Chapter 2

Phasor Estimation of Nominal

Frequency Inputs

2.1 Phasors of Nominal Frequency Signals

Consider a constant input signal x(t) at the nominal frequency of the power system
f0, which is sampled at a sampling frequency Nf0. The sampling angle θ is equal to
2π/N, and the phasor estimation is performed using Eqs. (1.25)–(1.27).

xðtÞ ¼ Xm cosð2pf0 t þ /Þ ð2:1Þ

The N data samples of this input xn:{n = 0, 1, 2, …, N − 1) are

xn ¼ Xm cosðnh þ /Þ ð2:2Þ

Since the principal interest in phasor measurements is to calculate the funda-

mental frequency component, we will set k = 1 in Eqs. (1.25)–(1.27) to produce the
fundamental frequency phasor obtained from the sample set xn. The superscript
(N − 1) is used to identify the phasor as having the (N − 1)st sample as the last
sample used in the phasor estimation.
pffiffiffi N 1 pffiffiffi N 1
2X 2X
XcN1 ¼ xn cosðnhÞ ¼ Xm cosðnh þ /Þ cosðnhÞ
N n¼0 N n¼0
pffiffiffi N 1  
2 X 1 Xm
¼ Xm cosð/Þ cos ðnhÞ  sinð/Þ sinð2nhÞ ¼ pffiffiffi cosð/Þ
2
ð2:3Þ
N n¼0
2 2

It is to be noted that the summation of the sin(2nθ) term over one period is
identically equal to zero, and that the average of the cos2(nθ) term over a period is
equal to 1/2.

© Springer International Publishing AG 2017 29

A.G. Phadke and J.S. Thorp, Synchronized Phasor Measurements
and Their Applications, Power Electronics and Power Systems,
DOI 10.1007/978-3-319-50584-8_2
30 2 Phasor Estimation of Nominal Frequency Inputs

The sine sum is calculated in a similar fashion:

pffiffiffi N 1 pffiffiffi N 1
2X 2X
XsN1 ¼ xn sinðnhÞ ¼ Xm cosðnh þ /Þ sinðnhÞ
N n¼0 N n¼0
pffiffiffi N1  
2 X 1
¼ Xm cosð/Þ sinð2nhÞ  sinð/Þ sin2 ðnhÞ
N n¼0
2
Xm
¼  pffiffiffi sinð/Þ ð2:4Þ
2

The phasor XN−1 is given by

Xm Xm
X N1 ¼ XcN1  jXsN1 ¼ pffiffiffi ½cosð/Þ þ j sinð/Þ ¼ pffiffiffi ej/ ð2:5Þ
2 2

It is to be understood that Eq. 2.5 gives the fundamental frequency phasor

estimate, even though the subscript k = 1 has been dropped for the sake of sim-
plicity. The result obtained in Eq. (2.5) conforms with the phasor definition given in
Chap. 1, and the phase angle ϕ of the phasor is the angle between the time when the
first sample is taken (corresponding to n = 0) and the peak of the input signal.

2.2 Formulas for Updating Phasors

2.2.1 Non-recursive Updates

Considering that the phasor calculation is a continuous process, it is necessary to

consider algorithms which will update the phasor estimate as the newer data
samples are acquired. When the Nth sample is acquired after the previous set of
samples has led to the phasor estimate given by Eq. (2.5), the simplest procedure
would be to repeat the calculations implied in Eqs. (2.3)–(2.5) for the new data
window which begins at n = 1 and ends at n = N.
pffiffiffi N 1
2X
XN1
¼ xn ½ðcosðnhÞ  j sinðnhÞ
N n¼0
pffiffiffi N 1 ð2:6Þ
2X
XN ¼ xn þ 1 ½ðcosðnhÞ  j sinðnhÞ
N n¼0

The two windows are shown in Fig. 2.1. Phasor 1 is the result of phasor esti-
mation over window 1, while phasor 2 is calculated with the data in window 2. The
first sample in window 1 is lagging the peak of the sinusoid by an angle ϕ, while the
2.2 Formulas for Updating Phasors 31

φ
φ+θ
n = N -1
n=0 n=N phasor 2 phasor 1
θ
n=1 φ

window 1
window 2

(a) (b)

