The basic motivation for state estimation is that

we want to perform computer analysis of the

network under the conditions characterized by
the current set of measurements.

Specifically, we want to know the values of the

bus voltage phasor magnitudes and angles |Vk|,
k
1=0
so we do not need to find that one).
We begin with linear least squares estimation.

2.0 Linear least squares estimation

The material in this section closely follows that
in [2, ch2]. Also, see 12.1, Ch12 app of W&W.

Consider the DC circuit given in Fig. 1 below

where current injections I1, I2, and voltage e are
unknown. Let R1=R2=R3=1.0
measurements are as follows:
meter A1: i1,2=1.0 Ampere
meter A2: i3,1=-3.2 Ampere
meter A3: i2,3=0.8 Ampere
meter V: e=1.1 volt

2
The problem is to determine the state of the
circuit, which in this case is nodal voltages v1,
v2, and the voltage e across the voltage source.
+ V -
R1
Node 1 Node 2
A1
+ -
e A3

I1 I2
R2 R3
A2

Node 3

Fig. 1
s write each one of the measured currents in
terms of the node voltages, and we may also
write down our one voltage measurement.
m v1 v2 e
i1, 2 v1 v2 e 1.0 (1)
1
m 0 v1
i3,1 v1 3.2 (2)
1
m v2 0
i2,3 v2 0.8 (3)
1
e 1.1 (4)
Expressing all of the above in matrix form:

3
 1 1 1 1.0
v1
1 0 0 3.2
v2
0 1 0 0.8
e (5)
0 0 1 1.1
(5) as A, x, and b, so:
Ax b (6)

How do we solve eq. (5)?

Observe that the multiplying matrix is not

square, i.e., there are 4 rows but only 3 columns.
The reason for this is because there are 4
equations but only 3 variables. This means that
the system of equations defined by eq. (5) is
over-determined. This is a standard feature in
state-estimation. Noticing that there is one
equation for each measurement, the implication
is that we will always attempt to obtain as many
measurements as we can.

4
There is no single solution to eq. (5), but there is
a single solution that is normally thought of as

solution which minimizes the sum of the

should be
computed by each equation, which is:
1.0
3.2
b
0.8 (7)
1.1
and what is computed by each equation, which
is:
1 1 1
v1
1 0 0
Ax v2
0 1 0 (8)
e
0 0 1
The difference, or error, is then:
1.0 1 1 1 1.0 v1 v2 e
v1
3.2 1 0 0 3.2 v1
b Ax v2
0.8 0 1 0
e
0.8 v2 (9)
1.1 0 0 1 1.1 e
The squared error is then

5
 2
1.0 v1 v2 e
2
3.2 v1
2
0.8 v2
2
1.1 e
and the sum of the squared errors is:
2 2 2 2
1.0 v1 v2 e 3.2 v1 0.8 v2 1.1 e
Careful tracking of the previous expression will
indicate that it could be written as
T
b Ax b Ax

name:
1 T
J
2
b Ax b Ax (10)
Our problem is then to choose x so as to
minimize J. Under requirements on the form of
J (convexity), we minimize it by setting its
gradient with respect to x to 0; then solve for x.
To do this, we can expand J as follows:
1 T 1 T T
J b Ax b Ax b Ax b Ax
2 2
1 T T T T (11)
b b Ax b b Ax Ax Ax
2
Using (Ax)T=xTAT, we have:
1 T
J b b xT AT b bT Ax xT AT Ax (12a)
2

6
Consider the second and third terms in (12a).
Using a 2x2 to illustrate,
A11 A21 b1
xT AT b x1 x2 b1 x1 A11 b1 x2 A12 b2 x1 A21 b2 x2 A22
A12 A22 b2

