Review of Various Sensor Technologies in Monitoring the Condition of Power Transformers

Abstract

Modern power grids are undergoing a significant transformation with the massive integration of renewable, decentralized, and electronically interfaced energy sources, alongside new digital and wireless communication technologies. This transition necessitates the widespread adoption of robust online diagnostic and monitoring tools. Sensors, known for their intuitive and smart capabilities, play a crucial role in efficient condition monitoring, aiding in the prediction of power outages and facilitating the digital twinning of power equipment. This review comprehensively analyzes various sensor technologies used for monitoring power transformers, focusing on the critical need for reliable and efficient fault detection. The study explores the application of fiber Bragg grating (FBG) sensors, optical fiber sensors, wireless sensing networks, chemical sensors, ultra-high-frequency (UHF) sensors, and piezoelectric sensors in detecting parameters such as partial discharges, core condition, temperature, and dissolved gases. Through an extensive literature review, the sensitivity, accuracy, and practical implementation challenges of these sensor technologies are evaluated. Significant advances in real-time monitoring capabilities and improved diagnostic precision are highlighted in the review. It also identifies key challenges such as environmental susceptibility and the long-term stability of sensors. By synthesizing the current research and methodologies, this paper provides valuable insights into the integration and optimization of sensor technologies for enhancing transformer condition monitoring and reliability in modern power systems.

1. Introduction

Electricity is a critical component affecting the economy and prosperity of any country, reflecting national growth. The ever-increasing demand for electricity combined with the aging infrastructure are two important challenges that affect the energy transition. In addition, climate change is increasingly causing extreme weather events, posing significant risks to physical assets. With the continuous rise in the global greenhouse gas (GHG) concentrations, climate change-related extreme weather events are expected to further increase in frequency, intensity, and duration. Since electrical power grids are amongst the critical infrastructures of our modern societies, it is crucially important to enhance condition monitoring tools. A healthy power system can meet accelerated power demands without compromising quality. Finding enhanced monitoring solutions is no longer a choice, but an obligation. High-voltage or power transformers, essential in electrical systems, are subjected to various operating stresses, and their failure can lead to significant power interruptions and economic repercussions [1].

Power transformers are among the most expensive equipment in power grids and are vital for reliable electricity delivery. As they age, the risk of power grid failure increases, making condition monitoring crucial in reducing unexpected downtime and enhance power availability. Effective condition monitoring must detect gradual or sudden deterioration and predict issues, allowing sufficient time for intervention before major failures occur. The continuous monitoring of changes in the electrical insulation system can prevent service-aged transformers from causing uncontrolled outages [2,3]. Traditionally, transformer maintenance relied on physical protection and minimal traditional techniques. However, modern requirements such as load demands, power quality, environmental constraints, and maintenance objectives have led to comprehensive (especially online) monitoring systems. Replacing transformers involves long-term planning and high costs, making extending their lifespan a beneficial investment [4,5].

Traditional condition monitoring methods include dissolved gas analysis (DGA) for fault identification, power factor testing for moisture and contamination detection, winding resistance testing for conductor losses and winding issues, winding ratio testing for defect identification, and thermography for overheating detection [6]. Modern monitoring systems employ various sensors—electrical, chemical, mechanical, acoustic, and optical—to track different parameters, necessitating extensive research in sensing technologies [7].

Recent review papers have highlighted the importance of various sensors used in transformer monitoring. For instance, ref. [8] discussed the broad applicability of fiber Bragg grating (FBG) sensors in detecting multiple issues in high-voltage assets. FBG sensors are used for monitoring temperature, partial discharges (PDs), oil assessment, and mechanical deformations in transformers. Another review [9] examined fiber optic sensors in power systems, emphasizing their use in detecting line faults and monitoring overheating. A review of polymer-based optical fiber sensors [10] highlighted their use in monitoring the refractive index (RI) of liquids, making them suitable for biochemical applications. Another review [11] on fiber optical (FO) sensors discussed their role in monitoring transformer oil conditions, enabling real-time, accurate assessments and predictive maintenance. Reviews have also explored advances in the online monitoring of transformer oil aging [12], emphasizing the benefits of non-destructive sensors over traditional offline methods. Similarly, Dongyan Zhao et al. [13] highlighted the advanced methods using FO and capacitive sensors for continuous moisture detection in transformer oil. FBG sensors’ effectiveness in detecting PD was analyzed in [14], noting their high sensitivity and immunity to electromagnetic interference. Additionally, ref. [15] reviewed ultra-high-frequency (UHF) sensors for PD detection, praising their high signal-to-noise ratio. Piezoelectric sensors have been highlighted for their cost-effectiveness and ease of installation for detecting acoustic emissions (AEs) generated by PD events, providing real-time monitoring and early fault detection [16]. The reviews have also discussed integrating fiber optic and DGA sensors for the real-time monitoring of transformer oil quality and fault detection, emphasizing their role in enhancing transformer reliability [17,18].

In addition to these insights, it is essential to consider advanced sensors that further enhance transformer monitoring. This review paper begins by discussing the current state of sensor technologies in use in transformer monitoring, covering both traditional and modern methods and the need for advanced systems. It then examines various sensor types, including electrical, chemical, mechanical, acoustic, and optical sensors, highlighting their roles in monitoring transformer parameters. Following this, the paper explores the practical applications of these sensors in detecting faults such as PDs, core conditions, temperature, and dissolved gases, supported by case studies. It identifies the limitations and challenges of current sensor technologies, including environmental susceptibility, long-term stability, and data interpretation issues. This review also discusses potential advances and future research directions. Finally, it summarizes the key findings, emphasizing the importance of advanced sensor technologies in ensuring reliable and efficient power transformers.

2. Background

Currently, sensors constitute a significant share in power system protection. When used in real-time monitoring, they help to avoid or to some extent minimize the impact of service outages. These enable a condition-based model of maintenance at a much lower cost compared to those of the traditional schedules. The sensors collect raw information about faults, which is very helpful in the control system. The data and parameters collected by sensors are available for categorization and prediction; and in turn, enable more accurate predictive and corrective analyses [8,19]. To better understand the maintenance strategies employed for power transformers, the following flowchart (Figure 1) delineates the decision-making process for determining the appropriate maintenance approach based on failure occurrence and condition assessment.

The following table (

Table 1. Conventional condition monitoring methods for transformers [6,7,12,19,22,23].

Types	Method	Description
Electrical Tests	Dissipation or Power Factor Testing	Measures the dissipation factor (tan δ) to indicate contamination and deterioration of the insulating fluid. Conducted by applying an AC voltage and measuring the resulting current, in line with IEC 61620 (or 60247 under DC [24]) [25] and ASTM 924 [26] standards.
Electrical Tests	Frequency Response Analyses (FRA)	FRA is a non-intrusive monitoring and diagnostic technique allowing the assessment of the mechanical integrity and validating the structural design. It involves measuring the impedance of transformer windings over a frequency range (usually from a few Hz up to a couple MHz). Interpretations are performed according to standard DL/T 911-2016 [27] or technical reports such as CIGRE [28], IEC [29] and IEEE C57.149-2012 [30].
Electrical Tests	Leakage Reactance Test	Measures the leakage reactance of transformer windings to detect mechanical displacement or deformation according to the CIGRE 445 standard [31].
Electrical Tests	Insulation Resistance Test	Measures the resistance of the transformer’s insulation to detect deterioration and contamination [32], as outlined in the CIGRE Guide for Transformer Maintenance [31].
Electrical Tests	Ratio and Polarity Test	Confirms the correct winding ratio and polarity of the transformer windings, in compliance with IEEE Std C57.12.90-2010 [33].
Electrical Tests	Impulse Tests	Assesses the ability of the transformer insulation to withstand high-voltage surges in accordance with IEEE C57.12.90 [33] and IEC 60076-3:2013 [34] standards.
Electrical Tests	Polarization Depolarization Current Measurement	Assesses the condition of the transformer’s insulation [2] based on CIGRE 254 Dielectric response methods for diagnostics of power transformers [35].
Electrical Tests	Recovery Voltage Method	Assesses the condition of the transformer’s insulation by measuring the recovery voltage after applying a direct current voltage, as specified in CIGRE 254 and the CIGRE Guide for Transformer Maintenance [31,35].
Electrical Tests	Frequency Domain Spectroscopy	Analyzes the frequency response of transformer insulation to assess its condition (moisture content and aging). Conducted by applying a range of AC frequencies and measuring the dielectric response, as outlined in IEEE C57.161-2018 [36].
Electrical Tests	Absorption Ratio/Polarization Index Test	Evaluates the condition of the transformer insulation [37] as specified in the CIGRE Guide for Transformer Maintenance [31] and IEEE Std 62-1995 [37].
Electrical Tests	Breakdown Voltage Test	Measures the voltage at which the insulation becomes conductive, indicating its dielectric strength. This test is conducted by gradually increasing the voltage until breakdown occurs, following ASTM D1816-12 [38], ASTM D877 [39], and IEC 60156 [40] standards.
Chemical Tests	Dissolved Gas Analysis	Diagnoses incipient failures by detecting dissolved gases like hydrogen, carbon monoxide, carbon dioxide, ethylene, acetylene, etc. This analysis involves extracting oil samples and analyzing gas content using gas chromatography as per ASTM D3612 [41], D3284 [42], and IEC 60567 [43] standards. Interpretation guidelines for gas can be found in technical references such as IEEE C57.104-2019 [44] and CIGRE brochure 443 [45].
Chemical Tests	Furan Analysis	Analyzes furan compounds in the transformer oil to assess the condition of the paper insulation [46], following IEEE Std C57.156-2016 [47], ASTM D5837 [46], and IEC 61198 [48].
Chemical Tests	Moisture content	Measures water content in insulating liquids. This is performed using Karl Fischer titration, which quantifies moisture content as per ASTM D1533 [49] and CEI 60814 [50].
Chemical Tests	Fourier Transform Infrared Spectroscopy	Identifies chemical changes in the insulation by analyzing its spectral fingerprints [51], as outlined in ASTM E2412-10 standard [52].
Chemical Tests	Total Acid Number	Measures the acid concentration in the oil, which increases with aging. The test involves titrating the oil with a base and determining the acid number, as specified by ASTM D974 [53] and IEC 62021 [54].
Chemical Tests	Interfacial Tension	Monitor the presence of polar compounds as per ASTM D971 [55] and ISO 6295 [56]. For in-service fluids, a decrease in this value indicates an increase in the concentration of contaminants, including oxidation by-products [55].
Chemical Tests	Dissolved Metals Analysis	Detects metals dissolved in the oil to identify wear and tear of transformer components. This is typically conducted using atomic absorption spectroscopy or inductively coupled plasma analysis according to ASTM D7151 standard [57].
Chemical Tests	Color/Visual Examination	This involves visually examining an oil sample by passing a beam of light through it to determine transparency and identify foreign matter. Poor transparency, cloudiness, or observation of particles indicates contamination [3] as specified in ASTM D974-22 standard [58] and STM D1500 [59].
Chemical Tests	Inhibitor Content	Measures the content of oxidation inhibitors in the oil. Inhibitors protect the oil from oxidation and extend its life. Common inhibitors include 2,6-di-tert-butyl-paracresol and 2,6-di-tert-butyl-phenol [60], following IEC 60666 Standard [61].
Chemical Tests	Corrosive Sulfur	Measures the presence of corrosive sulfur compounds in the oil, which can corrode metal surfaces and reduce the electrical strength of conductor insulation as per ASTM D2864-10e1 standard [62], IEC 62535 [63], and ASTM D1275 [64].
Chemical Tests	Particle Count	This measures the number and size of particles in the oil. Particles can significantly affect the dielectric strength of insulating liquids and increase the risk of static electrification, partial discharge activity, and tracking [3] as outlined in CIGRE 157 [65], ASTM D6786 [66], and IEC 60422 standard [67].
Chemical Tests	Turbidity Analysis	Measures the cloudiness of a liquid caused by suspended solids, indicating contamination levels. Higher turbidity signifies more suspended particles, affecting the fluid’s insulating properties and cooling efficiency as specified in ASTM D6181 [68].
Physicochemical Tests	Photoluminescence and Ultraviolet-Visible (UV-Vis) Spectroscopy	Assesses the optical properties of the oil as per ASTM D6802-02 [69].
Thermal Tests	Thermography	Detects overheating and pinpoints potential faults through temperature measurements. Infrared cameras are used to capture thermal images of the transformer, identifying hot spots as per ISO 18434-1:2008 [70], ASTM D1903-01 [71], and IEEE C57.156 [47].
Thermal Tests	Heat Transfer Properties	Evaluates thermal conductivity, specific heat, viscosity, pour point, and relative density to determine cooling efficiency. Improves heat transfer, while low viscosity aids in better flow. The pour point indicates the lowest temperature for oil flow, following ASTM D2717-95 [72].
Structural and Mechanical Assessments	Bushing Monitoring	Assesses the condition of transformer bushings. This involves measuring parameters like capacitance and power factor to detect insulation degradation, following IEEE C57.19.100-2012 standards [73] and IEC 60137 [74].
Structural and Mechanical Assessments	OLTC Monitoring	Monitors the performance and condition of on-load tap changers. This method involves measuring parameters such as operation times and contact wear, as specified by IEC 60214-1 [75] and IEE Std C57.137 [76].
Comprehensive Condition Assessments	Condition Assessment and Diagnostics	General guidelines for diagnostic testing of transformers. This comprehensive approach involves a range of electrical, chemical, and physical tests to assess transformer health, guided by IEEE 62-1995 [37], IEEE C57.143-2012 standards [77], IEC 60599 [78], and IEEE Std C57.152-2013 [79].

