Distinction Between Interturn Short-Circuit Faults and Unbalanced Load in Transformers

Abstract

Transformers are essential in electrical networks, and their failure can lead to the shutdown of a section or the entire grid. This study proposes a combination of techniques for early fault detection, distinguishing between small load imbalances and incipient interturn short circuits. An experimental setup was developed using a three-phase transformer bank with three single-phase dry-type transformers. One transformer was modified to create controlled short circuits of two and four turns and to simulate a load imbalance by reducing the winding by four turns. The main contribution of this research is the development of a combined diagnostic approach using instantaneous space phasor (ISP) spectral analysis and infrared thermal imaging to differentiate between load imbalances and incipient interturn short circuits in transformers. This method enhances early fault detection by identifying distinctive electrical and thermal signatures associated with each condition. The results could improve transformer monitoring, reducing the risk of failure and enhancing grid reliability.

1. Introduction

Transformers are static electrical machines that modify electrical energy parameters through electromagnetic induction, adapting the voltage levels to meet the requirements of the electrical grid that they are connected to. Transformers are fundamental machines, and their shutdown typically results in the outage of a part or the entirety of the electrical network [1].

An electrical transformer has a useful life of approximately 40 years, determined by its operating conditions. Overheating and overloading are factors that can reduce this lifespan [2,3]. These adverse conditions, combined with overvoltage issues, manufacturing defects, and poor insulation, can cause failures in the windings, tank, core, bushings, and other transformer components. According to statistics reported over the years, the most common transformer failures occur in the windings, with a failure rate between 30% and 40%, followed by tap changers, with reported failure rates ranging from 26% to 40%, and insulators, with rates between 14% and 30% [1].

Winding failures occur due to insulation degradation, winding deformation or displacement, and natural deterioration over time. These factors can lead to an incipient short-circuit fault, which may occur between turns of the low-voltage winding or the high-voltage winding. In any case, when a fault is incipient, it is very difficult to detect, as it appears suddenly and grows gradually. Consequently, classical detection methods are ineffective until the fault becomes significant, placing the transformer at imminent risk of total failure due to short circuit [3,4]. Currently, various methodologies are used for the detection of incipient short-circuit faults, including vibration signal analysis [5], which is useful in identifying mechanical problems in transformers, but its sensitivity is limited when it comes to detecting electrical faults, especially in the early stages. The analysis of active and reactive power under no-load conditions [6] and frequency response analysis [7,8] are two methodologies that are typically applied with the transformer completely offline, enabling initial detection but not monitoring during operation. Current signal analysis [9] often struggles to distinguish between similar phenomena, such as load imbalances and early-stage faults, particularly under noisy or variable network conditions. Infrared thermal imaging analysis [10], on its own, is a useful tool, but it requires the processing of many images to detect a fault, which entails high economic and computational costs. These faults have also been studied using different methodologies involving artificial intelligence techniques [8,11,12,13], although AI-based approaches require large volumes of labeled data and may not generalize well across different transformer types or operating conditions.

There is a large variety of methods for the detection of incipient turn-to-turn winding faults in transformers; however, few studies address how to distinguish between different types of faults or how to differentiate an inherent imbalance in the power grid or load from an incipient short-circuit fault. The authors of reference [14] study the distinction between internal faults and transient effects occurring in the network or load. Among the internal faults analyzed, incipient interturn short circuits are included, while the transient effects used for comparison include inrush currents, current transformer saturation, and the switching of nonlinear loads. The proposed method analyzes the waveform distortion caused by each condition, based on the duration of the event. Since these transients are brief compared to an incipient short circuit, the method can differentiate them based on their temporal profiles. The interturn short-circuit size considered in the study ranges from 20% to 60% of the winding, which corresponds to a severe short circuit, making it difficult to classify as incipient. The comparison between load imbalances and incipient interturn short circuits is also studied in [15]. Although the mentioned study is based on a three-phase permanent magnet synchronous generator, the behavior of the three-phase current signals is the same as in a transformer. Incipient interturn short-circuit faults involve 5, 10, and 15 turns in phase “a”, with an emphasis on the 15-turn short circuit, which represents 25% of the winding. To differentiate between incipient short circuits and load imbalances, a robust indicator, Δϕ₂, is used to discriminate faults by analyzing the phase difference between the voltage and current components. This indicator effectively distinguishes between imbalances and short circuits by detecting asymmetries in the phase angle, which tend to be more pronounced under incipient short-circuit fault conditions. This technique effectively distinguishes between incipient short circuits and load imbalances; however, it requires an external resistor to measure the voltage and current in the short-circuited loop, which is impractical under real operating conditions.