A11 A12 x1
bT A x b1 b2 b1 x1 A11 b2 x1 A21 b1 x2 A12 b2 x2 A22
A21 A22 x2
we observe these terms are equal, which we can
prove by using (Ax)T=xTAT to show that
[xTATb]T=bTAx and recognizing that these terms
are scalar in (12a). Therefore, (12a) becomes
1 T
J b b 2bT Ax xT AT Ax(12b)
2
which is identical to (12A.12), p. 504 of W&W.
To remind us about gradients, we recall that it is
given by (13):
J
x1
J
xJ x2
(13)
J
xn
Now (12) is written in compact notation, and it
may not be obvious how to differentiate each

7
term in it. To assist with this, your text, in the
Appendix to Chapter 12, develops the following
relations, which are summarized in (12A.25, pg.
506), and repeated here for convenience:
Function Gradient
#1 F xT b xF b
#2 F bT x xF b
#3 F xT Au xF Au
#4 F uT A x xF AT u
#5 F xT Ax xF 2 Ax
Relation #5 applies only if A is symmetric.

The above gradient relations apply to (12b) as

illustrated in (12c).
#4 #5
1 T
J b b 2bT A x xT AT A x
2 (12c)
Using the appropriate relations in the above
table (#4 to the second term and #5 to the third
term), the gradient of (12c) can be expressed as:
1
x J 2 AT b 2( AT A)T x (14)
2

8
(Applying #5 to the third term is acceptable
since ATA must be symmetric.)

Again, using (Ax)T=xTAT, the 2nd term in (14) is

2( AT A)T x 2 AT Ax
so (14) becomes
1
xJ 2 AT b 2 AT Ax AT b AT Ax (15)
2
The minimum of J is obtained when
J
AT b AT Ax 0 (16)
x
And this implies that
T
A b
T
A Ax (17)
Note:
Equation (17) is referred to in statistics as the
normal equations.
We could have obtained (17) by just
multiplying Ax=b through by AT.
ATA multiplies an m×n by an n×m to get an
m×m matrix Square!
(ATA)T=ATA, so the transpose of ATA is itself
(which means ATA is symmetric).

9
 Reference [3, p. 157] shows that if A has
linearly independent columns, then ATA is
invertible.

Solving eq. (17) for x results in

1
x
T
A A A b
T
(18)
Define the gain matrix G as
G A A
T
(19)
Also define the pseudo-inverse of A as
1
A
I T
A A A
T 1
G A
T
(20)

Now we can find the answer to our problem as.

1 T
T 1 T
x A A A b G A b Ab
I
(21)
First, the gain matrix is given as
1 1 1
1 1 0 0 2 1 1
T 1 0 0
G A A 1 0 1 0 1 2 1
0 1 0
1 0 0 1 1 1 2
0 0 1
The inverse of the gain matrix is then found
from Matlab as
3 1 1
1 1
G 1 3 1
4
1 1 3

10
The pseudo-inverse is then
3 1 1 1 1 0 0 1 3 1 1
I 1 T 1 1
A G A 1 3 1 1 0 1 0 1 1 3 1
4 4
1 1 3 1 0 0 1 1 1 1 3

We can obtain the least squares estimate of the 3

states from the 4 measurements as
1.0
1 3 1 1 3.125
I 1 3.2
x Ab 1 1 3 1 0.875
4 0.8
1 1 1 3 1.175
1.1
It is interesting to look at the difference between the
measurements we actually had, which is b, and the values
corresponding to those measurements that we would
compute using the state vector x, which is Ax.

This difference is referred to as the residual, r, and is

given by eq. (9) as
1.0 1 1 1
v1
3.2 1 0 0
r b Ax v2
0.8 0 1 0
e
1.1 0 0 1
1.0 1 1 1 1.0 1.075 0.075
3.125
3.2 1 0 0 3.2 3.125 0.075
0.875
0.8 0 1 0 0.8 0.875 0.075
1.175
1.1 0 0 1 1.1 1.175 0.075