) outlines the conventional methods used for monitoring the condition of transformers:

The condition monitoring methods discussed in the table above are closely aligned with current industrial standards such as IEEE, IEC, ASTM, and ISO. These standards ensure the accuracy, reliability, and effectiveness of the monitoring techniques used. Section 3 will elaborate on how each of these standardized methods is applied using various sensor technologies to monitor the condition of power transformers. This section will provide a comprehensive overview of their implementation and benefits in real-world applications, highlighting how these methods can facilitate and expedite these procedures. It is important to note that despite advances and improvements, there are still many challenges that these new inventions and enhancements have not fully addressed, which will be discussed further. While utilities have amassed considerable experience in transformer maintenance, the same cannot be said for sensing devices. Experience is growing alongside the increasing number of installations, operational time, and advances in technology. Sensors are being installed both in new and in-service transformers. For aged units, most sensors have been retrofitted onto units identified with incipient faults or due to their criticality within the system.

Table 2. Faults and parameters to be assessed for the main components [7,17].

Components	Description	Online Monitoring Possibilities
Active part	The core is prone to defects like displacement of blades due to electromagnetic forces and high eddy currents, causing mechanical deformation and efficiency loss. Failures due to mechanical, thermal, or dielectric stresses. Mechanical anomalies include loosening, displacement, or deformation due to improper maintenance, corrosion, or vibrations. Thermal stresses create “hot spots” that damage the windings. Dielectric anomalies arise from disruptions in the insulating material, leading to short circuits and local burning. The insulation degrades over time through hydrolysis, pyrolysis, and oxidation, reducing dielectric and mechanical properties.	Transformer load current, Core ground current primary/secondary/tertiary voltages Winding temperature, Short-circuit current of the transformer, Peak voltage of the transformer surge Top oil temperature Moisture (and temperature) in oil tank PD measurements Dissolved gases in oil
Tank, oil containment, and preservation	Oil leaks, corrosion, condensation, aging and degradation, welding defects, buckling/deformation, vibration-induced issues, faulty breather or pressure relief devices are some of potential faults of transformer tanks.	Oil level in the tank Sudden pressure Moisture and temperature in oil Ambient moisture
Bushings	Degraded due to contamination, water ingress, and aging, leading to PDs and overheating.	Capacitance and power/Dissipation factor, Leakage current, Bushing voltage from capacitive coupler
Tap-Changer	Failures due to mechanical wear, lack of maintenance, and electrical issues, disrupting voltage regulation.	Motor driving current and vibration signals, Current accumulated in individual taps, Tap position indicator, AC supply voltage, number of accumulated changes on each tap, Total number of operations of the OLTC, RMS phase-to-earth transformer voltage, OLTC oil level, OLTC oil temperature, Gas content in insulating oil, Moisture content in the OLTC oil.
Cooling System	Wears and tears may lead to cracks and oil leaks. Cooling system failures due to pump or fan malfunctions may cause overheating and increased pressure.	Oil pump motor current, Cooling system AC supply voltage status of the oil pumps (on/off), Transformer load current, Fan motor currents, Ambient temperature, Winding temperature (thermal imaging), Top oil temperature, Bottom oil temperature

details the critical components of the transformer and the associated faults that need to be assessed for effective transformer maintenance and operation. This table also includes actual sensing techniques available in industries [44,77,80,81]. The CIGRE technical brochure 343 [31] focuses on how currently existing industrial standards rely on these sensors. Both users and manufacturers generally agree that the current market state of condition monitoring systems supports the idea that standardizing facilities for later fitting a monitoring system on a new transformer would offer practical and economic advantages.

Recent advances in sensor technologies offer promising solutions for enhancing transformer condition monitoring. Effective in monitoring critical transformer parameters, FBG sensors measure hot-spot temperatures, detect AEs from PD events, and accurately gauge dissolved gases and moisture levels. These sensors are immune to electromagnetic interference, are lightweight, and capable of providing real-time data. Additionally, using techniques like Mach–Zehnder interferometry and optical spectroscopy, these sensors can measure transformer winding temperatures and detect changes in the oil’s RI, providing valuable insights into oil degradation. These capabilities make them suitable for high-temperature, high-pressure environments and crucial for maintaining transformer health [8,9,10,11]. Chemical sensors, including semiconductor-based gas sensors, electrochemical-based gas sensors, optical-based gas sensors, and field-effect gas sensors, were proposed to detect gas emissions indicative of faults. Optical chemosensors and surface plasmon resonance sensors detect oil degradation. Methane and acetylene are detected using metal oxide-based sensors and FBGs [19,82]. Wireless sensing networks like ultrasonic gas sensor nodes with ZigBee transceivers are used for remote monitoring. Humidity sensors, including capacitive and FBG types, were also proposed to detect the moisture in transformer oil. Acoustic sensors, like hydrophones and piezoelectric accelerometers, diagnose on-load tap changer conditions [19,83,84,85,86]. UHF Sensors capture electromagnetic waves from PD sources, providing high detection sensitivity and immunity to external noise, making them suitable for on-site monitoring. Various UHF sensor designs, including monopole, microstrip, fractal, and ultra-wideband antennas, offer specific advantages in bandwidth, size, radiation pattern, and sensitivity [15,87]. In other contexts, piezoelectric sensors convert mechanical stress into an electrical signal, making them ideal for detecting AEs from PDs. They offer high sensitivity, wide bandwidth, and cost-effectiveness, making them suitable for widespread deployment [16,88].

3. Sensor Technologies for Measuring Different Parameters

The diagram in Figure 2 illustrates various types of sensors utilized in transformer monitoring, including electrical sensors (current, voltage, and partial discharge), gas sensors (chemical and fiber Bragg grating), temperature sensors, moisture sensors, and oil-level sensors, each playing a crucial role in ensuring the efficient operation and reliability of power transformers.

3.1. Core Sensors

Monitoring the core of transformers is essential for ensuring their reliability and performance. The core’s integrity significantly impacts the transformer’s efficiency and operational stability. Advanced sensor technologies play a pivotal role in providing detailed insights into the core’s condition.

In [90], a triaxial strain gauge was employed to detect and measure magnetostrictive strain in transformer cores influenced by magnetic fields. The findings indicated significant strain variation across different core regions, particularly near the magnetic field source. This detailed and localized monitoring is crucial for assessing vibration deformation and enhancing transformer reliability. However, the study did not address the potential impacts of environmental factors like temperature and mechanical stress on sensor accuracy and reliability. Another innovative approach [91] involved using a “bendductor” sensor made from electrical steel to measure magneto–mechanical forces in transformers under bending loads. This sensor demonstrated high sensitivity, a good signal-to-noise ratio, and reduced hysteresis errors compared to previous sensors. The sensor was tested with different loads (0–500 g) and both sinusoidal and triangular waveforms at various frequencies, showcasing a measurement range of 1–5 N with an accuracy of 0.05 N and a total uncertainty of 0.11 N. This precision in force measurement is particularly beneficial for transformer applications [92,93]. Furthermore, the researchers in [94] explored the use of multichannel AE detection by combining fiber optic and lead zirconate titanate (PZT) sensors for PD monitoring in the core. This system was tested on a tank full of water. The approach consists of using a sparse network and measuring the time of arrival (TOA) of the AE signals, to locate PD sources with high accuracy. The integration of LabVIEW for real-time processing and MATLAB for localization algorithms allowed the system to perform efficiently under noisy conditions by achieving a mean localization error of less than 10 mm for 1% mean error in delay. LabVIEW in this hybrid system was used to capture signals from the AE and FO sensors in real-time to perform denoising through techniques such as fast Fourier transform and digital filtering to ensure clean data. Also, MATLAB was utilized for advanced signal processing, including TOA calculation via cross-correlation, trilateration, and particle swarm optimization (PSO) for precise localization. The PSO algorithm provided effective localization, where errors were primarily determined by TOA inaccuracies at the tank borders, demonstrating that the system could achieve efficient localization under noisy conditions. It should be noted that the PSO algorithm has provided effective localization where the errors were primarily determined by the TOA inaccuracies at the tank borders. For detecting full discharges (FD), another study [95] developed Rogowski coils with ferrite and Teflon cores. These coils were tested on a 30-kVA transformer and demonstrated effective FD detection, with the Teflon coil offering better noise immunity and lower production costs. The ferrite coil showed a higher sensitivity in the 0–1 MHz range, while the Teflon coil was more sensitive in the 5–10 MHz range. This makes the Teflon coil a promising option for non-destructive detection and more suitable for noisy environments testing and monitoring power transformers. To enhance the current measurement accuracy, Yuwen Wu et al. [96] constructed a novel current transformer (CT) using a virtual air gap (VAGCT) to address core saturation issues. The VAGCT demonstrated accurate current measurements across various conditions and provided high DC immunity and a wide dynamic range. This innovative design significantly enhances reliability in power systems.

In summary, advanced sensor technologies are crucial for monitoring transformer cores, offering precise, real-time, and localized data that enhance the understanding of core conditions. Despite these advances, challenges remain, such as environmental susceptibility of strain gauges, the need for further validation of the “bendductor” sensor, the complexity and cost of multichannel AE systems, the long-term reliability of Rogowski coils, and the optimal placement of piezoelectric accelerometers. Future research should prioritize addressing these gaps to enhance sensor robustness and practicality, thereby ensuring the operational health and longevity of power transformers.

3.2. Winding Sensors

Advanced sensor technologies for transformer windings play a crucial role in improving the monitoring and reliability of power transformers. Various sensors have been developed to measure different parameters and address specific challenges associated with transformer windings.