The main contribution of this study is the development of a combined fault detection approach using instantaneous space phasor (ISP) spectral analysis and infrared thermal imaging to differentiate between load imbalances and incipient interturn short circuits in transformers. This method enhances early fault detection by identifying distinctive electrical and thermal signatures associated with each condition. The results could improve transformer monitoring, reducing the risk of failure and enhancing grid reliability. Throughout the sections of this document, the behavior of the three-phase current signal is analyzed under load imbalance conditions and in the presence of incipient short-circuit faults of different magnitudes. Section 2 reviews the ISP, the tool used to transform the three-phase signal into a single signal for fault detection analysis. Section 3 details the methodology, describing the test bench setup, signal analysis, and thermal analysis. Section 4 discusses the results, and, finally, conclusions are presented in Section 5.

2. ISP for Fault Detection

The instantaneous space phasor (ISP), also known as the Park vector, is a tool used in power systems and has applications in fault detection for electrical machines and power systems [16]. The ISP represents three-phase voltages or currents in the complex plane, capturing the information of all three phases in a single phasor. Equation (1) presents the ISP of the current $\tilde{I}$.

$$\tilde{I}={\displaystyle \frac{2}{3}}\left(i_a+a{i}_b+a^2{i}_c\right),$$

(1)

where $a$ is the rotation operator $a=1\angle 120°$ and the three-phase currents are $i_a, { i}_b$, and $i_c$, with phase φ and angular frequency ω corresponding to a frequency $f$ as follows:

$$\begin{array}{c}{i}_a=Icos\left(\omega t+\phi \right); i_b=Icos\left(\omega t+\phi -120\right);\\ i_c=Icos\left(\omega t+\phi +120\right); \omega =2\pi f.\end{array}$$

The ISP can also be represented using the positive ${\tilde{I}}^{+}$, negative ${\tilde{I}}^{-}$, and zero-sequence ${\tilde{I}}^0$ components, as shown in Equation (2). This representation is widely used in power systems for fault and imbalance analysis.

$$\tilde{I}=\left({\tilde{I}}^{+}+{\tilde{I}}^{-}+{\tilde{I}}^0\right),$$

(2)

where

$$\begin{array}{c}{\tilde{I}}^{+}={\displaystyle \frac{2}{3}}\left({i_a}^{+}+{a{i}_b}^{+}+a^2{i_c}^{+}\right); {\tilde{I}}^{-}={\displaystyle \frac{2}{3}}\left({i_a}^{-}+{a{i}_b}^{-}+a^2{i_c}^{-}\right);\\ {\tilde{I}}^0={\displaystyle \frac{2}{3}}\left({i_a}^0+{a{i}_b}^0+a^2{i_c}^0\right).\end{array}$$

In a balanced three-phase system, the zero-sequence component is equal to zero. Therefore, under these conditions, the ISP can be expressed as shown in Equation (3):

$$\tilde{I}={\tilde{I}}^{+}+{\tilde{I}}^{-}.$$

(3)