11
3.0 A motivating system and some basics

Consider Fig. 2 where we obtain measurements

on the indicated quantities.
P12, Q12
V1 Bus 2

Bus 1
P23, Q23

Bus 3

Fig. 2
We denote the measured quantities as follows:
z1 V1
z2 P12
z3 Q12
z4 P23 (22)
z5 Q23
Then we can write
zi zi i (23)
where

12
 zi is the measured value
zi is the true value (unknown)
i is the error (unknown)
Not knowing zi and i is a problem. However,
we may obtain statistical information from
calibration curves (error as a function of
measurement) of measuring instruments. It is
usually assumed that i is a random variable
with a normal (Gaussian) distribution having
zero mean, as illustrated in Fig. 3.
i)

i
0

Fig. 3
Calibration curves for measuring instruments
enable determination of the variance i.

13
Recall the expectation operator, which we
denote as ). It
is defined as:
E ( x) xf ( x)dx x (24)
which gives x, the mean value of a variable x
described by the probability distribution
function f(x).

We also define variance as:

2 2
x var( x) x x f ( x)dx (25)
We relate variance to mean beginning with (25).
2 2
x var( x) x x f ( x)dx

x2 2x x ( x)
2
f ( x)dx

x 2 f ( x)dx 2 x xf ( x)dx ( x)
2
f ( x)dx
(26)
2 2 2
E ( x ) 2( x) ( x)

E( x2 ) ( x)
2

14
From eq. (26) (which is true for any random
variable x), we see that if the mean is 0
(E(x)=0), then the last term in eq. (26) is 0 and
2 2
x E ( x ) (27)
In regards to the calibration error, characterized
by the random variable i, we have then:
E ( i ) 0 (zero mean)
E ( i2 ) 2
i (variance)
Note that the larger the variance, the less
accurate is the measuring device.

Since we have multiple measuring instruments,

we also need to understand how statistics of one
random variable relate to statistics of another.

The covariance measure is effective in doing

this, and is defined as
2
xy cov( x, y) x E ( x) y E ( y ) f ( x, y )dx (28)
Note that variance is a special case of
covariance when x=y, i.e., x2 =cov(x,x).

15
It can be shown [4] from eq. (26) that
2
xy cov( x, y) E ( xy ) E ( x) E ( y) (29)
If x and y are independent, then
E(xy)=E(x)E(y).
Therefore, for two independent random
variables, the covariance is:
2
xy cov( x, y) 0
Two variables having 0 covariance are
uncorrelated, i.e., variation in one gives no
information about variation in the other.

A basic assumption:
The errors i and j for any two measuring
instruments i and j are independent. This means:
0, i j
cov( i , j ) 2 (30)
i , i j
Thus, we can define a covariance matrix R,
where the element in position (i,j) is cov( i j).
Given (30), the matrix will appear as:

16
 2
1 0 0
2
0 2 0 0
R
0 0 (31)
2
0 0 0 m

where it is assumed that we have m measuring

instruments. We use (31) in our development.

4.0 Problem for AC State Estimator

Define the state vector for an N-bus network as:
2

x1 N

x
x2 N 1 V1
(32)

VN

We have n=2N-1 states (N-1 angles, N

voltages).

For each parameter for which we have a

measurement, we want to write an equation in
terms of the states. In other words, if we have a
measurement zi, and if we know the correct state
x, then the true value of the measurement is

17
 zi hi (x) (33)
For a voltage measurement, the function hi is
very simple:
zi Vk (34)
where measurement i occurs at bus k.

For MW and MVAR flows, the function hi is

given by the expressions for power flow across
a line from bus p to bus q. These are given by
Ppq Vp2 g pq VpVq g pq cos( p q ) VpVqbpq sin( p q ) (35)
Qpq Vp2 (bpq bp ) VpVq g pq sin( p q ) VpVqbpq cos( p q ) (36)
where the line has
series admittance of gpq+jbpq; gpq>0, bpq<0 for
inductive line.
Shunt susceptance at bus p of bp (which
includes any reactive shunt at the bus plus half
of the line charging). If capacitive, then bp>0.