The researchers in [97] developed a novel built-in capacitive fast transient (FT) sensor for measuring transient terminal voltages at transformer windings, featuring a dielectric window (DW) and a capacitive voltage divider. The sensor, installed inside the bushing mounting of a 500 kV transformer, utilizes a dielectric window made from Nylatron MC 907 (Mitsubishi Chemical Advanced Materials) and surface-mounted capacitors for the low-voltage arm to ensure excellent performance. The experimental validation involved calibration tests with air, DW, and transformer oil, followed by switching and lightning impulse tests using a high-voltage impulse generator and field applications at a 500 kV pumped storage power plant. The FT sensor demonstrated high sensitivity and accuracy in measuring transient waveforms with minimal interference, effectively detecting high-frequency oscillations and transient overvoltages during field tests. In this field, Rathnayaka et al. [98] introduced an on-line impedance monitoring technique for transformers based on an inductive coupling approach. This novel method used inductive coils to measure the impedance of transformers without direct electrical connections to high-voltage components. The primary coil injects a high-frequency signal while the secondary coil measures the induced voltage by enabling impedance calculation. The experiments validated the method’s high sensitivity and accuracy in real-time fault detection, marking a significant advancement in non-invasive transformer monitoring. Wang et al. [99] also invented a capacitive voltage sensor array to measure transient voltage distribution in transformer windings. This non-invasive method employed stray capacitance between the sensing electrode and transformer discs, utilizing an RC integrator and impedance adapter to enhance performance. The system depicted high accuracy in capturing transient events, providing valuable insights into voltage gradients and potential insulation stress points without affecting the winding’s electrical properties. Song et al. [100] utilized commercially available fluorescent FO temperature sensors to monitor transformer winding temperature. One sensor was placed at the hotspot of the winding, and five other sensors were placed at critical points for verification. This method demonstrated high accuracy, with an error of less than 4 K between the measured and estimated temperatures and an average error between 0.24 K and 1.93 K. The real-time calculation process was efficient, taking only 0.38 s. The study identified optimal sensor placements using a sparse placement strategy based on proper orthogonal decomposition modes and QR factorization using the column pivoting (QR-pivot) method by minimizing the condition number to ensure stable and accurate temperature estimation. Recently, a novel coil model with innovative winding arrangements and an interference compensation scheme was designed [101] to measure high-amplitude currents with improved sensitivity and reduced errors. These coils, featuring wooden cores and concentric holes for windings, an outer conducting layer, and an Op-amp circuit for interference mitigation, showed up to 4.01 times higher sensitivity and 77.5% less non-linearity, and significant reductions in peak and rise times compared to conventional models. The interference compensation scheme reduced error by 78.79% with a 30 A interfering current. However, further field validation, long-term stability testing, and economic feasibility assessments are needed. The authors in [102] introduced novel airgap profiles and a distributed airgap arrangement to reduce losses in high-frequency current transformer (HFCT) sensors. Zigzag and stepped airgap profiles and a distributed airgap arrangement were designed to minimize fringing flux, eddy currents, and thermal losses. Using 3D finite element analysis, these designs reduced winding losses from 2.8 W to 1.7 W (zigzag) and 1.3 W (stepped) and decreased winding conductor temperatures from 44.8 °C to 32.0 °C (zigzag) and 27.8 °C (stepped). The stepped airgap profile also improved the induced voltage in the core. Finally, ref. [103] presents the design of a Fabry–Pérot (F-P) FO sensor array for detecting and localizing dual PD sources within transformer windings. The sensor, comprising a single-mode fiber, silica tube, and silica diaphragm, detects ultrasonic signals generated by PDs through changes in the F-P cavity’s interference pattern. Tested on a 35-kV single-phase transformer, the sensor array achieved high localization accuracy with errors within 18 cm using the MUSIC algorithm. While promising, further field validation, long-term stability testing, and assessments of integration, scalability, and economic feasibility are needed to confirm its practical applicability. Addressing these gaps will enhance the adoption of F-P sensor arrays for PD detection and localization in transformer windings.

In summary, by incorporating these advanced sensors, these technologies for transformer windings show significant promise in enhancing monitoring accuracy and efficiency. However, several challenges, such as long-term stability, scalability, etc., still remain.

3.3. Vibration Sensors

The research into vibration sensors for transformer monitoring has shown significant advancements, particularly with fiber optic technologies, emphasizing their precision and sensitivity to changes in load current caused by Lorentz forces and magnetostriction. For instance, a study [104] demonstrated the potential of fiber optic sensors for the real-time monitoring of transformer vibrations, aligning with the existing literature on their diagnostic value (Figure 3).

In another study [105], FBG sensors, packaged with polyether ether ketone (PEEK) and transformer board (TB) materials, were tested for their sensitivity and response time under a hydraulic press for static tests and a free-fall impact setup for dynamic tests in high-voltage winding transformers during events such as short-circuit and inrush current. Their results identified that the PEEK sensors showed superior sensitivity (2.39 pm/N, 5.61 times higher than the TB sensor), linear response, and high repeatability, though further real-world testing and temperature compensation are recommended.

The system’s capability to measure vibration using an integrated F-P sensor was highlighted in [106]. Utilizing frequency-modulated phase-shifting interferometry, the system achieved fast phase demodulation with a sampling rate of 100 kHz, effectively measuring vibrations up to 10 kHz with a signal-to-noise ratio of approximately 72.26 dB. This indicates high sensitivity and accuracy in vibration detection, suitable for the real-time transformer monitoring. An embedded FBG sensor in a carbon fiber-reinforced polymer (CFRP) clip for vibration measurement in power transformer iron cores was reported in [107]. The FBG strain sensor integrated into the spring-shaped CFRP clip allows the deformation caused by core vibrations to be measured, with the clip amplifying the vibrations for enhanced sensitivity. The experimental results demonstrated that the sensors accurately detected vibration frequencies up to 500 Hz and showed a 19% sensitivity improvement compared to unencapsulated sensors. This setup effectively measured fundamental and harmonic frequencies under various load conditions, proving its capability for precise vibration monitoring in transformer iron cores. In a similar field, ref. [108] investigated using FBG sensors designed by a 3D-printed cantilever structure to detect vibrations by measuring strain-induced changes in the FO’s RI. The sensors were installed inside a tight box that simulated the conditions of the dry power transformer and were tested under both dry and oil-filled conditions to simulate the environment of a transformer that is very sensitive to different frequencies and accelerations. The FBG sensors exhibited a linear response to acceleration, with a high correlation coefficient (R² = 0.995) and could discriminate between vertical and horizontal vibrations. However, the oil acted as a damper, reducing vibration amplitude, and the acrylonitrile butadiene styrene material deformed at higher temperatures. Thus, materials with higher thermal stability are needed for long-lasting applications.

Finally, an optimization ensured the high sensitivity and accurate detection of arc faults in low-voltage systems. A differential HFCT sensor was designed [109] with a differential threading method for the primary conductors and a centralized distribution pattern for the secondary windings for enhancing sensitivity and minimizing waveform distortion. It also included using an Ni-Zn ferrite core to ensure effective high-frequency signal detection. It is crucial to highlight that the flow electrification phenomenon can occur under unfavorable conditions in forced oil circulation systems. This phenomenon generates electrostatic charges that may accumulate and discharge, posing a potential risk to the solid insulation or components within transformers [110]. This presents a significant challenge for sensors, as the generated electrostatic charges can have adverse effects on their operation and longevity. So, despite those advances, future work should focus on improving long-term stability, optimizing sensor materials for thermal resilience, and validating sensor performance.

3.4. Temperature Sensors

RTD, thermocouples, Pt100, FO, and infrared are the most-used temperature sensors of transformers. Advanced transformer monitoring uses either fluorescing-tipped, or gallium arsenide (GaAs)-tipped fiber optic probes. Pt100 sensors are widely used in dry and cast resin transformers owing to their good accuracy and cost-effectiveness. Advanced FO probes with fluorescing or GaAs tips offer high precision and immunity to electromagnetic interference. These systems display real-time temperatures on digital monitors and can communicate data remotely through RS-485 or Ethernet by integrating into systems like SCADA. For preventing winding degradation, proper calibration and settings ensure temperatures remain below 220 °C. The Pt100 sensors have standard resistance values that simplify calibration and establish accurate monitoring and protection of transformer winding. The authors of [111] produced a temperature sensor by preparing Er3+/Yb3+ co-doped tellurite fluorescent glasses using the high-temperature melt-quenching method. They optimized the luminescence properties by adjusting the concentrations of rare earth ions and then fabricated a no-core tellurite optical fiber through mold-casting and fiber-stretching techniques. The sensor was attached between the coil and the core inside the transformer to monitor temperature changes online. An infrared thermometer was used as a reference thermometer. The results of the sensor showed excellent stability with a maximum measurement error within ±0.8 K and a fast response time of 2.1 s, proving its effectiveness for thermal monitoring of transformers. The distributed temperature sensors using the principles of Rayleigh, Raman, and Brillouin scattering have become widely applied techniques for measuring temperature. Raman scattering is particularly suitable for temperature measurement due to its wide bandwidth and sensitivity to temperature changes [112,113]. A Brillouin scattering process is induced by an acoustic wave, and corelates effectively to temperature and strain. Raman scattering with phonon participation is appropriate for temperature measurement; it features a wide bandwidth and is sensitive to temperature changes. Optical probes are very commonly employed in temperature monitoring because they provide a fast and accurate measurement. Over the past two decades, the direct measurement of hotspot temperatures based on fiber-optic temperature sensors by users and manufacturers of transformers has been popular [113,114]. Recently, temperature differences in insulation liquids were measured by fiber-optic-based sensors. These sensors were placed in the HV and LV windings to assess heat to be dissipated [115]. It should be mentioned that FO sensor temperature readings are comparable to traditional thermocouples, albeit with slight variances due to sensor placement [104]. However, further research is essential to explore aging effects, and by-products on sensor performance, and to ensure long-term durability under diverse operating conditions. Despite this, a fiber-optic-based transformer temperature sensor described in [115] demonstrated high precision (±0.5 °C) and a very short response time (1 s). The challenges faced by the system include the mechanical robustness of the sensor in the severe conditions of transformers, the complexity of the installation, and the dependence on the stability of the reference liquids requiring regular recalibrations. In addition, though insensitive to electromagnetic interference, the challenge of reliability in adverse conditions must be improved using fiber durability and system simplification. In another development, FBG-based sensors were embedded in a CFRP clip to monitor temperature. Encapsulated in a stainless-steel tube to isolate them from mechanical strain, these sensors demonstrated high precision with a temperature coefficient of 9.83 pm/°C, closely matching the data from an optical pyrometer [107]. This system effectively compensates for the thermal effects on the strain sensor, ensuring reliable temperature measurements in power transformer iron cores.

Importantly, FOs need to be compatible with transformer oil and must work stably under high temperatures. A distributed fiber optic sensor (DFOS) using Raman scattering was designed for temperature measurement in a 35 kV power transformer [116].