By squaring the ISP, its magnitude can be separated as shown in Equation (4):

$${\tilde{I}}^2={{(\tilde{I}}^{+}+{\tilde{I}}^{-})}^2={{(I}^{+})}^2+{{(I}^{-})}^2+2{(I}^{+}{I}^{-})\cos \left(2\omega t+{\phi }^{+}+{\phi }^{-}\right).$$

(4)

In this way, it is found that the mean value of the ISP is given by Equation (5):

$$〈\tilde{I}〉=\sqrt{{{(I}^{+})}^2+{{(I}^{-})}^2}.$$

(5)

The mean value of the squared ISP is used to detect incipient short-circuit faults in the three-phase windings of rotating machines, but it can also be applied to three-phase power systems and static electrical machines such as transformers. The analysis of the mean value of the squared ISP can be performed using various signal analysis techniques, such as Wavelet and Hilbert–Huang transforms [17], as well as with a simple and practical tool like the FFT [18].

3. Materials and Methods

In this work, experiments were conducted using a purpose-built platform designed for fault detection in a three-phase transformer bank, which consisted of three single-phase dry-type transformers rated at 120 VA, 240/24 V, and 60 Hz. This setup offered flexibility to intervene individually in each phase, inject faults in a controlled manner, and observe the electrical and thermal behaviors for each transformer [19]. One of these transformers was modified to intentionally induce incipient interturn short-circuit faults with magnitudes of 2 and 4 short-circuited turns, as well as a 4-turn imbalance in the supply. Four transformer conditions were evaluated to analyze the differences between a supply imbalance and an incipient interturn short circuit: a healthy transformer (no fault), a transformer with a 2-turn short circuit, a transformer with a 4-turn short circuit, and a healthy transformer with a 4-turn imbalance.

3.1. Modified Transformer

One of the transformers in the three-phase transformer bank was modified in its secondary winding. The original transformer had a winding with 45 turns of a 19-gauge magnet wire. The transformer was rewound on the low-voltage side, replicating its initial characteristics but with added taps providing access at turns 4, 6, 8, 28, 30, and 32. Figure 1 shows a schematic of the secondary winding of the three-phase transformer bank, indicating the available taps in the phase A transformer.

The proper operation of the modified transformer was verified by reproducing the healthy state. The primary was supplied with a variable three-phase voltage source at 240 V, obtaining 24 V at the output. The measured impedance parameter was 0.16 + j3.3 Ω, with a deviation of 0.3% from its original value and 0.24% from the average impedance of the three original transformers, which was 0.156 + j3.308 Ω.

3.2. Experimental Platform

The three-phase transformer used consisted of three single-phase transformers, each with the same specifications: 240 V, 60 Hz, 240 VA. The transformer was powered through a delta connection using a controlled three-phase voltage source. The modified transformer was placed in phase A, while the transformers in phases B and C were connected normally. The transformer’s load was purely resistive, connected in a star configuration with 240 W, utilizing 100% of the transformer’s capacity. Thermal–magnetic protections were installed on both the transformer’s power supply and the load side, along with an additional thermal–magnetic protection that functioned as a switch to induce a short circuit between the taps of the modified transformer. The current waveforms in the three primary windings were acquired at a sampling frequency of 5 kHz. The RMS voltage and current values were continuously measured in both the primary and secondary windings, as well as the current flowing through the short-circuited coils. Figure 2 shows the test bench, highlighting the components involved in its construction, including the voltage source, measuring instruments, three-phase transformer bank, three-phase load, high- and low-voltage protection elements, and the switch used to short-circuit the taps.

The infrared image was generated by the FLIR Systems, Inc. model E60 camera(Wilsonville, Oregon, USA), which operated within a range of −20 °C to 650 °C (−4 °F to 1202 °F) and featured a true infrared thermograph with a resolution of 320 × 240 pixels (76,800 pixels). The camera was mounted to focus on the transformer’s secondary winding, as shown in Figure 3, and was used in this configuration for all experiments.