Now define some vectors:

z1
z
Measured values: (37)
zm

18
 z1
z
True values: (38)
zm
1

Errors (39)
m

Generalizing, i.e., vectorizing (23), we have:

z z (40)
From (33), we also have for a vector of
functions expressing the measurement values in
terms of the states:
z h(x) (41)
Substituting eq. (41) into (40), we have:
z h(x) (42)
Now consider what we have here. The number
of unknowns is n=2N-1 (the states in x which
are the angle and voltage variables), and then
we have some number of measurements m.

that m>n, i.e., that we have more

measurements than states.
19
One thing we could do is
set =0
choose m=n equations (each corresponding to
a measurement)
Solve for x (it would need to be non-linear
solver but once done, solution is unique).
However, the tough question would be: Which
measurements to choose to keep? Which are the
best?

Since we do not know which measurements are

the best, we instead make sure that we have
more measurements than states, i.e., we will
solve the problem for m>n=2N-1.

So our strategy is as we saw in our earlier

example, to choose x so as to minimize the sum
of the squared errors between the measured
values and the actual values.

From eq. (42), we have that the error is

z h(x) (43)

20
Recall equation (10) which expressed our sum
of squared errors as
1 T
J
2
b Ax (10)
b Ax

Similarly we express the sum of squared errors as

m
1 2 1 T 1 T
J
2 i 1
i
2 2
z h( x) z h( x) (44)
We denote the above as because we will find
it convenient to modify a bit.

By minimizing , we are effectively choosing x

however, that some measurement devices are

better than others (which is a different statement
than some measurements are better than others).

It is reasonable, then, to place more weight on

the better measuring devices.
1
A good choice for this weight is i
2 since
2
1
Good device small i , large i
2

2
1
Bad device large i , small i
2

21
Therefore, we will modify eq. (44) to be:
m 2
1
(45)
i
J 2
2 i 1 i

Recall the covariance matrix given by eq. (31),

repeated here for convenience:
2
1 0 0
2
0 2 0 0
R
0 0 (31)
2
0 0 0 m

Because R is diagonal, its inverse is easy to find:

1
2
0 0
1
1
1 0 0 0
R 2
2
0 0 (46)
1
0 0 0 2
m

We can therefore express eq. (45) as:

m 2 m 2
1 i 1 T 1 1 T 1 1 zi hi ( x)
J 2
R z h( x ) R z h( x) 2
2 i 1 i 2 2 2 i 1 i

(47)

22
The problem then becomes to find x that
minimizes J. Note, however, that h is nonlinear,
and so our solution will necessarily be iterative.

5.0 Solution for AC State Estimator

So the problem is stated as follows:

mimimize
m 2 m 2
1 i 1 T 1 1 T 1 1 zi hi ( x)
J 2
R z h( x ) R z h( x) 2
2 i 1 i 2 2 2 i 1 i

(48)
We can apply first order conditions, which
means that all first derivatives of the objective
function with respect to decision variables must
be zero, i.e., that x J 0 . That is,
J
x1 0
J
xJ
x J (49)
0
xn
For a single element in x J, we have:
m 2
1 zi hi ( x)
J
2 i 1 i
2 (50)

23
 J 1 m 2 zi hi ( x) hi ( x) m
zi hi ( x) hi ( x)
x1 2i1 i
2
x1 i 1 i
2
x1 (51)
This can be written in matrix form as
z1 h1 ( x)
J h1 ( x) h2 ( x) hm ( x) 1 z2 h2 ( x)
R
x1 x1 x1 x1 (52)
zm hm ( x)
And we then see how to write the vector of
derivatives, according to:
h1 ( x) h2 ( x) hm ( x)
x1 x1 x1 z1 h1 ( x)
h1 ( x) h2 ( x) hm ( x) z2 h2 ( x)
J 1
x2 x2 x2 R
x
(53)
h1 ( x) h2 ( x) hm ( x) zm hm ( x)
xn xn xn
We recognize the matrix of partial derivatives in
(53) as a sort of Jacobian matrix but
(a) it is n×m, i.e., it is not square and
(b) unlike standard Jacobian, here the rows
vary with variable (x1, x2
(h1, h2