Fig. 2.1 Update of phasor estimates with N sample windows. Phasor 1 is calculated with samples
n = 0, …, N − 1, while phasor 2 is calculated with samples n = 1, 2, … N. θ is the angle between
successive samples based on the period of the fundamental frequency

first sample of window 2 (n = 1) lags the peak by an angle (ϕ + θ), θ being the
angle between samples.
It should be clear from Fig. 2.1 that in general, the phasor obtained from a
constant sinusoid of nominal power system frequency by this technique will have a
constant magnitude and will rotate in the counter-clockwise direction by angle θ as
the data window advances by one sample. Since the phasor calculations are per-
formed fresh for each window without using any data from the earlier estimates,
this algorithm is known as a non-recursive algorithm. Non-recursive algorithms are
numerically stable but are somewhat wasteful of computation effort as will be seen
in the following.
Figure 2.2 is another view of the non-recursive phasor estimation process. As
newer samples are obtained, the table of sine and cosine multipliers is moved down
to match the new data window. In this figure, the multipliers are viewed as samples
of unit magnitude sine and cosine waves at the nominal power system frequency.
The new data window has N − 1 samples in common with the old data window. In
actual computation, these are simply stored as tables of sine and cosine, which are
used repeatedly on each window as needed.

2.2.2 Recursive Updates

The formulas for calculating the (N − 1)th and (N)th phasors by the non-recursive
algorithm are
32 2 Phasor Estimation of Nominal Frequency Inputs

φ φ+θ Samples of unit magnitude sine

and cosine functions

Input
signal phasor 2
New sample,
window 2 φ2 = φ1 + kθ
t phasor 1
φ1

Fig. 2.2 Non-recursive phasor estimation. There are 12 samples per cycle of the power frequency
in this example. Fresh calculations are made for each new window as new samples are obtained.
The phasor for a constant input signal rotates in the counter-clockwise direction by the sampling
angle, in this 30°

pffiffiffi N 1
2 X jnh
X N1 ¼ xn e
N n¼0
pffiffiffi N 1 ð2:7Þ
2X
X ¼
N
xn þ 1 ejnh
N n¼0

The multipliers for a given sample are different in the two computations. For
example, the multiplier for (n = 2) sample in the first sum is e−j2θ, while the
multiplier for the same sample in the second sum is e−jθ.
It should be noted that samples xn: {n = 1, 2, …, N − 1) are common to both
windows. The second window has no x0, so that it begins with x1, and it ends with
xN, which did not exist in the first window. If one could arrange to keep the
multipliers for the common samples the same in the two windows, one would save
considerable computations in calculating XN. If we multiply both sides of the
second equation in (2.6) by e−jθ, we obtain the following result:
pffiffiffi N 1
_N 2X
X ¼ ejh X N ¼ xn þ 1 ejðn þ 1Þh
N n¼0
pffiffiffi
2
¼ X N1 þ ðxN  x0 Þejð0Þh ð2:8Þ
N

where use has been made of the fact that e−j(0)θ = e−jNθ, since N samples span
exactly one period of the fundamental frequency. The phasor defined by Eq. (2.7)
2.2 Formulas for Updating Phasors 33

differs from the non-recursive estimate by an angular retardation of θ. The

advantage of using this alternative definition for the phasor from the new data
window is that (N − 1) the multiplications by the Fourier coefficients in the new
window are the same as those used in the first window. Only a recursive update on
the old phasor needs to be made to determine the value of the new phasor. This
algorithm is known as the recursive algorithm for estimating phasors. In general,
when the last sample in the data window is (N + r), the recursive phasor estimate is
given by
pffiffiffi
_N þr 2
jh N þ r1
X ¼e X þ ðxN þ r  xr Þejrh
pffiffiffi N
_ N þ r1 2
¼X þ ðxN þ r  xr Þejrh ð2:9Þ
N

When the input signal is a constant sinusoid, xN+r is the same as xr, and the
second term in Eq. (2.8) disappears. The phasor estimate with data from the new
window is the same as the phasor estimate with data from the old window when the
input signal is a constant sinusoid. In general, the recursive algorithm is numerically
unstable. Consider the effect of an error in the estimate from one window—for
example, caused by a round-off error. This error is always present in all the phasor
estimates from then on. This property of the recursive phasor algorithms must be
kept in mind when practical implementation of these algorithms is performed [1].
Nevertheless, because of the great computational efficiency of the recursive algo-
rithm, it is usually the algorithm of choice in many applications.
Unless stated otherwise explicitly, we will assume that only the recursive form
of the phasor estimation algorithm is in use (Fig. 2.3).

