2

313\*

MERGEFORMAT (.)

In the impedance matrix

Z AA Z AB Z AC Z AN1 Z AN2

Z BA Z BB

Z CA Z CB
Z
Z
N1 A N1B
ZN A ZN B

2
2

Z BC Z BN1 Z BN2

Z CC Z CN1 Z CN2
Z N1C Z N1N1 Z N1N 2

Z N 2C Z N 2 N1 Z N 2 N 2

414\*

MERGEFORMAT (.)

the diagonal elements ( self-impedance terms) in according to Carson


correction terms formulae, depend on the finite Earth resistance R. In

general, it may be said that matrix and hence its eigenvalues


parametrically depend on the Earth resistance. Dynamic evaluation of
eigenvalues depending on the parameter variation is a fundamental
problem. Meanwhile, dependence of the eigenvalues on the parameters
variation cannot be obtained by numerical methods, which are generally
not effective for problems with the parameters. In such circumstances,
the possibility of an analytical representation of the eigenvalues may be a
decisive factor in the successful solution of the problem.

Reduction of matrix dimension (theory)

In the set of equations (1.1) the last two equations are homogeneous
(free terms equal zero). In this particular case the number of equations
may be reduced by the number of homogeneous equations.
The matrix equation (1.2) may be represented in the block form:

Z AA
VA
Z
VB BA
VC Z CA

L L L
0 Z
N1 A
Z
0
N2 A

Z AB Z AC MZ AN1 Z AN 2

Z BB Z BC MZ BN1 Z BN 2
Z CB Z CC MZ CN1 Z CN 2

L L L ML L L

Z N1B Z N1C MZ N1N1 Z N1N 2


Z N2 B Z N 2C MZ N 2 N1 Z N 2 N2

IA

IB
IC

K
IN
1
IN
2

515\*

MERGEFORMAT (.)
Introducing

Vabc

VA

VB
V
C

0
0
0

Z AA Z AB Z AC
A Z BA Z BB Z BC

CA Z CB Z CC

Z N1 A Z N1B Z N1C

Z N A Z N B Z N C

2
2
2

I abc

I A

I B
I
C

I N1
Ig
IN
2
Z AN1 Z AN2

B Z BN1 Z BN2

Z
Z
CN1 CN 2

Z N1N1 Z N1N2

ZN N ZN N

2 1
2 2

write down block matrix in the form

Vabc
A B I abc

C D I
0

616\* MERGEFORMAT (.)

After matrix multiplication

Vabc A I abc B I g
0 C I abc D I g

717\* MERGEFORMAT (.)


818\* MERGEFORMAT (.)

From (1.7)

I g D 1 C I abc

919\* MERGEFORMAT (.)

Substituting (1.8) into (1.6) yields

Vabc A B D 1
C

Iabc
10110\*

MERGEFORMAT (.)

A B D 1 C

Obviously, matrix
is the (3x3) modified
impedance matrix which comprises the impedances from the last two
rows of the initial matrix (1.3).
Finally

Vabc I abc
R

11111\* MERGEFORMAT (.)

where the superscript (R) denotes the double reduced matrix.


The reduced matrix equation (1.10) from maths point of view is fully
equivalent to the initial equation (1.2). The next step is to get elements of
R

in explicit form.

Kron Reduction (general case)


Consider a matrix equation of the form

Ax b

12112\* MERGEFORMAT (.)

where A is an

n n

complex valued matrix, x and b are vectors in the


Cn
complex vector space
. Assume that the b vector has two zero
elements in the last two rows. Then (1.4) may be written as

a11 a12 .............. a1n



a21 a22 ............. a2n

M M O

an-1,1 an-1,2 ............ an-1,n

an1 an2 .............ann

b
x1
1

x2 b2

M M
xn-1 0

xn 0

13113\*
MERGEFORMAT (.)
The last two equations are constraints by definition like any homogeneous
equations (free terms equal zero). In this case the order of matrix A may
be reduced to

n 2 n 2

by eliminating

xn-1

and

xn

using the

constraint equations

an-1,1 x1+ an-1,2 x2 .......an-1,n xn =0

14114\*

MERGEFORMAT (.)
and

an,1 x1+ an,2 x2 .......an,n xn =0

15115\*

MERGEFORMAT (.)
Reducing the order of the matrix will be given in two steps: in the first

xn

xn-1

step we eliminate
, then
. Generally speaking, the elements of
the vector x can be eliminated in any order, since the exchange of two
rows does not change a matrix.

From (1.6)

xn 1

n1

an, j x j
an, n
j 1

For the

i th

-row (

in

16116\* MERGEFORMAT (.)