3.3. Transformer Operation Tests

The modified transformer was tested under four conditions: a healthy state, an imbalance after removing 4 turns from the winding, a 2-turn short circuit, and a 4-turn short circuit.

• Healthy State: In tests with the healthy transformer, the taps created on the winding remain isolated, and the load is supplied normally through the winding terminals.
• Imbalance: The load can be supplied by removing the initial 4 turns, feeding it with an incomplete winding from turn 4 to turn 45. This setup simulates an unbalanced load supply equivalent to the loss of 4 turns in the winding. This reduces the output voltage in that phase and generates an asymmetric current flow, comparable to what occurs during a real load imbalance. This technique allows the imbalance to be induced in a precise, repeatable, and measurable manner in terms of the number of turns removed, as in the case of incipient short circuits, where the fault magnitude is also defined by the number of turns involved.
• Incipient Interturn Short-Circuit Tests: Using the taps shown in Figure 1, short circuits were induced between 2 turns at tap connections 4–6, 6–8, 28–30, and 30–32 and between 4 turns at tap connections 4–8 and 28–32. These 2-turn and 4-turn short circuits were conducted as experiments and represented 4.4% and 8.8% of the total winding, respectively.

3.4. Application of Fault Detection Techniques

For all four experiments, the primary current waveforms were acquired as three vectors. These three signals were then processed using the Clarke transform to obtain the oscillation of the mean value of the ISP, which results in a single vector or discrete time signal [16]. This vector depends on the values of the three sinusoidal current signals; when the signals are balanced, the mean value of the ISP is a direct current component located at the maximum value of the three current signals. When the three-phase signals are unbalanced, the mean value of the ISP oscillates at a frequency of $2f$, centered on a direct current component determined by the degree of imbalance present. The mean value of the ISP was analyzed using the FFT to obtain the frequency spectrum of the ISP mean value signal, thus moving from a time signal to the frequency domain, where the healthy transformer would ideally show a spectrum with only the direct current component at the zero component, but, in the case of imbalance, the $2f$ frequency appears. For the analyses in this work, this frequency was expected for both imbalance and the incipient short-circuit faults analyzed. In this way, differences were identified in the frequency components and magnitudes for each experiment—imbalance and short circuits—compared to the behavior of the healthy transformer. The combination of these concepts serves as a fault detection tool [17,18].

On the other hand, it is possible to take advantage of the fact that, in a turn-to-turn short circuit, the current flowing through the shorted turns creates a low-impedance loop that causes heating in those turns due to the Joule effect. This results in a progressive and localized temperature increase in the fault area. In contrast, during an imbalance, the current remains within nominal thermal limits, resulting in a stable temperature profile over time [10]. Thus, when an increase in the $2f$ frequency component is detected, the transformer is observed using an infrared image, which shows a stable temperature over time for an imbalance and an increasing temperature for an incipient short circuit, with a rate of increase that depends on the severity of the fault.

4. Results

This section presents the results of the experiments conducted to distinguish between an incipient short-circuit fault and a load imbalance in the transformer using signal analysis and thermographic analysis. All current measurements were taken on the primary side of the three-phase transformer operating under nominal load conditions, while the winding modifications were made to the secondary side of the single-phase transformer in phase A. The current signals of the transformer with induced two-turn short circuits exhibited the same behavior in terms of amplitude, frequency, phase, and temperature, regardless of the tap combination used: 4–6, 6–8, 28–30, or 30–32. Similarly, the four-turn short circuits behaved identically when induced between taps 4 and 8 and taps 28 and 32.

4.1. Time-Domain Analysis of the ISP and Current Signal

The time-domain measurements of the current signals for the healthy transformer, where the taps of the single-phase transformer remained open, are shown in Figure 4. The three-phase current had an average amplitude of 3.57 A, and the mean value of the obtained ISP appeared almost as a straight line at the same current level. Both signals were displayed in the time domain. The mean value of the ISP, shown in blue, remained nearly linear, which is characteristic of a healthy transformer, with minimal deviations due to the inherent imbalance of the real machine. This initial condition served as a reference for comparison with subsequent conditions.