24
 s define a matrix H that does not have the
second (b) attribute, i.e.,
h1 ( x) h1 ( x) h1 ( x)
x1 x2 xn
h2 ( x) h2 ( x) h2 ( x)
H x1 x2 xn
(54)
hm ( x) hm ( x) hm ( x)
x1 x2 xn
Note that H is m×n, and it is the transpose of the
first matrix in eq. (53).

Then we see that the optimality condition can be

written as:
J
x1
J 1
xJ H T ( x) R z h( x ) 0
x J (55)
xn
The solution to eq. (55) will yield the estimated
state vector x which minimizes the squared
error.

25
Because there are n elements in the partial-J
vector on the left, we observe that eq. (55) gives
n equations. Since there are n variables in x, it is
possible to solve eq. (55) explicitly for x.

Now we need to determine a solution procedure

for eq. (55 ne the left-hand-
side of eq. (55) as G(x), i.e.,
n m m m m 1
n 1
T 1
G ( x) H ( x) R z h( x ) (56)
0
Since H is m×n, HT is n×m. R-1 is m×m, and
z-h(x) is m×1. Therefore, G(x) is n×1, i.e., a
vector.

Perform a Taylor series expansion of G(x)

around a certain state x0.
G( x 0 x) G ( x 0 ) x G ( x) x x h.o.t. 0 (57)
0

Note that eq. (57) indicates that if x0+ x is to be

a solution, then the right-hand-side of eq. (57)
must be zero.

26
Recall that in a Taylor series expansion, the
higher order terms (h.o.t.) contain products of
x, and so if x is relatively small, terms
x will be very small,
and in fact, negligible. So we will neglect the
h.o.t. in eq. (57). This results in:
G( x 0 x) G ( x 0 ) x G ( x) x x 0 (58)
0

Since eq. (58) is nonlinear, we must resort to an

iterative algorithm to solve it. We will use a
Newton-type algorithm.

t we can make a pretty good

guess at the solution to eq. (58), i.e., that the
difference between our guess and the real
solution is relatively small.
Denote this guess as x(k). Because it is not the
solution, G(x(k)) 0.

So we want a better guess. Denote the better

guess as x(k+1). The difference between the old
guess x(k) and the new guess x(k+1) is x(k+1), i.e.,
( k 1) (k )
x x x (59)

27
Or we can write
( k 1) (k )
x x x (60)
Evaluating G at the better guess, we have
( k 1) (k ) (k )
G( x ) G( x x) G ( x )
(61) x G ( x) x
x(k )

We desire that G(x(k+1))=0. Under this desired

condition, eq. (61) becomes:
( k 1) (k ) (k )
G( x ) G( x x) G ( x ) x G ( x) x x 0 (62) (k )

Solving for x G ( x) x, we have:

x( k )
(k )
x G ( x) (63)
(k )
x G( x )
x

In considering eq. (63), we already understand

the right-hand-side, this is just the negative of
eq. (56), evaluated at x(k), i.e.,
(k ) (k ) 1 (k )
G( x ) H T ( x ) R z h( x ) (64)
There are n functional expressions in (64), one
for each derivative of J wrspt xj per (55).

But what is x G( x) x ? This is the derivatives, (k )

with respect to each of the n state variables, of

each of the n functional expressions in eq. (56):
1
G ( x) H T ( x) R z h( x) 0 (56)

28
Since there are n functional expressions and n
derivatives to take for each one, we can see that
x G ( x) x will be n×n, a square matrix.
(k )

Remembering eq. (55)

J
x1
J 1
xJ H T ( x) R z h( x ) 0
x J (55)
xn
reminds us that G(x) in (56) are also derivatives
with respect to each of the n state variables.
Therefore, x G( x) x are second derivatives of J
(k )

with respect to the state variables.