)
n1

ai, j x j ai, n xn bi

j 1

17117\* MERGEFORMAT

(.)

n1 a a
i, n n, j
ai, j x j
an,n
j 1
j 1

n1

x j bi
18118\*

MERGEFORMAT (.)

n1

j 1

ai, n an, j
x b
ai, j
an,n j i
19119\* MERGEFORMAT

(.)
Denote

a a

ci, j ai, j i,an n, j


n, n
ci, j

20120\* MERGEFORMAT (.)

in

j n

From (1.11) it follows that coefficients


turns zero for
or for
. Thus, the last row and the last column removed from the matrix A and
equation (1.4) may be written in the form

Cx b

21121\* MERGEFORMAT (.)

where

n1

ci, j x j bi
j 1

22122\* MERGEFORMAT (.)

The expression (1.12) may be written in the structured form as a


determinant of the second order normalized by the self-impedance of the
eliminated ground line.

ci, j

ai, j an,n ai, n an, j 1

an, n
an, n a

ai , j ai , n
n, j an, n

23123\*
MERGEFORMAT (.)
Expression (1.15) provides with the algorithm of the matrix reduction. An
element of the reduced matrix is expressed through the four elements of
the original matrix, brought together into a normalized determinant of the
second order. Each of these four elements is, in turn, the normalized
determinant of the second order (except the first reduction) as a result of
a previous reduction.

xn-1

To eliminate
the above procedure should be repeated using
constraint equation (1.4). As a result we will get the matrix

Dx b

24124\* MERGEFORMAT (.)

where

n 2

bi di, j x j
j 1

di , j

25125\* MERGEFORMAT (.)

ci , j ci, n1
cn1, n1 cn1, j cn1, n1
1

26126\*

MERGEFORMAT (.)
Now we have to express the elements of matrix D through the elements
of matrix A.

ci, j

ai , j ai , n
an, n an, j an,n
1

27127\* MERGEFORMAT (.)

ci,n1

cn1, j

ai ,n 1 ai , n
an, n an,n1 an,n
1

28128\* MERGEFORMAT (.)

an1, j an1, n
an, n an, j an, n
1

29129\* MERGEFORMAT
(.)

cn1,n1

an1, n1 an1, n
an, n an,n1 an, n
1

30130\* MERGEFORMAT

(.)
Substituting (1.19) (1.22) to (1.18) we obtain

di , j

an , n
an 1, n 1 an 1, n
an , n 1 an, n

1 ai , j ai , n
an, n an, j an, n

1 ai ,n1 ai, n
an, n an,n1 an, n

1 an1, j an1, n 1 an1, n1 an 1, n


an, n an, j an, n
an, n an, n1 an, n
31131\* MERGEFORMAT (.)

After simplification

di , j

ai , j ai, n
an, j an, n

ai ,n1 ai , n
an,n1 an, n

an 1, n 1 an1, n

an1, j an 1, n

an, n 1 an, n

an, j an, n

an1, n1 an1, n
an, n1 an, n

1
an, n

32132\* MERGEFORMAT (.)


The last expression is an embedded determinant. The factor in front of
the determinant is a constant, which depends only on the self and mutual
impedances of the eliminated conductors. The first term of the
determinant depends on the row and column numbers, the second one
only on the row number, the third one only on the column number, and
the fourth one is a constant.

Double Kron Reduction of impedance matrix


Apply the obtained formula for the equation (1.2).
Introduce the normalized coefficient

N Z N2 , N2

Z N1 , N1 Z N1 , N 2
Z N 2 , N1 Z N 2 , N 2

33133\*

MERGEFORMAT (.)

Now the impedance matrix

may be written in the Kron reduced form:

Zi, j
R

Zi , j

Zi , N2

Z i , N1

Zi , N2

Z N 2 , j Z N2 , N2 Z N2 , N1 Z N2 , N2
1

N Z N1 , j Z N1 , N2

Z N2 , j Z N2 , N2

Z N2 , N 2
34134\*

MERGEFORMAT (.)

where the superscript (R) denotes the double reduced matrix,

i , j A, B, C

.
After double Kron reduction the original set of five linear equations (1.1)
has been reduced to the set of three linear equations



VA Z AA
I A Z AB
I B Z AC
IC
R




VB Z BA
I A Z BB
I B Z BC
IC
R




VC Z CA
I A Z CB
I B Z CC
IC
R

MERGEFORMAT (.)

