Page 123 - IJEEE-2022-Vol18-ISSUE-1

Jasim & Jasim | 119

pass filter with impulse response (g), the output is
convolutional between the x and g:

??(??) = (?? * ??)(??) = ?8??=-8 ??(??)??(?? - ??) (27)

Fig.12 DWT decomposition tree Fig.13 Faults detection and classification using DWT
The classifier classifies fault types (line to ground,
A high-pass filter (h) is used to deconstruct the signal at
the same time. The detail coefficients correspond to the high- double lines to ground, three lines to ground, and line to line
pass filter's outputs, while the approximation coefficients faults) based on the values of detailed coefficients calculated
correspond to the low-pass filter's outputs. According to by the wavelet transform for current phases A, B, and C, as
Nyquist's rule, half of the samples can now be discarded well as for ground, where the classifier calculates the
because half of the signal's frequencies have been deleted. maximum values for these detailed coefficients and
compares them to the threshold value according to table 1 of
The low-pass filter (g) output is then subsampled by 2 and the highest detailed coefficients. Fig.14 shows the flow chart
of the classification process.
further processed by running it through a new low-pass filter
g and a high-pass filter (h) with half the previous one's cut- Fig.14 flow chart of the classification process

off frequency, i.e.[53] (28) III. CASE STUDIES
????????(??) = ??8?=-8 ??(??)??(2?? - ??) (29) The proposed microgrid model may be subjected to five case
??h????h(??) = ?8??=-8 ??(??)h(2?? - ??) studies:
? The first case study:
Because just half of each filter output characterizes the
System operation in Grid-Following mode, PQ control,
signal, the time resolution has been decreased. The frequency and irradiance fluctuation.
resolution has been doubled since each output has half the ? Second case study:
M-grid transitions from grid-following to grid-forming,
frequency spectrum of the input. followed by grid resynchronization.
? The third case study:
The summation above might be expressed more Grid-Forming mode with the power curtailment of a
solar plant.
succinctly. ? Fourth case study:
M-grid in grid-following mode, compensating for
???????? = (?? * ??) ? 2 (30) imbalance (negative sequence).
??h????h = (?? * h) ? 2 (31) ? Fifth case study:
Detecting and classification of the several fault types
Computing a complete convolution (x*g) with occurring in the location between main grid and
subsequent down sampling, on the other hand, would be a transformer (TR4).
waste of time.

The approximation coefficients are decomposed with
low pass filters and then down-sampled to boost the
frequency resolution even further. This is depicted as a
binary tree, with each node indicating a distinct time-
frequency localization of a sub-space. A filter bank is the
name given to the tree.

By adopting the wavelet transform, the threshold value
is used to classify fault type in the power system. The value
of the threshold is not universal and cannot be applied to all
types of power systems. This threshold value will be
calculated for each new power system in order to correctly
categorize the various kinds of faults. Here, the threshold
value is selected equal to 100 according to the maximum
value of detailed phase and ground current coefficients for

various faults as shown in table 1 (in result section).

The basic goal of the feature extraction is to supply
significant information and maximum detail coefficients to
the classifier so that it can categorize the kind of event using
the computed features. The fault type is classified using the

classifier block based on the threshold value.