Active Detection and Identification of Incipient Faults in Transformers

Presented By:
Emilio Morales Cruz
TechCon 2017


The failure of power transformers is always an area of significant concern because it can result in millions of dollars in costs, interruption of power, and possible environmental and safety impacts. Therefore, it is desirable to detect the existence of abnormal changes in the transformer’s internal condition and determine whether the changes could lead to an incipient failure. Active detection and identification of incipient faults is now possible through on-line monitoring of abnormal changes in some parameters, and the use of diagnostic methods provides an assessment of equipment condition as well as suggested actions. This paper gives an overview of a transformer monitoring system and measured signals for the diagnosis of the failure mechanism.


Online Monitoring

Online monitoring of substation assets is becoming an essential feature of electric utility systems. The justification for online monitoring is based on the need to increase the availability of substation assets, provide condition assessment and life management, move from time-based maintenance (preventive) to condition-based maintenance (predictive) and failure cause analysis.

Online Monitoring of Power Transformers

System outages due to failures in transformers have a significant economic impact on the operation of the power grid. Therefore, it is necessary to be able to do a good condition assessment of transformers. Techniques to diagnose integrity through non-invasive tests can be used to optimize the maintenance effort and ensure maximum availability and reliability. With the increase of the average age of the transformer population, there is a growing need to understand their internal state. For this purpose, online and offline methods and systems have been developed in recent years. Online monitoring can be used continuously during operation of the transformer and thus offers the possibility of timely recording of different events that can affect the life of the transformer. The automatic evaluation of this data allows early detection of future failures. In contrast, offline methods require an outage of the transformer and are mainly used during regular inspections or when the transformer has already been identified as having problems.

Since the 90s the online monitoring systems for transformers have been developed to the point that they can assess the condition of all the main parts of the transformer. These techniques cannot prevent the failure of transformers, but can prevent costly impacts associated with failures of transformers by allowing the user to take corrective measures during their operational life.

Total Transformer Monitoring System

A comprehensive monitoring system provides a complete view of the health of the transformer through the monitoring of individual parameters and the creation of simple models to compare expected performance vs. actual performance. This type of monitoring makes possible to notify the user if a limit value was reached. The user can then act according to preset company procedures.

A complete monitoring system shall integrate all relevant major components of the power transformer within a single system. The transformer database, algorithms, expert system and diagnostic function can all be in one intelligent electronic device. The correlation of the transformer data is then possible because all the data is in a single database. The interpretation of data provides information on the health of the transformer and all its components. The expert system and diagnostic functions help users make the right decisions regarding the future operation and maintenance of the power transformer.

Failure Statistics

The latest transformer reliability study (1) shows the windings with 45% as the major cause of failing transformers, followed by tap changers with 26%, bushings with 17% and lead exits with 7%. All other major components of transformers play a minor role in failure statistics (see Figure 1). Looking into catastrophic failures (3), bushings are the cause in 70% of the cases. There are other components that cause transformers to fail but with less risk of a catastrophic failure.

Latest transformer major failure statistic

Monitored Parameters Overview (2)

Each incipient fault in a transformer will somehow generate detectable signs of its appearance. These signs could be chemical, electrical, optical or acoustic in nature, but most of the time a combination of these. For new transformers being manufactured today, we find dozens of analog monitoring devices and several electronic monitoring solutions (Figure 1). While it may seem overwhelming, present day technology for monitoring of large power transformers (LPTs) can be narrowed down to eight main types of parameters:

  • Temperatures (top oil, bottom oil, ambient, winding hot spot and core hot spot)
  • DGA (gas formation and accumulation, total percent dissolved gas-in-oil)
  • Partial Discharge (PD)
  • Moisture (moisture in oil & sample temperature)
  • Cooling system parameters (top and bottom oil temperatures, fan and pump motor currents, pump flow status, cooling control and supply voltage)
  • Oil preservation system (Oil level, conservator tank bladder rupture, internal pressure)
  • Currents (line, excitation, core ground and GIC currents)
  • Bushing parameters (power factor/ capacitance)
  • OLTC parameters (position/ timing/ motor current/torque/temperature)