Figure 5 shows the three-phase current signal and the obtained mean value of the ISP, both in the time domain, corresponding to the transformer with an imbalance due to the removal of four turns from the winding connected to the load. In the three-phase signal, phases A and B decreased to 2.97 A, while phase C maintained the same magnitude at 3.57 A. The mean value of the ISP, shown in blue, lost the characteristic straight-line trend of a healthy transformer and took on a sinusoidal shape with an amplitude of 0.47 A at a frequency of 120 Hz.

The three-phase current signals and the mean value of the ISP corresponding to a two-turn short circuit in the transformer’s secondary winding are shown in Figure 6. In this case, phases A and B slightly increased to an amplitude of 4.01 A, while phase C remained unchanged at 3.57 A, as in the healthy state. Although the three-phase current signal changed, the mean value of the ISP remained similar to the signal resulting from the four-turn load imbalance, with an amplitude of 0.425 A and a frequency of 120 Hz.

The time-domain analysis of the mean value of the ISP shows that a short circuit involving only two turns produces the same effect on the signal as a four-turn load imbalance. The mean ISP value for the two-turn short circuit was 45 mA lower than that of the four-turn imbalance, representing a difference of 9.5%. Therefore, it can be concluded that incipient interturn short-circuit faults cannot be simulated by simply removing turns from the winding.

Figure 7 shows the three-phase current signals and the mean value of the ISP corresponding to a four-turn short circuit in the transformer’s secondary winding. It can be observed that the amplitudes of phases A and B increased to 5.13 A, while phase C remained at 3.57 A. These changes are reflected in the increase in the amplitude of the mean ISP value, which reaches 1.1 A.

When comparing the three-phase current signal and the mean ISP value for the four-turn short circuit in Figure 7 with the four-turn imbalance in Figure 5, an increase of 234% is observed. Additionally, when compared to the two-turn short circuit in Figure 6, the increase is 258%. These percentages highlight the difference between an incipient short circuit and an imbalance. However, this distinction is only evident when the cause is known; otherwise, it is not possible to determine whether the waveform change is due to an imbalance or an incipient short-circuit fault.

The results presented in this subsection were derived from observations performed on the three-phase current signal and the mean value of the ISP, both analyzed in the time domain.

4.2. Frequency Analysis

To further clarify the signal information, an analysis is conducted in the frequency domain. Only the mean value of the ISP is subjected to this analysis, as it is derived from the three-phase signals and inherently contains information from all three. For the analysis, the mean ISP signal is centered on the x-axis to remove the DC component that would otherwise appear in the FFT spectral analysis but is not relevant for fault detection.

A load imbalance or an incipient interturn short circuit induces changes in the magnitude and frequency of the ISP, as shown in Figure 5, Figure 6 and Figure 7. However, in a healthy and balanced transformer, the ISP appears almost as a ripple, as observed in Figure 4. Figure 8 shows the FFT analysis applied to the mean ISP signals of the currents for both the healthy transformer and the transformer with a four-turn load imbalance. A frequency component appears at 2f = 120 Hz with a magnitude of 0.36, representing a difference of 0.31 compared to the magnitude of the 2f frequency component in the mean ISP for the healthy current. The remaining spectrum does not show significant changes between the healthy transformer and the transformer with a load imbalance.

Figure 9 shows the spectrum obtained from the FFT analysis of the two-turn short-circuit signal, where the imbalance effect is again evident in the 2f = 120 Hz component, with a magnitude of 0.34. This represents a difference of 0.29 compared to the magnitude of the 2f frequency component in the mean ISP value for the healthy transformer. Similarly, to the load imbalance case, the rest of the spectrum remains unchanged, with no significant variations.