φ φ+θ Samples of unit magnitude sine

and cosine functions

Input New samples,

signal New sample, sine and cosine
window 2 φ2 = φ1
t phasor 1 & 2
φ1

Fig. 2.3 Recursive phasor estimation. There are 12 samples per cycle of the power frequency in
this example. Fresh calculations are made for each new window as new samples are obtained. New
sine and cosine multipliers are used on the new sample. The phasor for a constant input signal
remains stationary
34 2 Phasor Estimation of Nominal Frequency Inputs

Example 2.1 Consider the 60 Hz signal

xðtÞ ¼ 100 cosð120pt þ p=4Þ

sampled at the rate of 12 samples per cycle, i.e., at a sampling frequency of 720 Hz.
The first 18 samples, and the non-recursive and recursive phasor estimates obtained
using Eqs. (2.6) and (2.8) beginning with sample no. 12 (at which time the first data
window is completely filled) are shown in the following (Table 2.1).
As expected, the non-recursive phasor estimates produce a constant magnitude
of 100/√2 with an initial angle of π/4 (45°), and then for each successive estimate,
the angle increases by 30°.

2.3 Effect of Signal, Noise, and Window Length

The input signals are rarely free from noise. A spurious frequency component
which is not a harmonic of the fundamental frequency signal may be considered to
be noise. One may also have induced electrical noise picked up in the wiring of the
input signal. Leakage effect caused by the windowing function has already been
discussed in Chap. 1, and it too contributes to an error in phasor estimation and
should therefore be considered as a type of noise in the input.

Table 2.1 Phasor estimates of sampled data

Sample No. Sample xn Non-recursive phasor estimate Recursive phasor estimate
0 70.7107
1 25.8819
2 −25.8819
3 −70.7107
4 −96.5926
5 −96.5926
6 −70.7107
7 −25.8819
8 25.8819
9 70.7107
10 96.5926
11 96.5926
12 70.7107 70.701∠45° 70.701∠45°
13 25.8819 70.701∠75° 70.701∠45°
14 −25.8819 70.701∠105° 70.701∠45°
15 −70.7107 70.701∠135° 70.701∠45°
16 −96.5926 70.701∠165° 70.701∠45°
17 −96.5926 70.701∠195° 70.701∠45°
2.3 Effect of Signal, Noise, and Window Length 35

As an approximation, we will consider the noise in the input signal to be a zero

mean, Gaussian noise process. This should be a good approximation for the elec-
trical noise picked up in the wiring and signal conditioning circuits. The other two
sources of noise, viz. non-harmonic frequency components and leakage phenomena
need further consideration. A phasor measurement system may be placed in an
arbitrarily selected substation and will be exposed to input signals generated by the
power system which is likely to change states all the time. Each of the power
system states may lead to different non-harmonic frequencies and leakage effects,
and the entire ensemble of conditions to which the phasor measurement system is
exposed may also be considered to be a pseudo-random Gaussian noise process.
Consider a set of noisy measurement samples

xn ¼ Xm cosðnh þ /Þ þ en ; fn ¼ 0; 1; 2; . . .; N  1g ð2:10Þ

where εn is a zero-mean Gaussian noise process with a variance of σ2. If we set

(Xm/√2) cos (ϕ) = Xr and (Xm/√2) sin (ϕ) = Xi, the phasor representing the sinusoid
is X = Xr + jXi. We may pose the phasor estimation problem as one of the findings—
the unknown phasor estimate from the sampled data through a set of N overdeter-
mined equations:
2 3 2 3 2 3
x0 cosð0Þ  sinð0Þ e1
6 x1 7 6 cosðhÞ  sinðhÞ 7  6 e2 7
6 7 pffiffiffi6 7 Xr 6 7
6 7 ¼ 26 cosð2hÞ  sinð2hÞ 7 6 7
6 x2 7 6 7 X i þ 6 e3 7 ð2:11Þ
4  5 4   5 4  5
xN1 cos½ðN  1Þh  sin½ðN  1ÞhÞ eN1

or, in matrix notation

½x ¼ ½S½X þ ½e ð2:12Þ

Assuming that the covariance matrix W of the error vector is σ2 multiplied by a

unit matrix

½W  ¼ r2 ½1 ð2:13Þ

the weighted least squares solution of Eq. (2.11) provides the estimate for the
phasor
_
½X  ¼ ½ST W 1 S1 ST W 1 ½x ð2:14Þ