This self-produced sensor (Figure 4) was integrated with the transformer winding wire and used ethylene tetrafluoroethylene (ETFE) and polyimide as protective materials. By measuring the intensity of backscattered Stokes and anti-Stokes light, the DFOS provided temperature readings with a high accuracy. Conventional methods often have significant monitoring blind spots, particularly in complex internal environments where different sensor placements can yield varying results. Specifically, the layered winding structure of transformers is mentioned, where not all winding wires have oil channels between them. This arrangement can potentially disrupt the uniformity and continuity of temperature distribution, leading to the masking of hot spots. So, the DFOS was embedded in the transformer during manufacturing, providing a spatiotemporally continuous temperature measurement across the entire transformer and demonstrating effective real-time temperature monitoring along the windings. They employed a Gaussian convolution method to improve the detection accuracy of distributed fiber optic sensing (DFOS). Gaussian convolution aids in smoothing the data and reducing background noise, thereby enabling the precise localization of hot spots within the transformer. This approach notably enhanced temperature detection accuracy to ±0.2 °C and spatial positioning accuracy to 0.8 m. The results revealed that hotspots were around 90% of the winding height, suggesting a revision of the traditional hotspot location beliefs and the need for additional protection at these points. In this field, a similar observation, reported in [117], utilized a general DFOS system to monitor temperature distribution and hotspots in power transformers. The DFOS system, consisting of FOs with an ETFE sheath, was integrated into the transformer windings and validated through calibration with thermocouples and finite element method simulations. Testing on a 35 kV oil-immersed transformer showed the system’s high sensitivity and accuracy, effectively tracing hotspots and providing continuous temperature data. In addition to the Gaussian convolution method for noise reduction and enhancing detection accuracy, the authors also explored the use of various thermodynamic models to improve hotspot detection. They conducted a comparative analysis of four typical thermal models (IEC, Swift, Susa, and IEEE) against fiber-optical data to assess their performance. Comparison with thermal models identified the Susa model as the most accurate for predicting hotspot temperatures. Furthermore, field validation, long-term stability assessments, and evaluations of integration, scalability, and economic feasibility are needed to confirm the DFOS system’s practical applicability in real-world conditions. As a good case study, ref. [118] utilized ten FO sensors installed at various critical points within the transformer windings to measure instantaneous temperatures, specifically targeting hotspot temperatures. These sensors provided detailed empirical data, which validated the proposed non-uniform 3D CFD-based thermal analysis model by showing an error margin of only 0.11% for HST and less than 0.65% for top-oil and bottom-oil temperatures. To add to this, a novel programmable spectrum acquisition system for FBG and Fabry–Pérot sensors was presented in [106]. This system facilitated the simultaneous dynamic and static multiparameter measurements, including temperature. The FBG sensor measured temperature changes with a sensitivity of −1.4 GHz/°C, equivalent to about 11 pm/°C. The experimental results demonstrated high sensitivity and accuracy in temperature detection, showcasing the system’s potential for real-time temperature monitoring in power transformers. Beyond that, ref. [119] demonstrated that π-phase-shifted FBG (PS-FBG) sensors exhibited significantly higher sensitivity (17.5 times more) compared to traditional piezoelectric ceramic (PZT) sensors [120,121]. This research demonstrates that unlike PZT sensors, PS-FBG sensors can be directly installed in the fuel tank due to their superior insulation properties and high environmental resilience. These sensors successfully perform real-time monitoring through an optimized automatic wavelength scanning method for temperature compensation. PS-FBG sensors effectively detect ultrasonic signals, crucial for the early diagnosis of insulation faults, and show potential for moisture monitoring due to their reliable performance in varying environmental conditions. They offer superior sensitivity and reliability, effectively addressing the issues of electromagnetic interference and temperature-induced drift, thus enhancing the safety and efficiency of transformer monitoring. Further detailed analysis of how the sensor performs under prolonged exposure to aging by-products and other contaminants in transformer oil would enrich our understanding of its long-term reliability. In the end, for simultaneous temperature and humidity, ref. [122] developed a novel sensor based on tilted fiber Bragg grating (TFBG). As polyimide materials are affected by both temperature and humidity, the TFBG sensor, designed with a specific tilt angle and coated with polyimides, demonstrated high sensitivity and accuracy. Laboratory experiments and machine learning analysis, particularly using gradient boosting, confirmed its effectiveness.

3.5. Oil Quality Sensors

Oil sensors are essential for the real-time monitoring of transformer oil quality, helping to prevent failures and extend transformer life. Various studies have explored innovative sensor technologies to detect oil degradation and other critical parameters.

The researchers in [123] investigated the effectiveness of intensity-modulated plastic FO sensors in detecting aging in rapeseed transformer oil. Using polymethyl methacrylate plastic FOs, the research explored various sensing lengths (1.5, 2, and 3 cm) in a U-shaped configuration (Figure 5) to assess sensitivity to changes in the oil’s RI. The results showed that shorter sensing lengths produced more consistent and linear responses, suggesting they are less affected by environmental noise and disturbances compared to longer lengths. However, it should be noted that the study did not address the sensors’ long-term durability or their response to different aging by-products beyond changes in RI, which are critical for ensuring reliability and comprehensiveness in aging detection.

In another study [124], a tilted FBG sensor with a 130 nm Ge-Sb-Se-Te coating, using an ArF excimer laser and RF sputtering for uniform coating was fabricated to measure the RI of transformer oil. Utilizing an erbium-doped fiber amplifier (EDFA), polarization controller, and optical spectrum analyzer, the sensor showed a linear sensitivity of 9.952 dB/RIU for an RI range of 1.474–1.484 RIU by analyzing resonance amplitudes at 1560 nm. Additionally, a meta-surface-inspired complementary split ring resonator (CSRR) sensor operating at 2.94 GHz was designed and reported in [125] for the real-time monitoring of transformer oil quality. The sensor, which allows oil samples to flow through without direct contact, showed significant frequency shifts from 135.5 MHz to 470.5 MHz correlating with oil degradation over 24 to 504 h and enabling non-destructive monitoring by measuring changes in dielectric constant, viscosity, turbidity, and density. The CSRR sensor’s nearly linear response to varying oil volumes demonstrated its effectiveness in detecting different stages of oil degradation. This innovative approach allows for continuous transformer health monitoring, crucial for preventing failures. The authors in [126] implemented an evanescent field absorption-based fiber optic sensor (EFA-FOS) for detecting oil degradation. This sensor consisted of a plastic FO with a section of its cladding removed, using polymethyl methacrylate for the core, and measured changes in the optical power due to evanescent field absorption influenced by the oil’s RI. In testing, oil samples from seven transformers and one new sample were evaluated, showing high sensitivity and linearity, with the sensor output correlating well with the oil’s breakdown voltage. The EFA-FOS offers a low-cost, accurate method for the real-time monitoring of transformer oil degradation, providing an effective alternative to more complex methods. For another sensor assessing the aging condition of mineral oil, ref. [127] designed a novel optical sensor using carbon dots (CDs) synthesized from pomegranate peels. The CDs were prepared through a hydrothermal process and dispersed in olive oil, with their photoluminescence (PL) properties analyzed using a fluorescence spectrometer. The sensor exhibited high sensitivity and selectivity, showing significant emission intensity changes and peak wavelength shifts correlating with oil aging. Strong linear correlations were found between PL features and both aging time and dielectric dissipation factor (DDF) to validating the sensor’s potential for continuous online monitoring and transformer maintenance. As novel way to measure fluid degradation using frequency response analysis (FRA), ref. [128] designed a dual coil sensor. The sensor, resembling a single-phase transformer, used the fluid as the magnetic core and consisted of two 200-turn copper wire coils. FRA tests showed that the sensor could detect and quantify contamination levels in water with salt and oil in water mixtures, with resolutions of 15 ppm and 1 ppm, respectively. Changes in the frequency response indicated contamination levels. This study demonstrated a non-invasive, accurate method for monitoring fluid conditions in power transformers.

The researchers in [129] fabricated a distributed strain-sensing system using apodized π-phase-shifted FBG sensors to detect power transformer oil breakdown. The sensor, featuring a π-phase shift and Gaussian apodization, increased sensitivity to 60%/0.1 μs train and broadened the measurable strain range to 0.45 μs. The AP-PSFBG sensors successfully detected high-frequency acoustic waves generated by PDs, with the capability to sense frequencies up to 500 kHz. This system demonstrated high sensitivity, a wide dynamic range, and robustness to electromagnetic disturbances, making it suitable for real-time transformer condition monitoring. Beyond that, a high RI sensor based on a single-mode-multimode-silica rod-single-mode fiber structure was designed for monitoring transformer oil quality [130]. The sensor comprised a short segment of a multimode silica rod spliced between two single-mode fibers utilizing leaky mode interference to measure high RI and monitor oil degradation. It was tested with Cargill Series A RI liquids and thermally degraded transformer oil samples and demonstrating high sensitivity (115.12 dBm/RIU) and minimal temperature response. The sensor effectively detected oil quality changes, proving suitable for real-time transformer oil monitoring.

The real-time monitoring of viscosity, density, and dielectric constant of transformer oil is a comprehensive methodology that led researchers, in [131], to devise a quartz tuning-fork sensor. The sensor used a 32.768 kHz quartz tuning-fork crystal oscillator, modified to enhance sensitivity to the oil’s dielectric properties, and was calibrated using standard oil samples. The impedance analyzer measured changes in resonance frequency due to mass loading and viscous damping, providing accurate measurements with errors of 2.52% for density, 6.53% for viscosity, and 3.36% for dielectric constant. The sensor demonstrated good repeatability and matched standard sensor results across temperatures from 25 °C to 90 °C, making it suitable for continuous transformer oil condition monitoring.

In new combined sensor materials, molecularly imprinted polymers (MIPs), which are synthetic materials, were used in [132,133] where the materials were designed to have specific cavities that selectively bind to target molecules, mimicking the behavior of natural recognition entities like antibodies and surface plasmon resonance (SPR), provided an optical detection technique that measured changes in the RI near a sensor surface [134,135,136], and were used to study interactions between biomolecules or detect specific chemical compounds [137]. The study [138] examined the use of SPR optical fiber sensors with MIP to detect 2-furaldehyde (2-FAL), which is a by-product of thermal degradation in transformer cellulose insulation [3], providing a non-invasive and sensitive method for monitoring transformer health. The SPR sensor, made with plastic fiber optical (PFO) coated with an MIP layer specific to 2-FAL, showed high sensitivity and selectivity in various media, including air, water, and mineral oil, with performance unaffected by impurities. The research concluded that the PFO-SPR-MIP sensor is highly effective for real-time monitoring of transformer insulating oil, offering superior stability and robustness in harsh environments. Besides, researchers in [139] produced a self-produced sensor system by optimizing a resonant photoacoustic (PA) cell, reducing its volume by about 80%, and combining it with a handmade cantilever fiber acoustic sensor, an erbium-doped fiber amplifier, wavelength modulation, and harmonic detection technology. The system, aimed at detecting trace amounts of acetylene (C₂H₂) in transformer oil to identify high-energy discharge faults, achieved a detection limit of 6 ppb and an excellent linear range under 1000 ppm. The PA cell was constructed using brass for better noise resistance, with temperature control and dual-resonance enhancement to improve sensitivity. Finally, a novel cross-capacitive sensor for monitoring transformer oil was developed [140] by targeting moisture and 2-furfural (2-FAL) concentrations. The sensor, composed of four brass electrodes with PTFE insulation, measured capacitance changes correlating with oil’s dielectric constant variations using an using a commercial capacitance-to-digital converter. Testing with different oil types and varying moisture and 2-FAL levels showed high sensitivity, linear response, and temperature independence. The sensor demonstrated a maximum accuracy of 0.77%, making it a robust and effective solution for real-time transformer oil condition monitoring. In the end, it should be considered that for accurate oil measurement in transformers, sensors that can monitor the conductivity of the oil without direct contact are invaluable. Kantamani et al. [141] developed an inductive–capacitive probe with two concentric non-conducting cylindrical tubes, coils on high permeability soft ferromagnetic toroid cores, and capacitive electrodes. The excitation coil induced a current in the liquid, while the sensing coil detected the induced current, fortifying the current path, which remains cylindrical for accuracy. The signal conditioning circuit employs a dual-frequency approach to mitigate errors from coupling capacitance and external materials. The experimental results showed high accuracy with a worst-case error of less than 0.82%, demonstrating the probe’s effectiveness and reliability for transformer oil monitoring.

Future research should focus on enhancing sensor durability, validating performance under diverse conditions, and ensuring economic feasibility for widespread adoption.

3.6. Dissolved Gas Sensors

Three primary types of gas sensors are utilized for various applications: spectroscopic, optical, and solid-state. Spectroscopic sensors assess gases by examining their molecular characteristics [142], such as mass and vibrational spectra. Optical sensors function by analyzing how light is absorbed by gases after stimulation [143]. Solid-state sensors detect gases by observing changes in the properties of specific materials upon exposure to gases [144]. For identifying transformer issues, hydrogen (H₂) is critical. Higher levels indicate excessive dissolved gas, while a level under 100 ppm in oil suggests normal operation necessitating checks on the oil’s insulating capability [145]. Methods to measure H2 in oil include gas chromatography, photo-acoustic spectrometry, and calorimetry spectroscopy [3,146,147,148]. To address issues like size, complexity, and electromagnetic interference found in traditional gas analysis devices, ref. [145] demonstrated high sensitivity, stability, and repeatability for detecting hydrogen in transformer oil. They designed a micro–electro–mechanical systems (MEMS) hydrogen sensor using a Pd-Ni alloy with an Al₂O₃ passivation layer (that effectively prevented hydrogen infiltration into unwanted regions) by integrating a microheater and temperature sensor for optimal performance. The sensor was tested both in a gas chamber and in transformer oil, demonstrated a near-linear response to hydrogen concentrations, and showed a detection limit of 10 ppm and a sensitivity of 2.20 μV/ppm. This technology offers substantial improvements in transformer maintenance and safety through effective real-time hydrogen monitoring but does not address the sensor’s durability under real-world conditions, cost-effectiveness, or ability to detect gases other than hydrogen. Additionally, the environmental impact of the materials used and the sensor’s technical specifications were not fully discussed.

