Measurement Algorithms for Digital Protection - Digital Signal Processing in Power System Protection and Control - page 166

(E) Recommendations

On the strength of the above remarks one may try to recommend some of the
very many-described algorithms for practical applications, those with the
most favorable resulting virtues, fulfilling the standard or more specific
requirements.

• Magnitude and power measurement algorithms
Typical example of this kind applies full-cycle FIR filtration (sin/cos win-
dows) for both elimination of distortion and realization of signal orthogo-
nalization. The filters' gain values are identical and thus the final algorithm
has the simplest possible form. The recursive filtration or correlation proce-
dures assure effective filtration of unwanted signal components, with the
resulting estimation time equal 20 ms (a cycle of fundamental frequency
component). If the signal distortions are small, one can recommend Walsh
filters. In case when shorter measurement time is of importance, the methods
based on least mean square error, Kalman filter, shorter window orthogonal
filters or methods employing separated orthogonalization can be applied.
When the permissible measurement time is 25 ms or longer one can apply
single filter plus orthogonalization with time delay by a quarter of cycle.
Sometimes it may be important to use algorithms employing delayed
orthogonal components, characterized by decreased susceptibility to small
frequency distortions around nominal value. Their further usage in the quo-
tient type algorithms of other criteria measurement may also be advantageous.
All the considered algorithms perform well independently of the sampling
frequency level. Also in this respects there are no visible reasons for the
utilization of averaging algorithms.
• Quotient type impedance measurement algorithms
The discussion from previous point applies also to the algorithms of imped-
ance components estimation that employ standard algorithms for powers and
signal magnitudes as intermediate variables. It is recommended to use the
same filters for the measured variables in the numerator and denominator. The
best effects are obtained with application of algorithms based on delayed
orthogonal components, which offer practical independence of the frequency
deviations and comparatively simple final equations, with the filter gain
coefficients cancelled from the numerator and denominator.
• Frequency measurement by impulse counting
This algorithm differs from all others by the general principle of measure-
ment. The accuracy of measurement can be increased by changing the sam-
pling rate; however, for most of power system protection and control
applications the resulting precision is not sufficient. Necessary for getting
better results is an extension of the time of counting (over multiple cycles) or
application of improved method with zero-crossing correction. Advantage of
this approach is its simplicity, but one should remember that the result of
measurement is delivered every half-cycle only (not a continuous process).