Since this unit is in normal condition, the sharp spike near 2 MHz in the FRA plot could be due to poor electrical connections, loose tap changer contacts, or grounding issues. Improperly tightened bushing connections or high-contact resistance at terminals can introduce sudden impedance mismatches, causing resonance effects at high frequencies.
3.3. Statistical Analysis
This subsection presents a comparison carried out between the IFRA results of different SC and PD conditions on a statistical index basis. The absolute value of CSD, MM, and DABS ratios were applied for formulating the comparison. The presented parameters were used to provide distinct responsiveness in probing the variations between the compared cases. For this reason, it is advisable to employ multiple statistical parameters to reach a conclusion on the condition of the transformer. The measurement frequency of the test was categorized into five frequency bands to narrow fault conditions associated with each band to ensure maximum sensitivity. This method minimized the interpretation challenges.
In
Table 4, the statistical CSD method is used to analyze deviations in transformer response under different tap positions. In this case, deviations are observed across different frequency bands, with certain subbands showing significantly higher values. At 10–1000 Hz, the highest deviation is found in the lowest tap (−5%) at 10.38, followed by the highest tap (+5%) at 10.09. The mid-taps (±2.5%) recorded the lowest deviation at 7.06. This suggests that, at lower frequencies, tap position changes have a noticeable but relatively moderate impact. In the 1000–10,000 Hz band, the deviations decrease significantly. The highest tap (+5%) has a value of 3.81, while the lowest tap (−5%) recorded 4.05. Mid-taps remain at 3.14, showing that, within this range, there is greater uniformity in response across tap positions.
A major shift occurs in the 10,000–100,000 Hz range, where the highest tap (+5%) reaches an extreme deviation of 35.10, significantly higher than any other frequency band. The lowest tap (−5%) also exhibits a large deviation at 29.72, while mid-taps remain high at 26.99. This indicates that mid-frequency responses experience substantial shifts under different tap settings, suggesting winding deformations or resonance effects.
In the 100,000–1,000,000 Hz range, the deviations reduce but remain noticeable. The highest tap (+5%) records 12.09, while the lowest tap (−5%) follows closely at 11.11. Mid-taps settle at 9.79, showing a decrease in variation but still indicating measurable impact in this range.
Finally, in the 1,000,000–2,000,000 Hz range, another sharp rise in deviation is observed. The highest tap (+5%) reaches 34.71, with the lowest tap (−5%) at 34.53 and the mid-taps at 31.79. These values highlight that very high-frequency response changes significantly with tap position, likely due to increased capacitive and inductive effects. The most critical deviations occur in the 10,000–100,000 Hz and 1,000,000–2,000,000 Hz ranges, where extreme CSD values suggest significant alterations in the transformer’s internal impedance. These values warrant further investigation into possible structural changes in the winding system.
In
Table 5, the MM statistical method provides insight into how similar the frequency response remains across different tap settings. Unlike CSD, MM values are generally much smaller, indicating that, while variations exist, the magnitude response remains relatively stable. In the 10–1000 Hz band, the MM values are almost negligible, with the highest tap (+5%) at 0.14, mid-taps at 0.01, and the lowest tap at 0.01. These values suggest that the transformer’s response at low frequencies remains largely unchanged despite different tap settings. As frequency increases to the 1000–10,000 Hz range, a slightly larger deviation appears in the highest tap (+5%) at 0.43, while mid-taps and the lowest tap remain at 0.00. This suggests that some variations begin to emerge, but they are not substantial. In the 10,000–100,000 Hz and 100,000–1,000,000 Hz bands, the MM values stabilize at 0.01 for all tap settings. This indicates that, even in these frequency ranges, the overall magnitude response remains consistent.
Finally, in the 1,000,000–2,000,000 Hz range, the MM ratio drops completely to 0.00 across all tap positions, showing no meaningful deviation. This confirms that the MM method is not sensitive to structural or electrical changes observed in CSD or DABS. Given these results, the MM method appears to be ineffective at detecting tap-dependent variations in SC analysis, reinforcing the importance of using multiple statistical methods to gain a complete understanding of transformer behavior.
In
Table 6, the DABS method reveals absolute differences in frequency response, which can help identify variations caused by structural changes. Unlike MM, DABS tends to show more variation, especially in certain frequency bands. In the 10–1000 Hz band, the mid-taps (±2.5%) and lowest tap (−5%) show extremely high deviations at 61.78 and 65.19, respectively, while the highest tap (+5%) remains at 1.86. This is a significant finding, indicating that tap changes induce large variations in low-frequency response, possibly due to changes in leakage flux and winding movement. In the 1000–10,000 Hz band, deviations decrease but remain substantial. The highest tap (+5%) records 11.53, while the lowest tap (−5%) shows 12.91, suggesting that, while the impact is less severe than in the low-frequency band, tap changes still introduce variations. For the 10,000–100,000 Hz and 100,000–1,000,000 Hz bands, the deviations stabilize in the range of 6.43–11.62, indicating that variations persist but are not as extreme as in the lower frequency ranges. In the 1,000,000–2,000,000 Hz band, the deviations remain moderate, with values from 7.98–9.39. This suggests that, while high frequencies still experience variations, they are not as pronounced as at lower frequencies.