A TiO₂-SnO₂ composite hydrogen sensor was developed using TiO₂ quantum dots and SnO₂ nanosheets [149], with a 2 s response time, 5 s recovery time, high stability, and selectivity within a detection range of 4.8–5000 ppm. The sensor maintained performance across different humidity levels and was effective for real-time hydrogen monitoring in oil-immersed transformers. The study highlighted the sensor’s potential for fault diagnosis in transformers, showcasing its suitability for broader industrial gas detection applications. Alternatively, Lin et al. [150] tested Pd/C, Pd/C-R, and Pd/NC materials in sensors to monitor gases such as H₂, CO, and C₂H₂ in transformer oil while revealing varying sensitivity levels based on material modification and operating current. The Pd/NC sensors displayed superior performance, attributed to the synergistic effects of nitrogen doping. This demonstrates that modifying the chemical structure of carbon-based sensors can significantly enhance their gas-sensing capabilities, particularly for detecting specific gases in transformer oil. However, there is a noticeable lack of comparative data against other established gas-sensing technologies, which are crucial for validating the proposed sensors’ operational superiority and economic feasibility.

A highly sensitive low-pressure tunable diode laser absorption spectroscopy (TDLAS) sensor for detecting CO and CO₂ in transformer insulating oil was developed [151] using vacuum degassing and the dissolution equilibrium technique. For the sensor to be sensitive and selective in the detection system, the sensor employed a 1579.5 nm DFB laser. The calibration of the sensor was carried out using mechanical oscillation gas chromatography, which would be reliable calibration. This sensor obtained correct and reliable measurement, with MDL of 0.147 ppm for CO and 0.9 ppm for CO₂, whereas precision and repeatability were obtained. The measured error with the calibrated sensor was less than ±6 ppm or 6%, remarking that the designed sensor could monitor transformer health in an on-line method. With that knowledge, most recent papers have focused on determining the CO gas sensing properties using an atmospheric chamber [152,153,154,155,156], but in [157], the authors designed a room temperature gas sensor by synthesizing SnO₂ nanoparticles using a hydrothermal method and Pd nanoparticles using a sol–gel process, then depositing the SnO₂ on alumina substrates via screen printing and decorating with Pd. The sensor detected dissolved CO in transformer oil by measuring changes in electrical resistance. The results demonstrated high sensitivity, detecting CO concentrations as low as 13.3 ppm, with a clear correlation between CO presence and sensor resistance.

In [158], a portable photoacoustic spectroscopy [159] sensor was designed, using a distributed feedback laser (DFB) [160] and EDFA [161] for the high-sensitivity detection of C₂H₂ in transformer oil. They employed a combined headspace degassing method to extract dissolved gas and used wavelength modulation spectroscopy with second harmonic (2f) demodulation to reduce noise. The sensor demonstrated a 3.4 ppb detection limit. It also demonstrated high accuracy and stability, making it suitable for field applications in monitoring insulation status. The experimental results confirmed the sensor’s capability to provide reliable on-field measurements by meeting relevant standards in comparison to gas chromatography [162]. Also, Tang et al. [163] designed an MEMS gas sensor array with eight SWCNT-based sensing units, each functionalized or doped differently for enhanced sensitivity and selectivity. The sensor array was tested for H₂, CO, and C₂H₂ detection both in single and mixed gas environments by utilizing a deep belief network-deep neural network (DBN-DNN) model for pattern recognition. The results showed high sensitivity, accuracy, and reliability in gas detection. Moreover, for acetylene detection, Jing Wang et al. [82] designed and fabricated a solid-state electrochemical gas sensor by using yttria-stabilized zirconia (YSZ) as the solid electrolyte and cadmium titanate (CdTiO₃) as the sensing electrode (SE). The sensor was constructed on an alumina substrate with a Pt heater, and CdTiO₃-SE was synthesized using a sol–gel method and showed the highest response among tested materials. The sensor exhibited a response value of −126 mV to 100 ppm C₂H₂ with a detection limit of 500 ppb at 550 °C, demonstrating high sensitivity, selectivity, and stability. This makes it suitable for monitoring C₂H₂ in power transformers. As a novel method, ref. [164] invented a gas sensor based on LaFeO₃ (lanthanum ferrite) for the selective detection of acetylene (C₂H₂) and ethylene (C₂H₄) gases. The sensor was fabricated by a sol–gel method, with LaFeO₃ powders calcined at different temperatures to achieve optimal properties, and then deposited on platinum interdigitated electrodes. The LaFeO₃ sensor, particularly the one calcined at 600 °C, exhibited high selectivity and sensitivity for acetylene and ethylene, demonstrating strong responses to these gases at operating temperatures of 200 °C and 250 °C. This sensor shows promise for monitoring dissolved gases in transformer oil, offering reliable and reproducible detection of C₂H₂ and C₂H₄. Another study reported in [165] developed a novel system for detecting trace acetylene in transformer oil, using FO photoacoustic sensing. By optimizing the laser wavelength and employing a thermostatically controlled laser module, the system achieved a high sensitivity with a 0.5 µL/L detection limit. The experimental results confirmed the system’s capability to detect acetylene effectively, making it a compact, non-invasive, and electromagnetic interference-resistant solution for the continuous monitoring of transformer conditions by presenting a promising alternative to conventional methods. Even if we omit the discussion of cost-effectiveness, ease of integration, or performance under varying conditions, the study also does not address either long-term reliability or the ability to detect multiple gases simultaneously. Finally, Aasi et al. [166] reported that green phosphorene enhanced with palladium and platinum significantly improved the detection of dissolved gases like acetylene in transformer oil. The study reported that Pd-decorated green phosphorene shows a 65.9% sensitivity toward acetylene with a quick recovery time by highlighting its potential for real-time monitoring. These findings suggest that metal-decorated green phosphorene offers a promising and effective approach by detecting three main DGA gases (H₂, CH₄, and C₂H₂), thus making it suitable for enhancing the safety and efficiency of power transformers. Regarding that, the focus is mainly on acetylene, with little attention paid to other fault gases, the use of precious metals like palladium and platinum raises concerns about cost and scalability, and the environmental impact of these materials is not discussed. These issues suggest that while the sensor is effective for specific applications and performance in variable environmental conditions, its long-term stability, durability, and broader practical implementation require further evaluation.

3.7. Moisture Sensors

Moisture, which is enemy number one of the insulation system, must be monitored effectively to guarantee the operational efficiency of power transformers. It can be detected using various techniques, including Karl Fischer titration, polarization and depolarization current measurements, return voltage method, frequency domain spectroscopy, and other methods such as capacitance measurement, infrared spectroscopy, and microwave techniques [167,168]. Over the past decade, some of the sensor technologies developed to continuously monitor the level of moisture include water activity probes and capacitive sensors (polymeric film between two metal foils forming a capacitor) and FO sensors, and the successful integration of these sensors into online diagnostic systems require appropriate sensor placements for accurate moisture profiling [169]. Water activity probes, placed on top or directly immersed in the transformer oil, provide continuous measurements of water activity and temperature, which are used to calculate the water content in paper [170]. These probes offer the advantage of continuous measurement without requiring transformer disconnection, but their accuracy depends on achieving thermodynamic equilibrium with the surrounding oil. The response times of these probes in transformer oil are typically in minutes [171]. Polyimide-coated FBG sensors have shown effectiveness in detecting moisture, aligning well with the Karl Fischer titration results. These sensors work effectively with both mineral and synthetic ester oils, although further research is needed to explore the impact of fluid type and condition and ensure scalability for real-world use [104]. Similarly, FBG-based optical sensors, designed with polyimide coatings for stability, have been developed to measure moisture in transformer insulation [172]. These sensors, tested in two transformers filled with mineral oil and Envirotemp FR3 fluid, effectively tracked moisture migration during thermal cycles, demonstrating similar accuracy and sensitivity to commercial capacitive probes. It is crucial to consider the numerous challenges involved. These include the difficulty of placing sensors within the solid insulation, the dynamic nature of moisture levels during operation, and the necessity of testing under varied operational conditions to ensure sensor reliability and robustness. Therefore, accurately evaluating the condition of the paper insulation requires addressing these complexities [173,174].

The authors in [175] designed a humidity and moisture sensor by creating a single-mode-coreless-multimode-single-mode structure, with a coreless fiber coated in poly-vinyl acetate and it demonstrated high sensitivity to humidity in air (0.19 dBm/%RH for increasing and 0.16 dBm/%RH for decreasing humidity) and moisture in oil (0.15 dBm/ppm). Their results showed linear responses in the specified ranges, indicating the sensor’s potential for accurate humidity and moisture detection in various applications. In addition, a photonic crystal fiber sensor using SPR techniques with internal and external metal deposition methods was designed, as described in [176]. This sensor is designed to detect changes in the refractive index (RI) of analytes such as water and transformer oil by measuring shifts in resonance wavelength using an optical spectrum analyzer. The sensor’s core structure includes air holes arranged in hexagonal and octagonal patterns, coated with gold and titanium nitride, and utilizes a Penta core configuration to enhance sensitivity. The experimental results demonstrated high transmission and low absorbance for water, as well as low transmission and high absorbance for transformer oil, confirming its capability to accurately detect RI changes. When tested in a laboratory setup, the sensor showed potential for real-world applications in monitoring transformer oil conditions to ensure operational reliability. However, its material costs should be considered.

Capacitive humidity sensors were designed by anodizing high-purity aluminum sheets to form a porous Al₂O₃ layer to detect moisture via measuring changes in capacitance in response to varying humidity levels [177]. With pore morphology controlled by varying anodization parameters, silver electrodes were deposited on the anodized Al sheet to form a capacitor. The sensor detects moisture by measuring changes in capacitance in response to humidity level variations. It was shown that the sensor was sensitive to moisture levels from 180 ppm to 800 ppm in transformer oil and therefore suitable for the online monitoring of transformer oil conditions. Furthermore, the use of sensor and data analyses were reported in [178], who detailed the development of a high-frequency sensor system using an open-ended coaxial probe and an optimized artificial neural network model to measure moisture concentration in transformer insulation. This system included a microwave vector network analyzer and process reflection coefficients to determine dielectric permittivity and moisture content. The system was tested with three types of insulating oils (mineral oil, natural, and synthetic esters) with varying moisture levels (10, 20, 35, and 50 ppm) and demonstrated high sensitivity (0.027/ppm) and low errors (<±3%) that showed high accuracy and reproducibility. Conversely, the article [179] detailed the design and development of an S-taper fiber sensor for real-time moisture monitoring in oil. The sensor, constructed with specific structural parameters like waist diameter and axial offset, enhanced sensitivity to moisture-induced RI changes. Tested on transformer oil samples with varying moisture content, the S-taper fiber demonstrated high sensitivity (0.48 nm/ppm) and a low detection limit (2.19 ppm). Although the laboratory results are promising, further field validation, long-term stability testing, and economic feasibility assessments are necessary. Lastly, for real-time moisture monitoring in transformer oil, ref. [180] discussed the development of a micro-nano fiber (MNF) sensor. The MNF sensor, created by tapering a single-mode FO to diameters ranging from 800 nm to 125 μm, utilized the evanescent field to detect changes in the oil’s RI correlating with moisture content. Tested on oil samples with varying moisture levels, the MNF sensor showed high sensitivity (1.8 ppm) and accuracy, maintaining performance across temperatures from 30 °C to 80 °C. While promising, further field validation, long-term stability testing, and assessments of integration, scalability, and economic feasibility are needed to confirm its practical applicability. Addressing these gaps will enhance the understanding and adoption of MNF sensors for the real-time moisture monitoring in transformer oil.