35135\*

with corresponding impedance matrix


R
R
R
Z AA
Z AB
Z AC

Z BA Z BB
R R
Z CA Z CB

Z BC
R
ZCC

36136\* MERGEFORMAT

(.)
For convenience we drop the superscript (R) in (1.28)

Z AA Z AB Z AC

Z BA Z BB Z BC
Z

CA Z CB Z CC

37137\* MERGEFORMAT

(.)
but will keep in mind that elements of
according to (1.26).

should be calculated in

Similar matrices
The matrix equation

y Ax

where A is a

n n

complex valued matrix, x


Cn
and y are vectors in the same complex vector space
may be
considered from two points of view: algebraic and geometric.
From algebraic point of view this is the compact notation for a set of
linear equations.
From geometric point of view this is changing the length of the vector x
and its rotation by a certain angle (more scientifically, mapping ndimensional space onto itself).
Any vector space has a basis (coordinate system) or a set of linearly
independent vectors and all bases of a vector space have the same
number of elements, called the dimension of the vector space. Every
element of the vector space can be expressed uniquely as a finite linear
combination of basis vectors. In different bases, the elements of vectors x
and y should have different numerical values (provided the lengths and

directions remain unchanged). Correspondingly, the entries of matrix A


also depend on the selected basis.
If we consider the vectors x and y without reference to a specific basis,
the connection between them is the transformation (mapping) described
by a linear operator

y A x

38138\* MERGEFORMAT (.)

which is represented by different matrices of the same dimension in


different bases. These matrices are called similar and have a number of
specific properties.

y A x

Let mapping
matrix A so that

y Ax

in an arbitrary chosen basis in

Cn

is represented by

, and represented by matrix A in any other arbitrary chosen basis

y ' A' x '

in the same space, so that


where y and x are vectors in the
new basis. Connection between the vectors x and x, y and y is defined
by the transition matrix T from one basis to another

x' T x

y' T y

39139\* MERGEFORMAT (.)


40140\* MERGEFORMAT (.)

where T is any non-singular matrix.


As already mentioned, the matrices A and A are similar. The relation
between them can be easily determined. From

y ' A' x '


T y A' T x
or

y T 1 A' T x

Therefore similar matrices A and A related to each other through the


transformation

A T 1 A ' T

41141\* MERGEFORMAT (.)

The similarity transformation guides matrix A from the basis to the basis
in the same vector space.

Eigenvalue decomposition

As mentioned above, the impact of operator


on the vector x in the
general case is made of two independent actions: the changing the
vectors length and rotation of the vector by a certain angle. If angle of
rotation is zero, then operator
words

only change the vectors length, other

A x x

42142\* MERGEFORMAT (.)

In this case the vector x is called an eigenvector of the operator


and
is called an eigenvalue of the operator corresponding to the given
eigenvector.
From geometric point of view, an eigenvector x prescribes the direction
where the operators action is reduced to vectors expansion. Then the
coefficient of expansion is an eigenvalue .
If is complex then the geometric interpretation becomes pointless but
definitions remain in force. In this case only more general algebraic
interpretation exists.
Assume that for the certain basis equation (1.41) may be written in the
matrix form

Ax x

43143\* MERGEFORMAT (.)

Then the vector x is called an eigenvector of the matrix A and is called


an eigenvalue of the matrix corresponding to the given eigenvector.
In explicit form

a11 x1 + a12 x2 .......a ,n xn = x


a21 x1 + a22 x2 .......a ,n xn = x
1

.......................................................................
.......................................................................

an1 x1 + an2 x2 .......an,n xn = xn

44144\*

MERGEFORMAT (.)
One can see now that there is a product of two unknown in the right hand

xi

side of each equation: the eigenvalue and the component


of the
eigenvector x, that is the problem of finding eigenvalues and
eigenvectors of the matrix A is nonlinear. In addition, the number of
unknowns exceeds the number of equations by one.

xi

There is a method of separating and , based on the Cramer rule for


solving the set of simultaneous linear equations.
Collecting terms in the right hand side brings (1.43) to the set of linear
homogeneous equations

a x1 + a x2 .......a ,n xn = 0
a x1 + a x2 .......a ,n xn = 0
11

21

12

22

.......................................................................
.......................................................................

an1 x1 + an2 x2 ....... an,n xn = 0


MERGEFORMAT (.)