Besides the conventional temperature sensors, used to measure the top and bottom oil temperature, fiber optic sensors for direct measurement of the hot spot temperatures are now being more frequently used. Together with top oil and bottom oil temperature measurements, they provide a highly accurate thermal model for any transformer. They provide the foundation for calculating the actual dynamic rating for a transformer. Furthermore, they represent one of the bases for aging and moisture models as well as the bubbling temperature model. These measurements are instantaneous and complement DGA and PD.

Dissolved Gas Analysis

These sensors, either based on GC (Gas Chromatography) or PAS (Photo Acoustic Spectroscopy) technology, provide accurate ppm values of key gases dissolved in transformer oil. The resulting DGA model is now mature after almost two decades of application in the industry. More advanced DGA assessment models are nowadays providing condition assessment based on the combinations and concentrations of the different gases resulting in easier to understand information for the user. The analysis is often delayed by an hour to tens of hours depending on internal conditions and also the distribution ability of the gas inside the oil. The model only identifies the type of fault and requires certain additional analytics to identify the source of evolving problems. The use of other sensors to confirm and complement this model is the key to comprehensive monitoring.

Partial Discharges (PD)

Although relatively new in transformer application, UHF partial discharge techniques have proven themselves for accuracy and reliability on other electric assets (e.g. GIS) for almost two decades. UHF sensors are matched for continuous online monitoring, and offer more sensitivity when compared to Acoustic and IEC 60270 PD methods. The UHF PD monitoring has been proven to identify evolving problems much earlier than any other monitoring method. PD pattern recognition using ANN (Artificial Neural Networks) and Parameter Analysis can be verified by a DGA model to provide excellent reconfirmation. For transformers with three or more sensors, time of flight method can be used to localize and find the source very accurately. Localization of a PD source is one of the most important pieces of information in order to assess the condition of a transformer, its evolving defects and comprehensive risk assessment as well as to decide if the problem can be fixed on site or if the transformer needs to be replaced.

Monitoring parameters overview


A separate moisture sensor is usually combined with dissolved gas sensor. Simple reading of moisture in oil is used to create moisture-in-oil model. By comparing the trends of moisture in oil with actual transformer load and thermal model and by the use of moisture-in-insulation model, the moisture in insulation can be estimated and its influence in the acceleration on the cellulose aging rate, reduction in dielectric strength, bubble formation and dielectric failure, and partial discharges in the insulation can be determined. These models are relatively less accurate when compared to DGA, thermal and PD models. However, combined they accurately portray remaining life calculations and actual dynamic rating for a transformer.

Cooling System Parameters

The operation and efficiency of the cooling system efficiency are determined by the measurement of the cooler inlet and outlet temperature. Often it is also enough to use simply the top and bottom oil temperature. In order to detect evolving defects, usually the fan and pump currents and the flow are monitored. Together with the cooling efficiency model, a problem with the cooling system can be detected and the fans and pumps can be controlled accordingly.

Oil Preservation System

The monitoring of oil levels, conservator tank bladder rupture, condition of regenerating breather and tank internal pressure provide early detection of issues with the oil preservation system. Oil levels together with a thermal model can provide a much earlier indication of an oil leak. A conservator tank bladder rupture and a silica regenerating breather activated by relative humidity level can provide an indication of a malfunction in the conservator tank preservation system. An abnormal internal pressure will also indicate a malfunction of sealed, inert gas or conservator tank preservation systems.


Line current sensors provide a parameter that is used as an input for the winding hot spot temperature simulation by thermal image, thermal model, and to verify load dependency with other parameters. Excitation current gives an indication of a change of the magnetic circuit possibly due to through faults. Its measurement cannot be done continuously, but can be done online whenever there is no load on the transformer. Core ground current provides indication of the connection status as well as unintentional core grounds. The monitoring of GIC currents serves to validate the modeling of GIC flows in transformers as well as to verify dependency with other parameters.