Using (2.13) for W, and calculating {STS]−1 for the S in Eq. (2.11):

_ 1 T
½X  ¼ ½ST W 1 S1 ½ST W 1 ½x ¼ ½ST S1 ½ST ½x ¼ ½S ½x ð2:15Þ
N
36 2 Phasor Estimation of Nominal Frequency Inputs

Since the noise is a zero-mean process, the estimate given by (2.15) is unbiased,
and the expected value of the estimate is equal to the true value of the phasor. If X is
the true value of the phasor, the covariance matrix of the error in the phasor estimate
is
h_ _
i
E ½X  X½X  XT ¼ ½ST W 1 S1 ð2:16Þ

Substituting for [W] from Eq. (2.13), the covariance of the error in phasor
estimate is (σ2/N). The standard deviations of error in real and imaginary parts of
the phasor estimate are (σ/√N). We may thus conclude that higher sampling rates
will produce improvement in phasor estimates in inverse proportion of the square
root of the number of samples per cycle. Alternatively, if longer data windows are
used (multiple of cycles), then once again the errors in phasor estimate go down as
the square root of the number of cycles used. Thus, a four cycle phasor estimate is
twice as accurate as a one cycle estimate in with noisy input.
Example 2.2 Consider a 60 Hz sinusoid

xðtÞ ¼ 100 cosð120pt þ p=4Þ þ eðtÞ

in a noisy environment, with the Gaussian noise ε having a zero mean and a
standard deviation of 1. The source of noise could be electromagnetic interference,
quantization errors, or harmonic and non-harmonic components in the input signal.
If the phasing of the harmonic and non-harmonic signals is random, the noise model
may be approximated by a zero-mean Gaussian characteristic.
The signal is sampled at six different sampling rates: 8, 16, 32, 64, 128, and 256
times per cycle. The signal samples are created with appropriately modeled noise
input for 1000 cycles, and 1000 estimates of the phasor value are calculated. The
standard deviations of the errors in the 1000 phasor estimates as well as its theo-
retical value (σ/√N) are given in Table 2.2 and are also shown in Fig. 2.4.
These results show very good agreement with the expected results. As men-
tioned earlier, increasing the data window size at a fixed sampling rate, rather than
the number of samples in the same data window will produce similar results.

Table 2.2 Phasor estimation of a noisy signal

No. of samples Standard deviation of Standard deviation of phasor σ/√N
per cycle N input noise estimate error (volts)
8 1 0.3636 0.3536
16 1 0.2601 0.2500
32 1 0.1794 0.1768
64 1 0.1231 0.1250
128 1 0.0880 0.0884
256 1 0.0626 0.0625
2.3 Effect of Signal, Noise, and Window Length 37

Fig. 2.4 Standard deviation 0.35

of phasor error due to standard deviation of phasor error
zero-mean Gaussian noise in (volts)
the input. The result of 1000 0.25
phasor estimates at each
sampling rate is shown by the σ/√N
solid line, and the theoretical 0.15
value of the standard
deviation (σ/√N) is shown by
the dotted line 0.05
2 4 8 16 √N

2.3.1 Errors in Sampling Times

Another possible source of error in input data samples is the error in timing of the
sampling pulses. One possible source of errors is that the sampling clock is not precisely
at a multiple of the power system frequency. This case will be dealt with in the next
chapter, when we consider the phasor estimation problem at off-nominal frequencies.
In this section, we will consider the case where the sample times are corrupted by
a Gaussian random noise with standard deviation varying by up to 10% of the
sampling interval. Large errors of this type should not exist in modern measuring
systems. Nevertheless, when sampling pulses are generated with the help of software
clocks, it is possible to encounter random errors in sampling times. The following
numerical example considers errors of this type. The errors are truncated at 3 times
the standard deviation in order to eliminate impossibly large sampling clock errors.
Example 2.3 Consider a 60 Hz sinusoid

xðtÞ ¼ 100 cosð120pt þ p=4Þ

which is sampled at tn = nΔT + ε, where the Gaussian noise ε has a zero mean and
a standard deviation of bΔT, with the parameter b varying between 0.0 and 0.10.
The signal is sampled at a sampling rate of 32 times per cycle. The signal samples
are created with sampling time errors for 1000 cycles, and 1000 estimates of the
phasor value are calculated. The standard deviations of the errors in the 1000 phasor
estimates are given in Table 2.3 and are also shown in Fig. 2.5.