In matrix form

45145\*

a11 a12 ........... a1n


a21
a22 ....... a2n

M
an1

0
x1

x2 0

M M
0
xn




M O M

an2 ....... an,n

46146\*

MERGEFORMAT (.)

or

A I x 0

47147\* MERGEFORMAT (.)

n n

where I is the
unity matrix.
In according to the Cramer rule, (1.46) has a trivial solution

x1 x2 ....xn 0

A I

if the determinant of the matrix


is non-zero, and infinite
number of non-trivial solutions if the determinant is zero. The latter
condition

a11

a12 ............. a1n

a21

a22

an1

....... a2n
O

an2 ...... an,n


48148\*

MERGEFORMAT (.)

gives the equation for calculating the eigenvalues .


In according to algebra of determinants, (1.47) is equivalent to the
algebraic equation of power n

Pn ( ) 0

49149\* MERGEFORMAT (.)

which always has n complex different or multiple roots.


If all are different, then each

corresponds (with accuracy up to a

constant factor) to the eigenvector

fj

j 1, ... n

. Those n vectors are

linearly independent and form the basis in which the matrix


takes the diagonal form. It is easy to prove that

1
0
M
0

0 ...........0

2 ....... 0

M O M
0 ....... n

A T 1 A' T

50150\* MERGEFORMAT

(.)
if the transition matrix T made of eigenvector columns.
Vectors with the physical dimension and dimension
transformation
In equation

Vabc I abc

vectors

Vabc

and

I abc

51151\* MERGEFORMAT (.)

belong to the vector spaces of different physical

dimensions, which means that matrix


of currents to the space of voltages.

transforms vector from the space

A T 1 A' T

However, the similarity transformation


is defined only within
the same linear space and cannot transfer a vector from one space to
another. Since abstract algebra deals with vectors which do not have
physical dimensions it is not clear how the different physical dimensions
in the left and right hand sides of equation (1.42) may influence the
solution of eigenvalue equation. However, the general experience has
shown that neglecting some subtleties in mathematical reasoning while
solving a physical problem, may lead to fatal errors in the final result.

To comply with condition that vectors


the same physical dimension, matrix

Vabc

and

I abc

belong to the space of

must be made dimensionless.

Dimensionless of matrix
may be easily achieved by normalizing all the
elements to a certain number (numbers), which has the dimension of
impedance, the choice of which is more or less arbitrary. It is desirable,
however, to choose the parameters that have a physical meaning. Let us
write the equation (1.42) in explicit form

VA Z AA I A Z AB I B Z AC I C
VB Z BA I A Z BB I B Z BC I C
VC Z CA I A Z CB I B Z CC I C

52152\*

MERGEFORMAT (.)
It makes sense to divide each equation by the corresponding diagonal

entry of matrix , which has clear physical interpretation ( selfimpedance) of the line.

VA

Z AA

IA

IB

Z BC
IC
Z BB

VC
Z
Z
CA I A CB I B
Z CC Z CC
ZCC

IC

VB
Z
BA I A
Z BB
Z BB

MERGEFORMAT (.)
In matrix form

Z AB
Z
I B AC I C
Z AA
Z AA

53153\*

VA

Z
AA

VB

Z
BB

VC

Z
CC

Z AB
Z AA

Z BA
Z BB

Z CA
Z CC

Z CB
Z CC

Z AC

Z AA
Z BC

Z BB

I A
I B
I C

54154\*
MERGEFORMAT (.)
or

VA '
'
VB
V '
C

'

Z AC

'
Z AB

'
Z BA
Z'
CA

1
'
Z CB

I A

'
Z BC
I B
1 I C
55155\*

MERGEFORMAT (.)
where the normalized elements are marked by prime.

Now the matrix

'
' Z BA
Z'
CA

'
Z AB

1
'
Z CB

'

Z AC

'
Z BC

56156\*

MERGEFORMAT (.)

is dimensionless and vectors


space.

'
Vabc

and

I abc

belong to the same vector

The eigenvalue equation may be written as

'
Z AB

'
Z AC

'
Z BA

'
Z BC

'
Z CA

'
Z CB

1
57157\*

MERGEFORMAT (.)
Minor expansion along the first row yields

1
'
Z CB

'
Z BC

'
'
1 Z BC
Z CB
1
3

58158\*

MERGEFORMAT (.)

'
AB

'
Z BA

'
Z BC

'
Z CA

'
'
'
'
'
Z AB
Z BA
Z BC
Z CA
1 Z AB

59159\*

MERGEFORMAT (.)

'
AC

'
Z BA

'
Z CA

'
Z CB

'
'
'
'
'
Z AC
Z BA
Z CB
Z AC
Z CA
1

60160\*

MERGEFORMAT (.)

1 Z BC Z CB Z AB Z BA Z AC Z CA Z AB Z BC Z CA Z AC Z CB Z BA 0
'

'

'

'

'

'

'

'

'

'

MERGEFORMAT (.)

