L7 Lse Alg 1

Two Sample Estimation
Result of two sample estimation under ideal condition

Result of estimation in presence of noise
Summary
Least Square Method for Estimation of

Phasors - I
S. A. Soman
Department of Electrical Engineering,

IIT Bombay
S. A. Soman Department of Electrical Engineering, IIT Bombay Least Square Method for Estimation of Phasors - I
Summary
Outline
1 Two Sample Estimation
2 Result of two sample estimation under ideal condition
3 Result of estimation in presence of noise
4 Summary
Summary
Two sample estimation
In a numerical relaying setup,

voltage and current signals would
be sampled at appropriate
frequency and acquired by a
micro processor or a DSP.
For relaying decision making we
need to estimate the voltage and
current phasors.
For simplicity, imagine a single
phase circuit as shown and also
assume that the frequency of the
supply is known.
Summary
Two sample estimation....contd.
The voltage and current waveforms can be represented by

v (t) = Vm sin(ωt + φv );
i(t) = Im sin(ωt + φi ).
Suppose we need to estimate voltage phasor (Vm ∠φv ).
Since there are two unknowns, minimum number of
samples required in a cycle to estimate unknowns is two.
Let us assume that a sample is acquired after every ∆t
second. Let the first sample be obtained at t1 and the next
one at t2 .
Summary
Then, voltage samples at t1 and t2 are given by,

v1 (t1 ) = Vm sin(ωt1 + ωv )
v1 = Vm sin θ1 cos φv + Vm cos θ1 sin φv
v2 = Vm sin θ2 cos φv + Vm cos θ2 sin φv
where, θ1 = ωt1 and θ2 = ωt2 .
Assume that first sample is obtained at t=0 and angle
corresponding to j th sample is given by θj = jω0 ∆t.
Treating Vm cos φv and Vm sin φv as unknowns, we get

v1 sin θ1 cos θ1 Vm cos φv
= (1)
v2 sin θ2 cos θ2 Vm sin φv
Summary

Solution of that equation gives,

Vm cos φv 1 cos θ2 − cos θ1 v1
=
Vm sin φv sin(θ1 − θ2 ) − sin θ2 sin θ1 v2
Hence,
v1 cos θ2 − v2 cos θ1
Vm cos φv = (2)
sin(θ1 − θ2 )
v2 sin θ1 − v1 sin θ2
and Vm sin φv = (3)
sin(θ1 − θ2 )
It is clear from eqns. (2) and (3) that if θ1 − θ2 = ±mπ, m
an integer, then sin(θ1 − θ2 )=0, which implies the
singularity of coefficient matrix.
Summary
Two sample estimation... contd.
In such an event, it is not possible

to proceed with estimation.
For example, with m = 1, we get
ωt2 = ωt1 + π. Corresponding
samples are shown in the figure.
So sampling frequency has to be
higher than twice the supply
frequency which is in the
agreement with the sampling
theorem.
Summary

Consequently Vm and φv can be computed by following
equations.
q
Vm = Vm2 cos2 φv + Vm2 sin2 φv

−1 Vm sin φv
and φv = tan
Vm cos φv
A generic form of the estimation is given by the following
equation.
vk −1 cos θk − vk cos θk −1
Vmk cos φkv = (4)
sin(∆θ)
vk sin θ k −1 − vk −1 sin θk
Vmk sin φkv = (5)
sin(∆θ)
Summary
Example 1
Consider a signal V (t) = 10 sin(2π × 50 × t + 30◦ ), and a

sampling time of 5 msec i.e. sampling rate 4 samples per
cycle. Let the sampling be initiated at t0 = 0 sec.
(a) Write down the sequence of first 10 voltage samples.
Ans : The sequence is summarized below.
tmsec 0 5 10 15 20
V 5.00 8.66 -5.00 -8.66 5.00
tmsec 25 30 35 40 45
V 8.66 -5.00 -8.66 5.00 8.66
Summary
Example 1... contd.
Compute the voltage phasor for