The FFT analysis of the mean ISP value for the transformer with a four-turn short circuit is shown in Figure 10. A significant increase in the magnitude of the 2f = 120 Hz frequency component can be observed, reaching a magnitude of 1.08, which represents a difference of 1.03 compared to the 2f frequency component in the mean ISP value of the healthy transformer. Additionally, in the four-turn short-circuit fault, a 4f = 240 Hz frequency component appears with a magnitude of 0.09. This component is not present when a four-turn load imbalance occurs, and its appearance can be used as an indicator to differentiate between an incipient interturn short circuit and a load imbalance.

4.3. Infrared Analysis

As a definitive complement in distinguishing between incipient short-circuit faults and load imbalances, infrared imaging is used to analyze the winding temperature under different operating conditions of the modified transformer. Figure 11 shows the reference frame of the infrared image for the transformer under analysis. In this image, the modified transformer is located at the center, with its secondary winding being the focus of the infrared analysis.

Figure 12 shows the infrared image of the modified transformer in a healthy state, operating at full load. The initial temperature for the experiments was 26 ± 1.2 °C. The image was taken after 30 min of operation, at which point the temperature began to stabilize at 65 ± 3 °C. The average temperature increase rate was approximately 1.3 °C per minute.

Figure 13 shows the infrared image of the modified transformer supplying the load with four fewer turns in the secondary winding, operating at full load. The image was taken after 30 min of operation, at which point the temperature began to stabilize at 65 ± 2 °C. By the end of the 30 min period, this temperature did not show any significant difference compared to the transformer operating without a fault, and the average temperature increase rate remained at 1.3 °C per minute.

Figure 14 shows the infrared image of the modified transformer with a two-turn short circuit in the secondary winding, operating at full load. The image was taken after 9 min of operation, at which point the protection was triggered with 48 A circulating in the short-circuited loop. The temperature reached 75 ± 4 °C, exceeding the normal operating range. The infrared analysis of the two-turn short circuit clearly highlights the difference between a load imbalance and an incipient short circuit. The average temperature increase rate was 5.4 °C per minute.

Figure 15 shows the infrared image of the modified transformer with a four-turn short circuit in the secondary winding, operating at full load. The image was taken after 2 min of operation, at which point the protection installed in the short-circuited loop was triggered with 95 A circulating, and the temperature reached 79 ± 4 °C, exceeding the normal operating temperature range. Notably, the transformer’s thermal–magnetic protection device did not activate. The average temperature increase rate was 26.7 °C per minute.

The infrared analysis of the two-turn short circuit clearly highlights the difference between a load imbalance and an incipient short circuit. When the number of short-circuited turns increased to four, the temperature rise became hazardous for the transformer’s operation.

Finally, Figure 16 shows the temperature increase trends for each operating condition of the modified transformer. It can be observed that the temperature rise trends for the healthy condition and the four-turn load imbalance follow the same pattern. Although the three-phase sinusoidal signal, the mean ISP value, and the FFT analysis reveal differences between these conditions, the temperature remains as if the transformer were in a healthy state. The temperature trends for the transformer under incipient interturn short-circuit conditions show a rapid increase. As the number of short-circuited turns increases, the temperature rise rate becomes more significant. Consequently, if the fault is not effectively detected, the transformer will eventually suffer winding destruction.

5. Discussion

According to the results obtained in this study, the mean value of the ISP and infrared thermography allow us to distinguish between small load imbalances and incipient interturn short-circuit faults in transformers. The experimental platform was implemented to reproduce intentional and controlled interturn short circuits without inducing imbalances through external resistances or removing turns to simulate the fault. The transformer bank used adequately represents the characteristic electrical behavior of three-phase transformers. This type of configuration allows for the reproduction of real phenomena such as load imbalances and incipient turn-to-turn faults under controlled and measurable conditions. While previous studies report interturn faults in transformers with severe short circuits affecting 20% to 60% of the winding, which can cause damage before being detected [14], other studies have used infrared thermography for short-circuit detection but relied on continuous temperature measurement, requiring continuous image analysis with a higher computational cost [10]. Additionally, some works focus on distinguishing incipient interturn short-circuit faults in rotating machines, where various sources of imbalance may exist [15].