Let

X 1
'
'
'
'
'
'
p Z BC
Z CB
Z AB
Z BA
Z AC
Z CA

'
'
'
'
'
'
q Z AB
Z BC
Z CA
Z AC
ZCB
Z BA

'

'

61161\*

Now equation (1.47) may be presented in the form of the monic


depressed cubic

X 3 pX q 0

62162\* MERGEFORMAT (.)

The polynomial of degree n is called monic if its first coefficient is 1.


The polynomial of degree n is called depressed if its second coefficient is 0.

Considering that the solution of a cubic equation always starts with


bringing a cubic polynomial to the monic depressed form we have to
admit that well-chosen normalization allowed us to avoid tedious
algebraic transformations which should bring us to the same cubic
equation.

Analytical solution of cubic equation with complex


coefficients
1.

Fundamental theorem of algebra

The fundamental theorem of algebra states that every non-constant


single-variable polynomial with complex coefficients has at least one
complex root.
Corollary: every single-variable, degree n polynomial with complex
coefficients

P x an x n an 1 x n 1 ... a1 x a0
has, counted with multiplicity, exactly n roots
2.

x1 , x2 ,... xn

Vieta's formulae

Vieta's formulas relate the polynomial's coefficients { ak } to signed sums


and products of its roots { xi } .
For a cubic polynomial

P x a3 x 3 a2 x 2 ... a1 x a0
this relation looks as

x1 x2 x
3

a2
a3

x1 x2 x1 x x2 x
3

x1 x2 x
3

3.

a1
a3

a0
a3

63163\* MERGEFORMAT (.)

Depressed cubic polynomial

Analytical solution of cubic equation

a3 x 3 a2 x 2 ... a1 x a0 0

64164\*

MERGEFORMAT (.)
always starts with reducing cubic polynomial to the monic form and the
Tschirnhaus transformation which removes quadratic term from the
polynomial, bringing it to the depressed one

X 3 pX q 0

65165\* MERGEFORMAT (.)

In this case the Vieta's formulae are

x1 x2 x 0

x1 x2 x1 x x2 x p
3

x x x q
1 2
3

66166\* MERGEFORMAT

(.)

4.

Cubic roots of unity

Following the Fundamental theorem of algebra the equation

3 1 0
should have three complex roots.

67167\* MERGEFORMAT (.)

In trigonometric notation

1 cos 2 k i sin 2 k ei 2 k

68168\*

MERGEFORMAT (.)
3

1e

2 k
3

k 0; 1; 2

69169\* MERGEFORMAT

(.)
Correspondingly

0 e0 1
1 e

2 e

2
3

4
3

2
2 1 i 3
i sin

3
3
2
4
4
1 i 3
cos
i sin

3
3
2

cos

70170\*

MERGEFORMAT (.)
From this result immediately follows

2 12
71171\* MERGEFORMAT (.)
and

12 1 1 0
72172\* MERGEFORMAT (.)

5.

Solution of Cubic equation


(Method of D.Kalman and J.White)

Let start with not obvious but easily verifiable by inspection, identity

x b c x b c x b c x3 3xbc b3 c3
73173\* MERGEFORMAT (.)

Here b and c are arbitrary complex constants,

1
is

from (1.68).

If the following relations hold

3bc p
b3 c 3 q

74174\* MERGEFORMAT (.)


75175\* MERGEFORMAT (.)

then (1.64) may be factorized to linear factors

X 3 pX q x b c x b c x b c

76176\*

MERGEFORMAT (.)
which means that

x b c x b c x b c 0

77177\*

MERGEFORMAT (.)
The roots of the last equation obviously are:

x b c /

x b c

x b c

where b and c should be expressed in terms of p and q.


After cube (1.73), equations (1.73) and (1.74) may be rewritten in the
form

b3c 3 p 3 / 27
b3 c 3 q

It is apparent that

b3

and

c3

78178\* MERGEFORMAT (.)


79179\* MERGEFORMAT (.)

are the roots of the quadratic equation

p3
x qx
0
27
2

That leads to

80180\* MERGEFORMAT (.)

q q 2 4 p 3 / 27
b

q q 2 4 p 3 / 27
c

1
3

1
3

81181\*

MERGEFORMAT (.)
There are three complex cube roots among which to choose b and c, and
not all sets satisfy the original equation (1.73). To resolve this situation
one should choose any of three complex cube roots for b, and then define
c as

p
3 b

82182\* MERGEFORMAT (.)

You might also like