2.4 Phasor Estimation with Fractional Cycle Data

Window

The weighted least squares solution technique developed in Sect. 2.2 is a conve-
nient vehicle for calculating phasors from fractional cycle data windows. It should
be remembered that fractional cycle phasor estimates are necessary in developing
38 2 Phasor Estimation of Nominal Frequency Inputs

Table 2.3 Errors in phasor estimation due to noisy inputs

Coefficient ‘b’ of input noise described Standard deviation of phasor estimate error
above (volts)
0.00 0.0000
0.02 0.0705
0.04 0.1373
0.06 0.2111
0.08 0.2809
0.10 0.3432

Fig. 2.5 Standard deviation

Standard deviation of errors in

of phasor error due to 0.3

zero-mean Gaussian noise in
phasor estimate (volts)

sampling clock pulses. The

result of 1000 phasor
estimates at different values of 0.2
‘b’ coefficients are shown

0.1

0 ‘b’
0 0.02 0.04 0.06 0.08 0.10

high speed relaying applications, and not particularly useful in wide area phasor
measurement applications whereby measurement times of a few cycles are
acceptable. Nevertheless, it is instructive to include a discussion of fractional cycle
phasor estimation.
Consider the use of M samples of a sinusoid for estimating phasors, the sinusoid
having been sampled at a sampling rate of N samples per cycle. M < N produces a
fractional cycle phasor estimation algorithm.
As before, the input is set of noisy measurement samples

xn ¼ Xm cosðnh þ /Þ þ en ; fn ¼ 0; 1; 2; . . .; M  1Þ ð2:17Þ

where εn is a zero-mean Gaussian noise process with a variance of σ2. The sampling
angle θ is equal to 2π/N.
2 3 2 3 2 3
x0 cosð0Þ  sinð0Þ e1
6 x1 7 6 cosðhÞ  sinðhÞ 7   6 e2 7
6 7 pffiffiffi6 7 Xr 6 7
6 7 ¼ 26 cosð2hÞ  sinð2hÞ 7 6 7
6 x2 7 6 7 Xi þ 6 e3 7 ð2:18Þ
4  5 4   5 4  5
xM1 cos½ðM  1Þh  sin½ðM  1ÞhÞ eM1
2.4 Phasor Estimation with Fractional Cycle Data Window 39

or, in matrix notation

½ x ¼ ½S½ X  þ ½e ð2:19Þ

As before, the weighted least squares solution of Eq. (2.17) provides the esti-
mate for the phasor
_
½X  ¼ ½ST W 1 S1 ST W 1 ½x ð2:20Þ

Using (2.13) for W, and calculating {STS]−1 for the S in Eq. (2.11):
_
½X  ¼ ½ST W 1 S1 ½ST W 1 ½x ¼ ½ST S1 ½ST ½x ð2:21Þ

Unlike in the case of the full cycle phasor estimation, [STS]−1 is no longer a
simple matrix:
2 3
P
M1 P
M1

6 cos2 ðnhÞ cosðnhÞ sinðnhÞ 7

½ST S ¼ 26
4 M1
n¼0 n¼0 7
5 ð2:22Þ
P P
M1
cosðnhÞ sinðnhÞ sin ðnhÞ
2
n¼0 n¼0

It can be shown that for a half-cycle estimation, with M = N/2, the least squares
solution is very similar to the DFT estimator.