This study demonstrates that incipient faults in a transformer can be detected from as little as 4.4% of the winding affected, corresponding to a two-turn short circuit, and that the change that this short circuit induces in the mean ISP signal of the three-phase currents is similar to that caused by a four-turn load imbalance, where the thermal effects are the key differentiating factor. A more severe fault, affecting 8.8% of the winding, is detectable using FFT analysis of the mean ISP, but, at this stage, the transformer is already experiencing the negative effects of a temperature increase rate of 26.7 °C per minute, whereas the four-turn load imbalance does not cause a significant temperature change compared to the healthy transformer. These results confirm that the mean value of the ISP and its spectral content are suitable indicators to capture the asymmetry and instability introduced by the fault. Additionally, the thermal image helps to confirm whether these electrical anomalies are due to the progressive development of a fault or to normal load conditions, thereby enhancing the diagnostic capabilities of the method. The ambient temperature and the cooling method of the transformers can influence the thermal evolution of an incipient short-circuit fault, but the appearance of the $2f$ component in the spectrum of the ISP mean value allows for the identification of the presence of a turn-to-turn short circuit, setting the basis for determining whether there is a significant temperature increase in the transformer. This complementarity between the electrical response and thermal behavior enhances the reliability of the detection, regardless of the transformer’s thermal dissipation capacity.

The effects of imbalances and short-circuit faults highlight the importance of implementing redundant fault detection, ensuring continuous monitoring for imbalance detection through the three-phase current signal and fault verification using infrared imaging to rule out or confirm an interturn short-circuit fault at an early stage.

The ability to distinguish between incipient interturn short circuits and small imbalances contributes to diagnostic and preventive maintenance tools, enhancing the reliability of industrial and power distribution systems in electrical grids. Additionally, this work sets the groundwork for further research on fault detection with lower failure percentages, testing additional fault detection and diagnostic techniques based on distinguishing between these fault types, as well as exploring the simultaneity of faults and imbalances and adding imbalances caused by capacitive loads and nonlinear loads. New wireless temperature sensor technologies can also be explored. Another direction that research can take is real-time implementation in non-laboratory conditions, with some application in industrial situations. In the future, scalability tests can be conducted using higher-capacity transformers.

6. Conclusions

The combination of fault detection techniques of different natures, such as signal analysis and thermography, provides effective redundancy in detection. In this work, this redundancy is achieved, and, additionally, the dual detection approach allows the identification of false positives when dealing with load imbalances. The mean value of the ISP, through FFT analysis, effectively detects both imbalances and faults but cannot distinguish between them when dealing with incipient faults.

It was found that the amplitude of the 2f frequency component in a two-turn short-circuit fault has a similar effect to that of a four-turn imbalance, with only a 5.55% difference in amplitude. However, the same frequency component in the four-turn short-circuit test showed a 300% increase compared to the four-turn imbalance test. The operating temperatures of the transformer under full load in two-turn and four-turn short-circuit conditions showed temperature increase rates of 5.4 °C per minute and 26.7 °C per minute, respectively, making this the key discriminating factor between an incipient short-circuit fault and a small imbalance.

It is important to note that fault detection or diagnostic techniques in experimental platforms or modeling approaches that introduce external resistances to create an imbalance or remove turns from the winding to simulate an imbalance cannot be considered equivalent to an incipient short circuit with the same number of removed turns. This would be an incorrect practice.

The effective discrimination of these types of faults is essential in ensuring the continuity of electrical service in power systems and can be achieved using basic signal analysis techniques combined with thermography.