Can we obtain a form for these second
derivatives? ):
1
G ( x) H T ( x) R z h( x) (56)
So what we want is:
G ( x) 1
x G ( x) H T ( x) R z h( x ) (65)
x x
The differentiation of what is inside the brackets
of eq. (65) is formidable. We will make it easier
for ourselves by assuming that H(x) is a constant
matrix. This implies H(x+ x H(x), i.e., the

29
derivatives of the measurement equations do not
change. Is this a good assumption? Remember,
from eq. (54),
h1 ( x) h1 ( x) h1 ( x)
x1 x2 xn
h2 ( x) h2 ( x) h2 ( x)
h( x)
H x1 x2 xn
x (54)
hm ( x) hm ( x) hm ( x)
x1 x2 xn
where the functions hi are the expressions for
the measurements (voltages, real and reactive
power flows) in terms of the states (angles and
voltage magnitudes).

So H is very similar to a power flow Jacobian

matrix. It is well known that the power flow
Jacobian is relatively insensitive to small
variations in state.

With the above assumption, differentiating the

right-hand-side of eq. (65) becomes not-so-bad:

30
 G ( x) 1
x G ( x) H T ( x) R z h( x )
x x

H T ( x) R
1 h( x )
H T ( x) R
1 h( x ) (66)
x x
h( x)
But we recognize from eq. (54) the term x in
eq. (66) as H. Therefore, eq. (66) becomes:
1
x G ( x) H T ( x) R H ( x)
(67)
Making this substitution into eq. (63) results in:
(k )
x G ( x ) x G ( x(k )
) (63)
x
1 (k )
H T ( x) R H ( x ) x(68) G( x )
x( k )

Finally, replacing the right-hand-side of eq. (68)

with eq. (54) evaluated at x(k) yields:
1 1
H T ( x) R H ( x ) x H T ( x ) R z h( x)
(k ) (69) (k )
x x

Equation (69) provides a way to solve for x.

31
6.0 Solution Algorithm
Given:
measurements z [z1 m]
standard deviations 1 m
the network
T
Compute : state estimate x=[x1 n]
(all voltage magnitudes and all voltage angles
except for swing bus angle)
1. Form measurement expressions h(x)
h( x )
2. Form derivative expressions H x
3. Form R
4. Let k=0. Guess solution x(k).
5. Compute H(x(k)), h(x(k))
1
1
6. Compute A H ( x) R H ( x) x , b
T
H T ( x) R z h( x)
(k )
x( k )

7. Solve A x=b for x.

8. Compute x(k+1)= x(k)+ x
9. If max
i
xi then
k=k+1
Go to 5
Else Stop

32
Example:
Consider the system below. Real power measurements are taken as follows: P12=0.62 pu , P13=0.06 pu,
and P32=0.37 pu .
3 angle is reference. So the state vector is therefore x 1 2]T. Your textbook solves this problem using
DC power flow equations on pp. 467-471, which are reviewed below. Your homework requests that you
repeat the analysis using AC power flow equations.
Bus 1 Bus 2
P12 x12=0.2pu

P12=0.62pu

P13=0.06pu x23=0.25pu

P32=0.37pu Bus 3
X13=0.4pu

a) Determine the vector of measurement expressions h(x), the derivative expressions H h( x) ,

x
and the weighting matrix R.

Recalling that Pjk=Bjk( j - k), and with 3=0, we write

1
h1 ( x) P12 1` 2 5 1` 5 2
0.2
1
h2 ( x) P13 1` 3 2.5 1`
0.4
1
h3 ( x) P32 3` 2 4 2
0.25
b) Compute H(x(0)), h(x(0)) for an estimate of x(0) = 1 2]
T
=[0.024 -0.093]T, (in radians)

Evaluating h1, h2, and h3 given x(0) =[0.024 -0.093]T

h1( x) P12 5 1` 5 2 5(.024) 5( .093) 0.585

h2 ( x) P13 2.5 1` 2.5(.024) 0.06 The example in the text,
h3 ( x) P32 4 2 4( .093) 0.372 p. 469, assumed the
initial estimate to be
T
1 2] =[0 0]T.