2.5 Quality of Phasor Estimate and Transient Monitor

Phasor estimates obtained from a data window represent the fundamental frequency
component of the input confined to the data window. When a fault occurs on the
power system, there is a series of data windows which contain pre- and post-fault
data. This is illustrated in Fig. 2.6 for an assumed voltage waveform during a fault.
It should be clear that although a phasor estimate will be available for all data
windows (including the ones that are shaded in Fig. 2.6), only phasors which
belong entirely to the pre- or post-fault periods are of interest. The phasors com-
puted for the shaded windows of Fig. 2.6 do not represent any meaningful system
state, and a technique is needed to detect the occurrence of mixed states within a
data window.
A technique known as ‘Transient Monitor’ [2] provides a measure to indicate a
‘quality’ of the estimate and can also be used to detect the condition when a data
window contains mixed state waveforms. Consider the process of computing the
_
data samples ðx n Þ in a window from the estimated phasor which has been estimated
from a sample set (xn):
40 2 Phasor Estimation of Nominal Frequency Inputs

Pre-fault
Post-fault

Windows
with all
pre-fault data

Windows with all

post-fault data

Fig. 2.6 Transition from pre-fault to post-fault waveforms. The shaded windows contain mixed
waveform data

2 3
cosð0Þ  sinð0Þ
6 cosðhÞ  sinðhÞ 7" _ #
pffiffiffi6 7 Xr
½xn  ¼ 26  sinð2hÞ 7
_
6 cosð2hÞ 7 _ ð2:23Þ
4   5 Xi
cos½ðN  1Þh  sin½ðN  1Þh

Substituting for the phasor estimate from Eq. (2.15)

2 3
cosð0Þ  sinð0Þ
6 7
h i pffiffiffi6 cosðhÞ  sinðhÞ 7
6 7
x n ¼ 26 7
_
6 cosð2hÞ  sinð2hÞ 7
6 7
4   5
cos½ðN  1Þh  sin½ðN  1Þh
pffiffiffi  
2 cosð0Þ cosðhÞ cosð2hÞ  cos½ðN  1ÞhÞ
 ½xn  ð2:24Þ
N  sinð0Þ  sinðhÞ  sinð2hÞ   sin½ðN  1ÞhÞ

Multiplying the matrices and simplifying

2 3
1 cosðhÞ cosð2hÞ  cos½ðN  1ÞhÞ
h i 266 cosðhÞ 1 cosðhÞ  cosð0Þ 7
7
xn ¼ 6 7½xn 
_
cosð2hÞ cosðhÞ 1  cosðhÞ ð2:25Þ
N6
4
7
5
   1 
cos½ðN  1ÞhÞ cosð0Þ cosðhÞ  1
2.5 Quality of Phasor Estimate and Transient Monitor 41

where use has been made of the fact that N = 2π. The difference between the input
data and the re-computed sample data from the phasor estimate is the error of
estimation [tn]:
_
½tn  ¼ ½xn  x n 
2 3
1  N2  N2 cosðhÞ  N2 cosð2hÞ   N2 cos½ðN  1ÞhÞ
6 7
6  N2 cosðhÞ 1  N2  N2 cosð3hÞ   N2 cosð0Þ 7
6 7
¼66  N cosð2hÞ
2
 N2 cosð3hÞ 1  N2   N2 cosðhÞ 7½xn 
7
6 7
4    1  N2  5
 N2 cos½ðN  1ÞhÞ  N2 cosð0Þ  N2 cosðhÞ  1  N2
ð2:26Þ

If the input signal is a pure sinusoid at fundamental frequency, all entries of [tn]
will be identically equal to zero. However, when the input signal is noisy or
contains a composite window of two different sinusoids, [tn] is not zero, and one
may use the sum (Tn) of the absolute values of its elements as a measure of the error
of estimation.

X
N 1
Tn ¼ jtk j ð2:27Þ
k¼0

This sum has been referred as a ‘Transient Monitor’ and can be used as a
measure of the ‘quality’ of the phasor estimate.
Example 2.4 Consider a composite 60 Hz voltage waveform samples described by

xn ¼ 100 cosðnh þ p=4Þ; for n ¼ 0; 1; 2; . . .; 35

xn ¼ 50 cosðnh þ p=8Þ; for n ¼ 36; 37; 38; . . .; 71

with a sampling rate of 24 samples per cycle; thus, θ = π/12.