3.8. Partial Discharge Sensors

Detecting PD in transformers is crucial as it indicates potential insulation failure. Transformers experience three primary types of PDs: the Corona type, which occurs within gas bubbles suspended in oil or at transformer terminals in air, producing hydrogen and methane; the sparking type, manifesting as small arcs that degrade oil or paper insulation, producing acetylene and hydrogen; and the surface type, occurring at the interface of different insulation materials, indicating points of weakness [181]. Various methods have been developed for PD detection. Ultra-high-frequency PD detection is known for its high sensitivity and reliability in locating discharge sources, although it faces challenges such as susceptibility to electromagnetic interference [84,182]. FO technologies, including fiber optic interferometers (Michelson, Fabry–Perot, Sagnac, and Mach–Zehnder) and FBG sensors, are used for PD monitoring through acoustic signals, offering high sensitivity and a broad response range [112,183], but it should be mentioned that according to IEEE Std C57.127 [184], the frequency range for PD acoustic emission detection is between 20 and 500 kHz, guiding the setup and calibration of detection equipment [113]. Based on this information, Figure 6 categorizes various types of partial discharge sensors.

The researchers in [185] introduced a PD detection system using a distributed feedback fiber laser (DFB-FL) based on an asymmetric 3 × 3 coupler and by removing the direct impacts of the NPS method, an ultra-high SNR was proposed that improved the NPS (INPS) scheme for FO sensors used in PD detection. This system achieved an ultra-high average signal-to-noise ratio of 38.30 dB, which is 24.2 dB higher than traditional methods and significantly surpassed the performance of PZT sensors by addressing issues related to coupler asymmetry and enhancing signal integrity. Tested on an 80 kVA oil-immersed transformer, the system demonstrated superior performance in detecting PD signals without distortions. Moreover, the AEs refer to the release of transient elastic waves within a material due to stress or deformation [186] as discussed in Ref. [86]. For detecting PD in oil-immersed power transformers, the authors evaluated low-cost piezoelectric sensors, specifically PZT and microfiber composite (MFC). The MFC sensor displayed superior sensitivity and signal quality and captured a dominant frequency around 2 MHz and when compared to the PZT sensor, it had a dominant frequency around 1 MHz. These sensors were mounted on the transformer wall and effectively distinguished PD signals from environmental noise and electromagnetic interference using time-domain, frequency-domain, and time-frequency analyses. By employing methods such as the short-time Fourier transform, researchers were able to extract both frequency and time-domain information from the partial discharge (PD) phenomenon. This approach aids in distinguishing PD signals from environmental noise and vibrations. However, it should be mentioned that the MFC sensor demonstrated better overall performance. This research suggests that such low-cost acoustic sensors are a viable and cost-effective option for continuous PD monitoring in transformers, potentially enhancing their reliability and safety.

The Hall effect is the production of a voltage difference across an electrical conductor, transverse to an electric current and an applied magnetic field [187]. The [188] study involved PD detection and differentiation with the Hall effect and AE sensors in dry-type transformers; the frequency peaks of current oscillations recorded have shown that it is possible to distinguish the mechanical waves, whose amplitude and frequency content differences of PDs that could be discerned. Both sensors were tested for samples of epoxy resin and Nomex paper to differentiate PDs, depending on the insulation material while using DFT and DWT techniques. In this context, in [189], an extrinsic Fabry–Perot interferometer sensor attached to the drain valve of a transformer was utilized to detect PDs by measuring the membrane vibrations induced by ultrasonic waves. Numerical analysis showed that the fundamental frequency of the membrane was reduced by about 60% in transformer oil; in addition, the vibration amplitude was considerably damped. In contrast to higher-order modes, which are less sensitive, the first-order vibration mode of operation of the sensor is more sensitive and offers superior performance for PD detection. The study also emphasized that in the real-time sensor design, it is crucial to include viscous damping and thermal effects to ensure proper monitoring. For accurate measurement by the sensor, in-service material and design optimization should also be considered.

To detect PD ultrasonic signals in transformer oil, the authors of [83] developed a microfiber coupler sensor (MFCS) for utilizing coupled FOs for high sensitivity and anti-interference capabilities. The MFCS system included a light source, conditioning circuit, and oscilloscope to measure PD-induced changes in optical power. By employing the beetle swarm optimization–support vector machine algorithm, the system achieved 93% classification accuracy for various PD types, outperforming traditional methods in terms of accuracy and convergence speed. Other sensors developed for PD detection in power transformers include the HFCT, active dielectric windows (ADWs) with ultrasonic transducers, and meandered planar inverted-F antennas (MPIFA) [190]. The HFCT detected high-frequency currents from PDs, the ADW combined AE and UHF detection, and the MPIFA optimized UHF signal detection. Laboratory and field tests showed the ADW had over five times higher sensitivity for AE detection, and the MPIFA was up to 7.8 times more sensitive for UHF signals than traditional antennas. The hybrid system combining HF, UHF, and AE detection significantly improved PD monitoring reliability and sensitivity, enhancing transformer maintenance and fault prevention. In addition, an innovative study in [191] designed an ultrawideband (UWB) PD sensor for high-voltage power transformers. The sensor featured a multilayered design optimized for a wide frequency range (300–3000 MHz) and was mounted on transformer inspection hatches using a circular window and oil sealant gaskets to prevent leakage. Performance validation included frequency-domain testing with a vector network analyzer and PD sensitivity testing using five emulated PD sources, demonstrating high sensitivity and effective PD detection across a broad frequency range. This UWB sensor significantly enhanced PD detection, offering improved monitoring for early warnings and preventive maintenance in HV power transformers. Similarly, ref. [192] detailed the design of a DFB-FL sensor for detecting PD acoustic emissions in oil-immersed transformers. This sensor, composed of an erbium-ytterbium co-doped DFB fiber laser and a Michaelson fiber interferometer, achieves ultra-high sensitivity by detecting wavelength changes caused by structural strain. The DFB-FL sensor demonstrated a peak-to-peak response value 637% higher than traditional piezoelectric transducer sensors, with effective noise suppression methods reducing average noise amplitude by 91.67%. While laboratory results show promise, further field validation, long-term stability testing, and economic feasibility assessments are needed. Addressing these gaps will enhance the practical application and reliability of DFB-FL sensors for real-time transformer monitoring.

One of the most comprehensive studies of sensor design and testing is that in [193], where the authors designed a combined in-oil PD sensor integrating AE and UHF methods. Consisting of a piezoelectric AE sensor and a UHF antenna, it was installed using an air extraction method to avoid air entry during installation, ensuring non-destructive implementation. Vacuum pressure tests validated its sealing performance, while in-air and in-oil tests confirmed its sensitivity and effectiveness in detecting PD signals from various insulation defects. Additionally, its ability to measure AE and UHF signals simultaneously and determine the time difference (∆t) between them enables accurate PD source localization. Moreover, for effective PD detection, in addition to the design of winding, the location and type of sensor are also crucial [194]. It should be mentioned that HFCT consists of a magnetic core, a sensing element (like a Hall-effect sensor or Rogowski coil), and electronic circuitry that are working together to detect and convert high-frequency currents into proportional signals for monitoring and analysis [195]. The AE sensor for PD detection (referenced in [196]) successfully determined the location of the PD source. Additionally, the sensor used the time differences of arrival (TdoA) method to calculate the distance of the PD, as described in [197]. In [198], the AE sensor effectively captured PD sound waves, particularly on the high-voltage side. The HFCT sensor captured corresponding electromagnetic signals, and with a time gap of 6.289 ms, it was confirmed as a significant PD event. The relocation of these sensors, using the TdoA, enabled an accurate location of the PD source at a point approximately 0.9 m from AE1. The application of AE and HFCT sensors will enable the precise online detection of PDs, thereby enhancing transformer maintenance and preventing severe damage.

The authors of [199] developed a novel lattice–Rogowski–coil (RC) sensor to enhance PD monitoring and localization in power transformers, featuring a thin, flat-shape PCB design that balanced low resonant frequency and high mutual inductance. This RC sensor, composed of multiple electrically conducting layers and interspersed with FR-4 and Teflon substrates, detects the magnetic flux produced by PD pulses and is installed inside the transformer tank parallel to the grounded wall to avoid operational interference. In a specially prepared 20 kV/0.4 kV, 500 kVA distribution transformer, the sensor successfully localized PD sources by analyzing the peak sensor output voltage and demonstrated adequate sensitivity up to 30 cm from the winding. The lattice-PCB-RC sensor’s high resonance frequency and flat design make it an effective tool for PD localization within transformer tanks. Beyond that, to determine the optimal placement of UHF sensors on the tank wall of transformers, the research in [200] utilized a three-dimensional finite element method (FEM) simulation model [201] in both air and oil environments. They concluded that for effective PD detection, four sensors (two on the top and two on the vertical walls) are suitable; meanwhile, for enhanced PD source localization, six sensors (two on top and four on the vertical walls) should be used. In other research about optimized UHF sensor placement on transformer tanks for accurate PD localization, the authors of [87] used hyperbolic surfaces based on TDOA. By employing a genetic algorithm for optimization, commercially available UHF sensors were strategically placed to minimize localization errors caused by non-line of sight (NLOS) effects and signal reflections. Monte Carlo simulations [202] and CST Microwave Studio [203] confirmed that this optimized UHF sensor placement greatly improved localization accuracy for PD sources in transformers and enhanced real-time diagnostics.

In the end, it has been discovered that arcing faults in transformers result in very high temperatures and high rates of gasification of the insulating oil, with formation of explosive gas mixtures consisting of mainly hydrogen (H₂), methane (CH₄), and acetylene (C₂H₂). This fast process causes sudden internal pressure, which can lead to structural failure or even catastrophic explosions if not well managed [204,205].

3.9. Bushing Sensing

Capacitance and dissipation (power) factor measurements of transformer bushings are essential for assessing the insulation condition, where an increased dissipation factor indicates potential aging or deterioration of the bushing insulation [206]. The researchers in [207] developed a method using non-invasive capacitive sensors (NICSs) and an air gap capacitor (AGC) for the real-time monitoring of transformer bushings. The system employs NICSs to measure the voltage signal’s amplitude and phase from the high-voltage conductor, while the AGCS provides a reference voltage. This setup allows for continuous online monitoring and detecting faults without taking the transformer offline. The sensors were tested on various transformers and showed high accuracy and reliability in detecting bushing faults, short circuits, and variations in oil conductivity. The method significantly improves transformer maintenance and longevity by providing real-time diagnostics.

Building on these advancements, other researchers have also focused on innovative sensor technologies to enhance transformer bushing diagnostics. For instance, Jiang et al. [208] designed a UHF sensor to detect PD in transformer bushings. One of the PD faults in simulation was corona discharge at the top of bushing and the sensor effectively detected PD in lab tests with a 96% classification accuracy. This technique enhances the transformer maintenance and reduces failure risks, with future work focusing on optimization and cost reduction. Another research study with a comprehensive view [209] invented multi-parameter sensors to monitor hydrogen, temperature, pressure, micro water content, dielectric loss, capacitance, and PD in transformer bushings. As an advantage of this work, the four-in-one sensor was designed to be compact and lightweight to fit within the bushing flange. Tested in high-voltage substations, the sensors showed high accuracy and reliability, significantly enhancing fault detection and transformer safety.

In the same way, as a novel Sagnac interference optical sensor for detecting PD in transformers, the authors of [210] fabricated a sensor that used FOs to detect ultrasonic signals generated by PD, leveraging phase changes for high sensitivity detection. Laboratory experiments on a 72.5 kV bushing confirmed the sensor’s effectiveness, demonstrating accurate and reliable PD detection with high sensitivity and comparative measurements using a high-frequency current transformer. This advancement offers significant potential for improving transformer condition monitoring.