The most significant deviations appear in the 10–1000 Hz range, where mid-taps and the lowest taps show extreme variations. This highlights that low-frequency behavior is highly sensitive to mechanical changes in SC conditions.
In
Table 7, the CSD method identifies deviations in frequency response, making it useful for detecting electrical faults such as PD. In the 10–1000 Hz range, the values remain relatively low, with the highest tap (+5%) at 5.74, mid-taps (±2.5%) at 3.62, and the lowest tap (−5%) at 5.94. These values suggest that PD effects are minimal in the low-frequency range, which aligns with the general expectation that PD-induced deviations are more pronounced at higher frequencies. In the 1000–10,000 Hz range, deviations slightly decrease, with values ranging from 2.46 to 3.06. This suggests that while PD activity is present, it does not induce significant changes in this band. A major shift occurs in the 10,000–100,000 Hz band, where mid-taps (±2.5%) exhibit the highest CSD value of 26.12, followed by the lowest tap (−5%) at 24.50 and the highest tap (+5%) at 19.04. This is a critical subband where PD effects become highly prominent, suggesting that PD modifies the impedance characteristics of the winding at these frequencies.
In the 100,000–1,000,000 Hz range, the deviations drop again, but mid-taps still record a notable value of 7.85, which is higher than the neighboring frequency bands. This implies that, while PD effects are less significant at these frequencies, there is still some residual impact. The 1,000,000–2,000,000 Hz band shows another major increase in deviation, with mid-taps reaching 20.24, the second-highest deviation in the entire dataset. This suggests that high-frequency resonance effects due to PD are particularly dominant in the very high-frequency spectrum.
Key Findings from CSD Analysis
The most significant deviations appear in the 10,000–100,000 Hz and 1,000,000–2,000,000 Hz ranges.
Mid-taps (±2.5%) consistently show higher deviations than the other tap settings, indicating that this tap position is more sensitive to PD-related variations.
The increase in deviations at high frequencies suggests that PD alters the transformer’s high-frequency impedance response, possibly due to insulation degradation and increased dielectric losses.
In
Table 8, the MM method is used to compare the similarity in magnitude response across different cases. Unlike CSD, MM values remain relatively low, indicating that, while PD causes some variation, it does not significantly change the overall response shape. In the 10–1000 Hz range, MM values remain low, with the highest tap (+5%) at 0.14 and lowest tap (−5%) at 0.14, showing little distinction across tap settings. The 1000–10,000 Hz band records slightly higher values, peaking at 0.46 for mid-taps (±2.5%), suggesting that some variations are introduced by PD. For the 10,000–100,000 Hz and 100,000–1,000,000 Hz bands, MM values stabilize at 0.01 across all tap positions, indicating that PD does not significantly alter the magnitude response in these regions. At 1,000,000–2,000,000 Hz, MM values drop to 0.00, showing that, in this very high-frequency range, PD does not influence the similarity of the magnitude response.
Key Findings from MM Analysis
MM values remain low throughout all frequency ranges, indicating that the magnitude response remains mostly unchanged despite PD presence.
The highest MM deviation appears at 1000–10,000 Hz (0.46 at mid-taps), suggesting that minor PD-induced differences exist in this range.
The lack of significant MM deviations confirms that PD effects are better captured by CSD and DABS, rather than MM.
In
Table 9, the DABS method quantifies the absolute deviation in FRA response, making it a valuable parameter for identifying changes caused by PD. This method is particularly useful in isolating extreme deviations. At 10–1000 Hz, DABS values remain low (≤0.66), confirming that PD does not strongly influence this region. A noticeable increase occurs in the 1000–10,000 Hz range, where the lowest tap (−5%) reaches 6.92, while the highest tap (+5%) follows closely at 6.15. This suggests that PD begins to introduce measurable deviations in this frequency band.
In the 10,000–100,000 Hz band, the highest tap (+5%) reaches 7.60, showing that PD effects continue to grow in magnitude. A major anomaly is found in the 100,000–1,000,000 Hz range, where the lowest tap (−5%) records an extreme deviation of 107.04, which is drastically higher than any other recorded value. This suggests a significant PD-induced alteration in transformer response, likely due to dielectric breakdown or insulation aging. No other frequency band shows such a severe deviation. In the 1,000,000–2,000,000 Hz range, the DABS values return to moderate levels, with the highest tap (+5%) at 8.45 and lowest tap (−5%) at 7.61, indicating that the extreme PD-induced variation was concentrated in the previous frequency range.
Key Findings from DABS Analysis
The most critical anomaly occurs at 100,000–1,000,000 Hz, where DABS reaches 107.04 at the lowest tap (−5%).
High-frequency response is more affected by PD than low-frequency response, with increasing deviations from 1000–1,000,000 Hz.
DABS is highly effective in capturing PD-induced changes, more so than MM.
The extreme deviation suggests potential insulation failure or a localized PD event causing major changes in winding response.
The statistical analysis confirms that CSD, MM, and DABS exhibit distinct sensitivities to transformer condition variations, reinforcing the necessity of a multi-parameter approach for accurate diagnostics. The most critical deviations were observed in the 10,000–100,000 Hz and 1,000,000–2,000,000 Hz ranges, where CSD and DABS revealed extreme shifts, particularly at mid-taps and the lowest tap (−5%), indicating severe impedance alterations linked to SC and PD phenomena.