The data samples, recursive phasor estimates, and the function Tn are shown in
Table 2.4.
Note that in the interest of saving space, several rows which do not show
interesting transitions have been omitted. The phasor estimates and the transient
monitor are plotted in Fig. 2.7.
Note that the phasor estimate remains stationary at (50 + j50) and
(32.6641 + j13.5299), while the input signal is 70.7∠45° and 50∠22.5°, respec-
tively, and the transition from one value to another takes 24 samples, the width of
phasor estimation window. The transient monitor provides a good indication of the
quality of the phasor estimate, it being high during the transition period when the
phasor estimate is unreliable.
42 2 Phasor Estimation of Nominal Frequency Inputs

Table 2.4 Transient Monitor Sample Sample Phasor Tn

for a transient signal No. value
1 70.7107 0
2 50.0000 0
23 96.5926 0
24 86.6025 50.0 + j50.0 0.0000
36 −86.6025 50.0 + j50.0 0.0000
37 −46.1940 48.5553 + j50.0 51.4677
38 −39.6677 47.9672 + j50.1576 67.5480
39 −30.4381 48.1998 + j50.0233 68.1205
40 −19.1342 48.9970 + j49.2261 80.9808
41 −6.5263 49.9518 + j47.5723 129.1521
42 6.5263 50.6149 + j45.0978 195.2854
43 19.1342 50.6149 + j42.0587 267.1946
44 30.4381 49.7583 + j38.8619 331.2446
45 39.6677 48.0811 + j35.9570 382.1787
46 46.1940 45.8392 + j33.7151 413.6845
47 49.5722 43.4397 + j32.3297 430.7721
48 49.5722 41.3320 + j31.7650 436.9438
49 46.1940 39.8874 + j31.7650 432.8577
50 39.6677 39.2993 + j31.9225 426.1713
51 30.4381 39.5318 + j31.7883 431.3592
52 19.1342 40.3290 + j30.9910 433.7538
53 6.5263 41.2839 + j29.3372 424.5001
54 −6.5263 41.9469 + j26.8628 405.5234
55 −19.1342 41.9469 + j23.8236 367.7588
56 −30.4381 41.0903 + j20.6269 314.0214
57 −39.6677 39.4132 + j17.7219 243.3327
58 −46.1940 37.1713 + j15.4800 162.9157
59 −49.5722 34.7718 + j14.0947 77.7374
60 −49.5722 32.6641 + j13.5299 0.0000
72 49.5722 32.6641 + j13.5299 0.0000

2.6 DC Offset in Input Signals

Fault currents in a power system often have an exponentially decaying dc com-

ponent, which is generally known as the dc offset. Occasionally, voltage waveforms
may also have a dc offset due to capacitive voltage transformer transients. In both
cases, the dc offsets decay to negligible values in a few cycles. If the phasor
estimate is performed while a dc offset is present in a waveform, one is likely to get
significant errors in phasor estimate while the dc offset is nonzero. The transient
monitor described in Sect. 2.4 can be used to alert the user that the phasor estimate
thus obtained is unreliable.
2.6 DC Offset in Input Signals 43

450
40

20

0 0
0 20 40 1 24 36 60 72
(a) (b)

Fig. 2.7 Result of phasor estimation in a data stream with mixed input signals. a Phasor
estimates. The transition from a solid phasor at 50 + j50 to a new phasor of 32.6641 + j13.5299 is
shown by open circles. b The transient monitor Tn. Note that it is high during the transition from
one phasor value to another. When the input signal is a pure sinusoid, the Tn becomes zero

Since many phasor applications are dedicated to relatively slow phenomena, it is

not essential that dc offset be handled in any special way; one only needs to be alert
to the estimate quality indicated by the transient monitor. However, in computer
relaying applications, powerful techniques have been developed to remove dc
offsets before phasors are estimated, and in very specific applications of phasors
which require very high speed of response, it may be necessary to employ algo-
rithms which will remove the dc offset from the signals. This section provides a
brief summary of the available techniques for this purpose. We will consider only
the dc offset in current waveforms when a fault occurs. Similar techniques are
applicable to the handling of dc offset in the voltage waveforms as well.
The earliest technique used in relays for removing the dc offset from fault
currents is the one of using a ‘mimic’ circuit in the secondary winding of a current
transformer [3]. Figure 2.8 shows the primary fault circuit, and the current trans-
former secondary winding with a burden (r + jωl) such that the ratio R/L is matched
exactly by the burden ratio r/l. In this case, the dc offset in the current is not present
in the voltage e2(t) across the burden, and the burden voltage can be used as a signal
which is proportional to the current and is free from the dc offset. The primary fault
circuit and the CT secondary burden are both primarily inductive in nature, and
hence, the mimic circuit acts as a differentiator. Thus, it has the property of
amplifying any high frequency noise that may be present in the current. However, it
should be remembered that the primary fault current is itself is produced by the
R–L circuit, and thus has attenuated the high frequency noise that may be present in
the voltage signal. Thus, although the mimic circuit is a differentiator, the noise
content of the voltage across it is similar to that in the primary voltage.
For computer relays, there is a least squares solution technique available for
eliminating the dc offset, which is free from the noise amplification properties of the
mimic circuit. Consider a fault current i(t), containing a dc offset is given by