33
 h1 ( x) h1 ( x) P12 ( x) P12 ( x)
x1 x2 1 2 5 5
h( x ) h2 ( x) h2 ( x) P13 ( x) P13 ( x)
H 2.5 0
x x1 x2 1 2
h3 ( x) h3 ( x) P32 ( x) P32 ( x) 0 4
x1 x2 1 2
In this particular case, H is a constant matrix. In general, it will be a function of the
states and will thus change from iteration to iteration. It is constant in this case since
we have used linearized (DC) power flow expressions.
1 1
c) Compute A H T ( x) R H ( x) , b H T ( x) R z h( x)
x( 0 ) x( 0 )
First, get R, which is given by
2
1 0 0
2
0.0001 0 0
0 2 0 0
R 0 0.0001 0
0 0
2
0 0 0.0001
0 0 0 m
1
0.0001 0 0 5 5
5 2.5 0
A H T ( x) R 1 H ( x ) 0 0.0001 0 2.5 0
x
( 0)
5 0 4
0 0 0.0001 0 4
10000 0 0 5 5
5 2.5 0 312500 - 250000
0 10000 0 2.5 0
5 0 4 - 250000 410000
0 0 10000 0 4
1
b H T ( x) R z h( x ) ( 0)
x

10000 0 0 0.62 0.585 Because the example in

5 2.5 0 the text, p. 469, assumed
0 10000 0 0.06 0.06
5 0 4 the initial estimate to be
0 0 10000 0.37 0.372 1 2]
T
=[0 0]T, this last
vector [z-h(x)], in the text,
10000 0 0 0.035 was just [z]=[.62 .06 .37]T.
5 2.5 0
0 10000 0 0 This would result in
5 0 4 b=[32500 -45800]T.
0 0 10000 0.002
1750
1670
d) Solve A x=b x.
312500 - 250000 1 1750 1 0.0046
- 250000 410000 2 1670 2 .0013
And the new value of x would then be
T T T
1 2] =[0.024 -0.093] +[.0046 -.0013] =[0.0286 -.0943]
which is the same solution obtained in the text, pg. 469. The solutions are the same, even

34
 though we started from differ i functions are linear (and therefore
the H matrix is constant).

We may now compute the residual from

r z h( x) x
(k )

T
To get this, we need to recompute h(x), using x= 1 2] =[0.0286 -.0943]T as follows:
h1( x) P12 5 1` 5 2 5(.0286) 5( .0943) 0.6145
h2 ( x) P13 2.5 1` 2.5(.0286) 0.0715
h3 ( x) P32 4 2 4( .0943) 0.3772
The measurements are P12=0.62 pu , P13=0.06 pu, and P32=0.37 pu, and so residuals
are given by
0.62 0.6145 0.0055
r 0.06 0.0715 0.0115
0.37 0.3772 .0072
Compare to our previous residual
0.62 0.585 0.035
0.06 0.06 0
0.37 0.372 0.002
and we see that it got better in r1 and worse in r2 and r3. However, the squared error
for this iteration is 0.000214 and that for the previous iteration is 0.0012, and so, in
fact, our solution has improved. We are able to look at just squared error in this example
instead of weighted squared error because all of the weights are the same.

Note that your text, at the bottom of page 469, calculates J. This is the objective function,
the weighted squared error, which is a measure of how good our solution is. They
get 2.14, which can be compared to our above calculations by multiplying by 2=.0001,
which gives 0.000214, the same as what we obtained.

References:
[1
Computer Conference (PICA), Denver, Colorado June, 1969.
[2]
[3] G. Strang,
[4] A. -Hill, New York, 1984.

35