Also, for monitoring internal pressure in transformer bushings, a novel graphene piezoresistive pressure sensor was developed [211]. The sensor, made from graphene oxide-coated handkerchief paper, exhibited high sensitivity, fast response times, and durability. The handkerchief paper was chosen for its highly porous structure and fibrous tissue, which contribute to superb sensitivity. Through repeated pressure tests, using a universal testing machine and a precision source meter, the field tests demonstrated its effectiveness in real-time pressure monitoring and fault detection. The integrated hierarchical alarm system provided early warnings, enhancing transformer safety and reliability. As a further improvement, they should consider sensor design, reduce costs, and improve durability for broader application. Lastly, the researchers in [212] produced a capacitive coupling device for measuring PDs in transformer bushings. The device, designed for ABB-type GOB bushings, included coupling capacitance and protection circuits with gas discharge tubes and thyristors to handle high-frequency signals and overvoltage protection (Figure 7).

Laboratory tests validated the device’s ability to detect PD signals without significant attenuation and withstand high voltage and impulse stress. Field applications demonstrated the reliable real-time monitoring of PD activities in substations. The device enhances transformer monitoring, contributing to proactive maintenance and fault prevention by ensuring signal integrity and robust protection against surges and interference.

3.10. Tap-Changer Condition Sensing

The recent research into on-load tap-changer (OLTC) monitoring has seen significant strides, emphasizing both innovative sensor development and the application of existing sensor technologies. Notably, there has been relatively little research dedicated to inventing or developing new sensors specifically for tap-changer monitoring. This suggests that currently available sensors largely fulfill the existing diagnostic requirements.

Based on the work reported in [213], researchers designed an innovative OLTC detection system integrating wireless sensor network (WSN) technology with embedded platforms. This system included sensors for monitoring OLTC’s mechanical vibrations, rotation angle, and driving motor current. Utilizing ZigBee technology and an ARM microcontroller, the system effectively monitors OLTCs in real time, diagnosing faults accurately. However, further validation and enhancements in data security and wireless robustness are needed. In another study [214], the authors created a novel method for diagnosing mechanical faults in OLTCs using existing vibration sensors. They used short-time Fourier transform for signal processing and optimized support vector machine algorithms, achieving over 99% diagnostic accuracy even under noisy conditions. The method proved effective in a controlled environment, with future work required to validate its robustness in real-world settings. Additionally, Lin Cheng et al. [215] used high-precision pressure sensors to study pressure changes during high-energy arc discharge faults in OLTC oil chambers. They installed sensors at various positions within a cylindrical transformer oil chamber and applied high voltage to simulate an arc fault, capturing detailed pressure distribution and evolution characteristics. The study underscored the importance of pressure relief mechanisms in OLTC oil chambers to mitigate structural damage and potential explosions. In the research of [216], an OLTC monitoring system was installed on nine 370 MVA autotransformers, using accelerometers, temperature sensors, and current clamps to measure vibro-acoustic signals and motor current, as shown in Figure 8.

These studies highlight the effectiveness of existing sensor technologies in OLTC monitoring and fault diagnosis. The limited focus on developing new sensors indicates that current sensors are sufficient for diagnostic needs, with ongoing improvements in application and data analysis techniques.

3.11. Commercially Available Sensors

Selecting and positioning sensors for power transformers involves considering several key principles to ensure effective monitoring and protection. The following are the main principles guiding sensor selection:

• Compatibility with Transformer Specifications: Sensors must be compatible with the transformer’s operating specifications such as voltage levels, current ratings, and frequency ranges.
• Positioning: Sensors should be strategically placed throughout the transformer, so operators can gather comprehensive data on its condition, enabling predictive maintenance and enhancing operational reliability. For example, monitoring core and windings involves placing sensors near the core and windings to monitor temperature and detect hot spots, while gas and moisture sensors should be placed in areas prone to gas and moisture buildup, such as near insulation materials or gaskets.
• Measurement Accuracy: Sensors should provide accurate measurements of parameters critical to transformer health, such as temperature, oil level, and vibration.
• Reliability and Durability: Sensors must be reliable and durable to withstand the harsh operating conditions typical of transformer environments, including temperature extremes and electromagnetic interference.
• Response Time: Sensors should have a fast response time to promptly detect and respond to changes in transformer conditions, helping to prevent damage or failure.
• Ease of Installation and Maintenance: Sensors should be easy to install and maintain, minimizing downtime during installation or replacement.
• Compatibility with Monitoring System: Sensors should be compatible with the monitoring and control system used for centralized monitoring of transformers, ensuring seamless integration and data transmission.
• Cost-Effectiveness: Consideration of the initial cost of sensors and their long-term operational costs should align with the budget constraints of the asset management program.
• Safety Standards: Sensors must comply with relevant safety standards and regulations to ensure safe operation within the transformer environment.

By adhering to these principles, utilities and asset managers can select sensors that optimize transformer performance, enhance reliability, and extend operational life.

In the field of power transformer monitoring, commercially available sensors play a crucial role in ensuring the reliability and efficiency of transformers. These sensors, developed and refined by various companies, are used to monitor a wide range of parameters including temperature, dissolved gases, vibrations, and AEs.

Table 3. Commercially available sensors used in monitoring.

Sensor Type	Key Findings	Parameter
Bragg grating-based fiber optic sensors	Found slightly higher winding temperatures in transformers with natural ester oil compared to mineral oil, confirmed by CFD simulations [217].	Temperature
Fluorescing-tipped or gallium arsenide (GaAs)-tipped fiber optic sensors	Validated a new temperature calculation method for high current busbar leads in large transformers with a maximum error of 0.78%, outperforming traditional methods [218].	Temperature
Optical frequency-domain reflectometry system with single-mode fibers	Achieved high spatial resolution (1 cm) and temperature accuracy (0.1 °C) for transformer core temperature monitoring, validated against infrared thermography and FEM simulations [218].	Temperature
Fiber optic sensors and Kalman filter	Embedded in SCADA systems to monitor hotspot temperatures on transformer windings using a FBG-based quasi-distributed thermal sensing method [219].	Temperature
Photoelectric infrared thermal imaging and discharge circuit detection sensors	Measured internal temperatures and discharge signals, using a fuzzy data fusion model to improve fault diagnosis precision and accuracy, reducing data fluctuation and increasing diagnostic reliability [220].	Temperature
PT1000 platinum resistance sensors	Combined with PCA, LSTM neural networks, and decision trees to diagnose sensor faults with over 96% recognition rate and under 1 millisecond diagnosis time; LSTM model predicted temperature with an error margin within 0.1 °C [221,222].	Temperature
Pt100 temperature sensors	Strategically placed on transformer oil pipes for precise measurements of oil temperature to monitor and control transformer cooling processes. Modular and scalable, allows for remote monitoring and updates [223].	Temperature
Tunable diode laser absorption spectroscopy (TDLAS)	Detects acetylene dissolved in transformer oil with high sensitivity (7.1 mV/ppm), low detection limit (0.49 ppm), quick response, and no need for carrier gas [147].	Temperature
Robotic system with multiple sensors	Used thermal cameras, RGB cameras, AEs, depth cameras, LiDAR, humidity, and temperature sensors to monitor transformer lifecycle and detect hot spots, mechanical irregularities, and environmental conditions [224].	Temperature
Piezoelectric Accelerometers	Measured vibration acceleration in transformer tanks to assess core and winding conditions, finding optimal sensor placement in the middle of the tank, and indicating higher vibration acceleration in older transformers [225].	Core
Impedance Analyzer	Detected Disk-Space Variation faults in transformer windings with high accuracy using transfer function results and artificial neural networks (94.72% for Multi-layer Perceptron and 87.22% for Group Method of Data Handling) [226].	Winding
Acoustic Sensor	Compared Time Difference of Arrival and Acoustic Time Reversal methods for measuring PD [85].	Partial Discharge
Acoustic Emission Sensors	Prototype AE sensor displayed greater sensitivity and stability in detecting PDs, with higher amplitude PD pulses and a wider range of detectable PD signals compared to commercial sensors [227].	Partial Discharge
High-speed optical sensors	Detected PD in medium-frequency transformers under high dv/dt switching transients, finding that higher PWM frequencies increased PD susceptibility and reduced insulation lifespan [228].	Partial Discharge
UHF Disk Sensor, UHF Drain Valve Sensor, Commercially Antennas	Detected PDs in transformer insulation systems, with the log-periodic antenna achieving the highest defect-recognition efficiency. Machine learning algorithms classified PD types with high accuracy [229].	Partial Discharge
UHF sensors and dipole antennas	Measured electromagnetic waves from PDs, achieving high accuracy in locating PD sources using the inverse filter method with a 3D localization error of 5 mm [230].	Partial Discharge
Microfiber Composite sensor PZT sensor	Designed for detecting AEs from PDs, with the MFC sensor showing higher sensitivity and the PZT sensor providing detailed frequency analysis. Combining RMS and STFT analyses improved PD detection [88].	Partial Discharge
Piezoelectric ultrasound sensors	To detect acoustic signals from PDs in transformer oil, with sensor placement optimization reducing measurement uncertainty due to temperature variations. MUA software version 1.0 helped quantify and minimize uncertainties [231].	Partial Discharge
High-Frequency Current Transformer Sensors	Characterized PD systems using HFCT sensors with a modular test platform. Evaluated sensitivity, noise rejection, and defect localization. Ensured accurate PD diagnostics [232].	Partial Discharge
Commercially microphones	Measured low-frequency noise generated by power transformers, with preamplifiers and digital signal meters registering sound pressure levels [233].	Vibration
Piezoelectric acceleration sensors)	Measured vibration signals in a ±500 kV HVDC converter transformer, finding increased vibration intensities with load current and significant components at 100 Hz, 200 Hz, 300 Hz, and 400 Hz [234].	Vibration
Buchholz Relays	Detected gas accumulation and oil surges in oil-immersed transformers, providing early fault detection [235,236].	Gas Accumulation
Transformer Protector	Involves a fast-acting rupture disk that opens within milliseconds to depressurize the transformer tank, preventing explosive gas production and channeling gases to a remote area where they can safely burn [237,238].	Prevention of explosive gas production
Smart Photodiodes Array	Monitors color changes in silica gel of transformer breathers, indicating saturation levels [239].	Breather Health Monitoring
Commercially DGA Sensors	DGA is the leading technology in industry for early detection of many incipient transformer failures and degradation mechanisms. Measured dissolved gas concentrations in transformer oil, achieving high prediction accuracy with mean absolute percentage errors ranging from 1.525% to 5.763% using a wavelet-like transform and autoregressive neural network model [240].	Dissolved gas in Oil
PL spectrometers and UV-Vis spectrophotometers	Measured PL and UV-Vis spectra of transformer oil samples for condition assessment, with PL spectroscopy showing higher sensitivity and accuracy (98% and 99% correlation with DDF results) [241].	Oil’s degradation
Thin-Film Capacitive Sensor	Improving the capacitive sensor for real-time moisture measurement in transformer oil [242,243].	Moisture in oil
UV light source (365 nm), digital camera, fluorescence images	Captures fluorescence images of transformer oil under UV light for accurate oil leak detection using a U-net model [244].	Oil leak detector
Fringing Field Capacitive Sensor	To measure liquid levels in vessels without direct contact, showing linear response and high sensitivity [245].	Oil level detector
AE sensors	Evaluated effectiveness of various AE in detecting and classifying defects in OLTCs. The sensors demonstrated the best overall performance with ensemble subspace discriminant (ESD) algorithms [246].	Tap-Changer’s condition
Acoustic Emission Sensors	Developed a method using AE signals and machine learning to detect OLTC faults. AE signals were recorded with piezoelectric transducers and analyzed using wavelet decomposition and ML models, achieving high classification accuracy [247].	Tap-Changer’s condition
Accelerometer, Temperature, Current Clamp sensor	Detecting faults in OLTC by using vibro-acoustic signal analysis, Hilbert Transform, and Low Pass Filter to simplify the complex signal [216].	Tap-Changer’s condition

summarizes key studies involving different commercially available sensors, highlighting their applications, key findings, and further needs. These sensors have been instrumental in enhancing transformer maintenance strategies by providing real-time data and improving fault detection accuracy.