100 samples.
Figure shows the estimate of
voltage magnitude Vm as a
function of samples. It is obvious
that in the absence of noise,
estimate gives the right value of
Vm .
Summary
Example 1... contd.
Figure shown here also

introduces the concept of data
window for estimation. This
window contains the ‘active’ set of
samples which are currently being
processed for phasor estimation.
In the present case, we say that
we are using a 2-sample window.
Each consecutive window, differs
from the previous window by
adding a new sample and by
removing the oldest active
sample.
Summary
Example 2
The noise has a zero mean, and

it’s standard deviation measures
the accuracy of the meter.
Typically, noise is modeled by
zero mean Gaussian distribution.
Consequently, if a 0-100V
voltmeter has a standard
deviation of 1%, then it implies
that a measurement of a signal
having magnitude of 100V will be
measured any where between
97-103V, 99% of time.
Summary
Example 2... contd.
Let us model the voltage samples by
v (k ∆t) = 10 sin(2π × 50 × t + 30◦ ) + E ∗ randn (6)
Where, E models the standard deviation of noise.

Now we conduct an experiment to estimate voltage phasor
in presence of noise for 100 samples using 2-sample
method. The mean and standard deviation of Vm for
different levels of noise is shown in table.
Summary
Effect of Gaussian noise on estimation
Randn Mean Standard

Multiplier(E) Deviation
0.1 10.0069 0.1596
0.4 10.0426 0.6392
0.7 10.1011 1.1182
1.0 10.1824 1.5938
1.3 10.2871 2.0624
1.6 10.4169 2.5186
2.0 10.6346 3.0919
2.3 10.8336 3.4940
2.7 11.1430 4.0080
2.8 11.2277 4.1241
3.0 11.4055 4.3830
Summary
They are calculated as per following equations. Let

X1 , X2 .....XN be N samples under consideration.
Then, mean
N
P
Xi
i=1
M=
N
And standard deviation
v
u N
u (Xi − M)2
uP
t i=1
σ=
N
the square of standard deviation (σ 2 ) is called variance.
Summary
Figure shows the plot of estimated

value of Vm as a function of
sample number for E = 0.5.
From this figure and table shown
before, it is evident that
1 As magnitude of zero mean
noise increases, the standard
deviation associated with
magnitude increase.
2 Mean of Vm is nearly 10.
3 Actual estimates seldom match
with 10V.
Summary
Summary
1 With bare minimum number of measurements, the noise

affects the accuracy.
2 Noise has to be filtered to estimate Vm and φv .
3 To filter out noise, we need to consider redundant
measurement.
4 Redundancy in measurement is defined as ratio of actual
number of measurement used for estimation to minimum
number of measurement required for estimation.
Summary
Thank you

L7 Lse Alg 1

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

L7 Lse Alg 1

Uploaded by

Copyright:

Available Formats

Two Sample Estimation

Result of two sample estimation under ideal condition

Least Square Method for Estimation of

Department of Electrical Engineering,

1 Two Sample Estimation

2 Result of two sample estimation under ideal condition

3 Result of estimation in presence of noise

Two sample estimation

In a numerical relaying setup,

Two sample estimation....contd.

The voltage and current waveforms can be represented by

Two sample estimation....contd.

Then, voltage samples at t1 and t2 are given by,

Two sample estimation....contd.

Two sample estimation... contd.

In such an event, it is not possible

Two sample estimation....contd.

Consider a signal V (t) = 10 sin(2π × 50 × t + 30◦ ), and a

Example 1... contd.

Compute the voltage phasor for

Example 1... contd.

Figure shown here also

The noise has a zero mean, and

Example 2... contd.

Let us model the voltage samples by

v (k ∆t) = 10 sin(2π × 50 × t + 30◦ ) + E ∗ randn (6)

Where, E models the standard deviation of noise.

Effect of Gaussian noise on estimation

Randn Mean Standard

They are calculated as per following equations. Let

the square of standard deviation (σ 2 ) is called variance.

Figure shows the plot of estimated

1 With bare minimum number of measurements, the noise

You might also like