Bushing Parameters

Accurate power factor and capacitance measurements are possible using the phase shift method. Comparing leakage current with a reference signal from the same phase is the key to a reliable bushing condition assessment. Besides drastic changes in terms of partial breakdowns or power factor, the accuracy of this method allows the detection of small power factor changes, even the detection of evolving moisture. Certain analytic and correlation models allow distinguishing between different types of defects. The PD monitoring method, using the bushing as coupling capacitor (IEC 60270 method), is less efficient for bushing condition assessment because it is difficult to distinguish the origin of a discharge (external/ internal) and because of the normally low level discharges of a bushing compared to other discharge sources.

On Load Tap Changer Parameters (OLTC)

The main parameters for a successful condition assessment of a LTC are tap position, switch count, switching time, motor current, motor torque, temperature difference between the main tank and OLTC compartment.


Each incipient fault in a transformer will generate detectable signs of its appearance. These signs could be chemical, electrical, optical or acoustic in nature, but most of the time a combination of these. The active detection of these signs and their combination may lead us to identify the incipient fault. The following are some examples of possible faults and the useable parameters for their early detection:

Overheating of Laminations and/or Core Joints

This phenomenon may take place if the forces due to a through fault were able to displace the upper core yoke, resulting in higher losses at the core joint, an increase in temperature that may or not be detectable by the top and bottom oil sensors but possible by a core hot spot sensor and there will be a change in the pattern of combustible gas generation indicating more likely hot metal gases. Depending on the fault severity there could be gas accumulation in the gas accumulation relay, and it would be expected not to have correlation with core ground current, partial discharge, winding hot spot temperature, line currents and parameters used to monitor the cooling system.

Unintentional Core Ground

This phenomenon may occur if the magnetic circuit gets grounded in a place other than the intended core ground connection. This will lead to a circulating current through the transformer tank. This circulating current will be detectable, and depending of the fault severity, there could be a small change in the top and bottom oil temperatures. The combustible gas generation will indicate more likely hot metal gases with presence of cellulose overheating due to the nature of the magnetic circuit insulation system. Depending on the fault severity there could be gas accumulation in the gas accumulation relay, and it would be expected not to have correlation with excitation current, partial discharge, winding hot spot temperature, line currents and parameters used to monitor the cooling system.

Partial Discharge in Winding Insulation

This phenomenon may occur if the dielectric withstand capability of the winding insulation system has been exceeded due to possible insulation ageing combined with a through fault or due to moisture contamination. Partial discharges will be detected in real time and their pattern may indicate an insulation defect. The combustible gas generation will indicate partial discharge with presence of cellulose overheating, there may or may not be gas accumulation in the gas accumulation relay. There could be correlation with moisture in the insulation and there should not be correlation with any of the other parameters measured. As mentioned before, once partial discharges have been detected and identified it is very important to determine the source location to decide if the problem can be fixed on site or if the transformer has to be replaced.


The power industry today demands a more comprehensive monitoring approach in order to increase the reliability and availability of electrical assets, allow the application of condition based maintenance, detect the existence of abnormal changes in the transformer’s internal condition and determine whether the changes could lead to a failure, and improve the assets use efficiency. Modern monitoring solutions support monitoring of several parameters, include analytics models with diagnostics capabilities, which allow active detection and identification of incipient faults in transformers.


  1. A2.37, CIGRE WG. Transformer Reliability Survey: Interim Report, No. 261, ELECTRA. 2012.
  2. IEEE Std C57.143™-2012 “IEEE Guide for Application for Monitoring Equipment to  Liquid-Immersed Transformers and Components”
  3. H.-P. Berg, N. Fritze. RELIABILITY OF MAIN TRANSFORMERS. Salzgitter, Germany : Bundesamt für Strahlenschutz, 2011.

Join our email list

We use cookies to give you the best online experience. By using this website you agree with our cookie policy.