iðtÞ ¼ A cosðxtÞ þ B sinðxtÞ  Cet=T for t  0

¼AC t ¼ 0 ð2:28Þ
44 2 Phasor Estimation of Nominal Frequency Inputs

Fig. 2.8 Fault circuit and the

e(t) L R Fault
mimic burden in CT
secondary to eliminate the dc
offset in fault current i(t)

i2 (t) l
e2(t)
r

This expression assumes that the current just before the occurrence of fault is
(A − C), and that the dc offset decays with a time constant T, which for the circuit
of Fig. 2.8 is equal to L/R seconds.
Consider a sample set of M data points obtained from this current waveform

in ¼ A cosðnhÞ þ B sinðnhÞ  Crn ; for fn ¼ 0; 1; 2; . . .; M  1Þ ð2:29Þ

where θ is the sampling angle equal to 2π/N, N being the number of samples per
cycle of the nominal frequency, and r is the decrement factor for the decaying dc
component in one sample time

r ¼ eDT=T ð2:30Þ

If we now assume that the decrement factor r is known, the only unknowns in
Eq. (2.30) are A, B, and C. Taking the overdetermined set of M data points,
2 3 2 3
i0 1 0 1 2 3
6 i1 7 6 A
6 7¼6 cos h sin h r 7 74 B 5
4 5 4 5 ð2:31Þ
C
iM1 cosðM  1Þh sinðM  1Þh r M1

As usual, the above equation can be solved for A, B, and C, and then by adding
(Cr−n) to each sample of the current, the dc offset can be removed from the
waveform.
It is possible that the value of the time constant T is not known exactly and an
approximate value must be used for ‘r’. The algorithm is tolerant of reasonable
errors in the value of ‘r’, as is seen by the following numerical example.
Example 2.5 Consider a fault current waveform with full dc offset given by

iðtÞ ¼ 100 cosð120ptÞ þ 100 sinð120ptÞ100et=0:05

The current is zero before the occurrence of the fault. The dc offset decay time
constant is 50 ms.
2.6 DC Offset in Input Signals 45

Fig. 2.9 Errors in phasor 0.15 assumed ‘T’ = 45 ms

estimate caused by errors in
time constant estimate. Two
cycles worth of data are (100 – j100)
shown. The solid arrows
show the direction of 0
increasing time

assumed ‘T’ = 55 ms
-0.15
-0.15 0 0.15

The true value of the phasor must be (100/√2)(1 − j1). The dc offset is removed
by applying Eq. (2.31) for a window of one cycle at a time. It is assumed that an
error of ±10% is made in the decrement factor ‘r’ used in Eq. (2.31). The resulting
error in (√2 × phasor) is shown in Fig. 2.9.
It can be seen from Fig. 2.9 that the errors of estimation of phasors are less than
0.2% even though the time constant errors are of the order of 10%.
The least square solution described above is some times described as a ‘digital
mimic’ procedure. However, it must be pointed out that this process is not a
differentiator, and consequently, there is no amplification of noise in the current
signal in this process.

2.7 Non-DFT Estimators

A number of papers dealing with the problem of computing phasors from sampled
data have been published over the last several years. Several papers consider
variations on the Fourier transform method, with special emphasis on the problem
of dealing with off-nominal frequency signals. We will consider such signals in the
next chapter, where the performance of the fixed frequency Fourier transform on
off-nominal frequency input signals will be discussed. Among the variations of the
basic Fourier technique, least squares methods, Kalman filter methods, and Prony
methods have been discussed. As the main thrust of these variations is to deal with
off-nominal frequency signals, we will defer their discussion to a later chapter.

References

1. Phadke, A. G., & Thorp, J. S. (1994). Computer relaying for power systems (pp. 127–129).
New York: Research Studies Press Ltd., Wiley.
2. ibid. pp. 151–152.
3. Mason, C. R. (1956). Art and science of protective relaying. New York: Wiley.
http://www.springer.com/978-3-319-50582-4