The use of commercial sensors in transformer monitoring has significantly advanced the ability to detect and diagnose faults early, thereby improving maintenance strategies and overall transformer reliability. Each sensor type brings unique advantages, such as high sensitivity, accuracy, and the ability to operate under various conditions. However, further field validation, long-term stability assessments, and integration with existing systems are necessary to maximize their effectiveness and practical application. These studies highlight the ongoing evolution and importance of sensor technology in maintaining the health and efficiency of power transformers.

3.12. Theoretical Sensors

In the quest for more efficient and accurate methods to monitor the condition of power transformers, researchers have turned to advanced theoretical models to develop novel gas sensors. These sensors, based on cutting-edge materials and innovative designs, show great promise in detecting dissolved gases in transformer oil as critical factor in diagnosing and preventing transformer faults. Utilizing density functional theory (DFT) and other computational methods, these studies explore the potential of various monolayer materials doped with different atoms to achieve high sensitivity and selectivity for specific gas molecules. The following table (

Table 4. Recent non-commercially available sensors.

Material	Target Gases	Key Findings	Notes
Pd-doped Janus HfSeTe monolayers	H2, CO, CH4, C2H2, C2H4	High stability and selectivity, particularly for C2H4 detection [248]	Promising for resistance-type and work-function-type gas sensors in transformer oil
Pd-C3N monolayers	HCHO, C2H3Cl	Enhanced conductivity and sensitivity with strong binding energies, stable under moisture [249]	Effective for real-time monitoring in dry-type transformers
Ir-decorated MoS2 monolayers	CH4, C2H4, C2H2	High sensitivity to C2H4 and C2H2, significant electronic property changes [250]	Enhances transformer condition monitoring, weak sensitivity to CH4
Ni-doped WS2 monolayers	H2, C2H2, CO	High sensitivity and selectivity, substantial improvements in adsorption energy and conductivity [251]	Suitable for DGA in transformer oil
Pd-modified Ti3C2O2	C2H2, C2H4, CH4	High adsorption energy and improved conductivity for C2H2 and C2H4 [252]	Effective for fault detection and maintenance in transformer oil
Cu-embedded PtSe2 monolayers	CO, HCHO	Strong chemisorption of CO, physisorption of HCHO, significant conductivity changes [253]	Requires experimental validation for practical application
Cu-decorated ZnO monolayers	CO, HCHO	Enhanced conductivity and effective detection of sensor’s layer [254]	Needs further experimental research for practical applicability

) summarizes the findings from these theoretical studies, highlighting the key materials, target gases, and potential applications.

These theoretical studies offer valuable insights into the development of advanced gas sensors for transformer oil applications. By leveraging the unique properties of various monolayer materials and doping elements, these sensors can potentially provide more accurate and reliable condition monitoring, aiding in the early detection of faults and enhancing the maintenance strategies for power transformers. However, it is crucial to conduct experimental validations to confirm the practical applicability and long-term stability of these promising sensor designs.

4. Utilization of Failure Modes and Effect Analysis

Failure modes and effects analysis (FMEA) is a structured approach used across various industries to proactively identify potential failure modes of a system, product, or process, and assess their effects on system performance. This approach typically involves the following steps [255]:

• identification of all potential failure modes that could occur within the unit.

  Table 5. Potential failure modes that could occur within the unit.

  Failure Location	Failure Likelihood (%)	Example of Typical On-Line Monitored Values
  Active part/main tank	55	DGA, load current, over-current, short circuit current, top/bottom oil temperature, hot-spot/winding temperature, over-voltage, transient over-voltages, moisture in solid/liquid insulation, oil level, partial discharge, oil pressure, aging, humidity of air inside conservator, condition of oil preservation system, gas in Buchholz relay.
  OLTC (Tap changer)	27	Active power consumption/torque of the OLTC motor drive, oil level, oil temperature, position, number of operations, operation time, inrush current, contact erosion, DGA.
  Bushings	17	Capacitance, power factor, transient over-voltages, oil/SF6 pressure, partial discharge.
  Cooling system	1	Cooling medium temperature (oil/water), status/condition of fans and pumps, oil flow, cooling efficiency.

  summarizes the main failure modes;
• analysis of the impact of each identified failure mode on the overall system performance;
• determination of the underlying causes of each failure mode;
• ranking of failure modes based on their potential impact, likelihood of occurrence, and detectability;
• development and implementation of strategies to mitigate or eliminate identified failure modes;
• documentation of the FMEA process and regular review and update as new information becomes available.

By systematically identifying and addressing potential failure modes, FMEA enhances the reliability and safety of systems, products, and processes, leading to improved performance and reduced risk of failures.

Understanding the root causes helps in developing effective mitigation strategies. Each failure mode is then evaluated based on its potential severity or impact on the overall system performance. This helps prioritize which failure modes are most critical to address. This step helps determine the likelihood of the failure being detected before it affects the system. By combining the severity, likelihood of occurrence, and detection capability (often quantified using a risk priority number or RPN), high-priority failure modes can be identified. These are the ones that warrant immediate attention and mitigation efforts. Based on the analysis, appropriate actions and mitigation strategies are developed to reduce the occurrence or impact of high-priority failure modes. The effectiveness of the mitigation strategies is implemented, and the component is monitored to ensure that the identified risks are adequately controlled over time. It should be mentioned that FMEA is not a one-time exercise but rather a proactive stance that helps organizations enhance reliability, safety, and efficiency across their operations. Furthermore, the location, climate, and load demand significantly affect transformer conditions. Urban transformers, for example, face periodic overloads, while industrial transformers experience constant overloads. FMEA can help reduce maintenance costs, enhance sensor performance, and improve the overall efficiency of transformer monitoring systems.

5. Discussion

Through a meticulous analysis of the existing literature and current methodologies, this paper provides a comprehensive understanding of how modern sensors enhance the reliability and efficiency of power systems. The discussion critically evaluates key findings, highlights the advantages and disadvantages of current sensor technologies, and proposes future research directions to address existing gaps in the field. This approach aims to synthesize insights from various studies to illuminate the advances and challenges in sensor technology within power systems.

5.1. Key Findings and Advancements

The integration of advanced sensor technologies such as FO sensors, and wireless sensing networks represents a significant leap forward in transformer condition monitoring. These technologies provide unparalleled sensitivity and precision in measuring critical parameters like temperature, moisture, PDs, and dissolved gases. FBG sensors, capable of monitoring multiple variables concurrently including mechanical strain and acoustic emissions, have proven particularly valuable in detecting early signs of transformer faults. Wireless sensor networks enhance this capability by enabling remote monitoring and real-time data transmission, crucial for proactive maintenance. Chemical sensors, notably semiconductor and electrochemical gas sensors, have demonstrated effectiveness in detecting gas emissions indicative of potential faults. Precise detection of hydrogen, methane, and acetylene gases is crucial for diagnosing insulation and oil degradation issues, thus preventing catastrophic failures. Furthermore, advances in UHF sensors and piezoelectric sensors have improved the accuracy of PD detection, a critical indicator of insulation health.

5.2. Advantages of the Review

This review stands out for its comprehensive and systematic approach to analyzing the vast array of sensor technologies used in transformer monitoring. By collating and evaluating diverse sensor types and their applications, this paper provides a holistic view of the current state of technology. It underscores the importance of integrating various sensing mechanisms to achieve a more robust and reliable condition monitoring system. The detailed examination of both established and emerging sensor technologies offers valuable insights into their practical applications and potential improvements.

5.3. Challenges and Limitations

Despite these advancements, several challenges persist in the practical deployment of these sensor technologies. Environmental susceptibility remains a significant concern, as sensors are often exposed to harsh conditions such as extreme temperatures, high electrical stress, and chemical contaminants. These factors can adversely affect sensor performance and longevity. For instance, while FBG sensors offer high sensitivity, their readings can be distorted by temperature variations, necessitating sophisticated compensation mechanisms to ensure accuracy. The long-term stability and durability of sensors are also critical issues. Continuous exposure to operational stressors can degrade sensor materials and performance over time, raising concerns about their reliability and maintenance costs.

Electrostatic charging and discharging can potentially affect sensors inside transformers, especially if the sensors are not properly shielded or if they are sensitive to electromagnetic interference (EMI). Careful selection of sensors and implementation of protective measures can help mitigate these effects and maintain reliable monitoring capabilities. Moisture content in transformers is crucial for maintaining their operational reliability and longevity.

Assessing moisture in the solid insulation online remains challenging due to the difficulty of accessing the solid insulation materials directly. However, several indirect methods and technologies can be employed for continuous or periodic monitoring of moisture levels in transformer solid insulation. Direct measurement can be performed offline by dielectric response analysis (DRA). This method measures the dielectric properties of the insulation, which change with varying moisture content. It can be used as an online monitoring tool to detect changes in insulation condition that may indicate moisture ingress. Hygroscopic and fiber-optic-based sensors, calibrated for transformer provide accurate water content measurements in oil. The concept of oil–paper equilibrium curves provides a framework for estimating moisture content in solid insulation. However, it is important to note that in-service transformers are seldom in equilibrium because of the dynamic nature of moisture within the insulation system. It is generally used to guide maintenance practices, help diagnose transformer conditions, and inform design considerations to ensure reliable transformer operation over its lifespan.

End users should keep in mind that no sensor is “set-it-and-forget-it”; regular maintenance, replacement, and calibration are always necessary. Moreover, integrating multiple sensors into a cohesive monitoring system poses logistical and technical challenges, including issues related to data fusion and real-time processing.

5.4. Future Directions

To address these challenges, future research should focus on developing more resilient sensor materials and designs that can withstand harsh operational environments. Innovations in nanomaterials and advanced composites hold promise for creating sensors with enhanced thermal stability and resistance to chemical degradation. Furthermore, research should aim at enhancing the long-term stability and multifunctionality of these sensors to ensure their consistent performance under varying operational conditions.

The integration of artificial intelligence and machine learning algorithms with sensor networks can significantly enhance fault detection and predictive maintenance capabilities. These technologies can analyze vast amounts of sensor data to identify patterns and predict failures, enabling more effective and timely interventions. Moreover, the development of multi-functional sensors capable of monitoring multiple parameters simultaneously will also be crucial. Such sensors can provide a more comprehensive assessment of transformer health, leading to better-informed maintenance decisions and extended transformer lifespans. Furthermore, collaborative efforts between academia and industry will be essential to accelerate the adoption of these advanced technologies in real-world applications.

In addition, enhancing sensor calibration methods and placement techniques will be vital to improving data accuracy and reliability. Standardized calibration protocols and advanced signal processing algorithms can help filter out noise and irrelevant signals, reducing the risk of false positives and data misinterpretation [256,257].

6. Conclusions

This review has provided a detailed and comprehensive analysis of the current state of sensor technologies for the monitoring of transformer condition. The integration of advanced sensors has significantly improved fault detection and maintenance strategies, enhancing the reliability of power systems. Sensors can significantly enhance an asset management program, offering substantial advantages. The data collected can be strategically used as a part of an organized process/program. However, challenges related to environmental susceptibility and long-term stability need to be addressed to fully realize the potential of these technologies. Future research should focus on developing more resilient sensor designs, leveraging artificial intelligence for data analysis, and creating multi-functional sensors. These advancements are crucial to ensuring the continued reliability and efficiency of power transformers, which are the costliest machine on the grid.

Actually, there are a large number of sensors available commercially. Since it might be too expensive to incorporate all of them, risk management methods are used to optimize their number given a specific application (operating conditions, location, importance/criticality, etc.).

Overall, this review paper not only highlights the advances but also sets the stage for future innovations in this critical field. The future of sensor applications in power transformers monitoring holds great promise. The ongoing trend towards the digitalization of networks and the development of digital twins strongly justify the use and advancement of new, more efficient sensors.