Conference Agenda

This paper presents a comprehensive evaluation of mineral oil-filled power transformers and reactors by utilizing various diagnostic frameworks, including the Duval Triangle, the Duval Pentagon, IEEE, IEC, and CIGRE TB guidelines. The study investigates the efficacy of Dissolved Gas Analysis (DGA) in condition assessment and its complementary role alongside conventional electrical testing during maintenance and research activities. Through diverse case studies involving field maintenance, workshop inspections, and laboratory fault investigations, the correlation between electrical test results and DGA findings is examined. Cases showcasing both consistency and discrepancies between these methods are analyzed to delineate the boundaries of DGA-based fault localization. Furthermore, the role of DGA as a primary trigger and supportive tool for fault detection is emphasized, illustrating its impact on the optimization of maintenance strategies. Leveraging an extensive dataset from real-life applications, this article establishes a baseline for comparing the diagnostic accuracy and statistical success of established DGA methodologies.

Since the global population of high-voltage direct current converter (HVDC) transformers is constantly growing, their reliable operation becomes more and more critical. The present paper aims to enhance the reliability of the results of the condition assessment of HVDC transformers using DFR (Dielectric Frequency Response) analysis and, consequently, to increase the overall reliability of HVDC transformers.

DFR (Dielectric Frequency Response) measurement is a non-intrusive, non-destructive off-line testing technique developed as an advanced diagnostic tool for transformer insulation condition assessment. One important milestone for wide acceptance of DFR was the publication of IEEE Std C57.161-2018 "IEEE Guide for Dielectric Frequency Response Test" [1] in November 2018. The Guide includes recommendations for instrumentation, procedures for performing the tests and has proven its effectiveness in relation to transformers of various applications and voltage levels. However, when it comes to HVDC converter transformers, this document may confuse the readers / assets owners.

One example, there is an error in the recommendations of the Guide considering three-winding transformers if it is related to one-phase three-winding HVDC transformers (with one line winding and two valve windings). According to the chapter 5 of the Guide, it is recommended estimating moisture content in cellulose based on measurement results obtained from insulation between valve windings. In fact, different valve windings are usually located on different limbs of the transformer core and therefore the insulation between these windings cannot be represented as a cylindrical capacitor. This means that implementation of the “X-Y model” described in the Guide will lead to erroneous results (examples will be given in the paper), and the measurement cannot be used for moisture in cellulose estimation. Therefore, the determination of moisture content in cellulose insulation of one-phase three-winding HVDC transformers is only possible based on DFR measurement of insulation between the line winding and the valve winding.

Another example, chapter 7 of the Guide is very much focused on interpretation of UST (ungrounded specimen test) measurements and provides no guidance for GST (grounded specimen test). Meanwhile, usually valve winding bushings of HVDC transformers are located inside a valve hall and line winding bushings – outside. Due to this reason, it is often not feasible to connect a DFR instrument to both windings simultaneously and then the UST measurement is not applicable.

For cases when GST measurement is the only option, the interpretation of DFR results can be based on a comparison between the measured trace and a reference (time-based, type-based or phase-based comparison). An approach to analysis of GST measurement results and examples of successful condition assessments of HVDC transformers using DFR will be given in the paper.

The results of this work will be reported to the IEEE working group C57.161 which has been formed to conduct revision of the DFR Guide.

REFERENCE:

[1] IEEE Transformers Committee. “IEEE Guide for Dielectric Frequency Response Test” (IEEE Std C57.161-2018).

Gas insulation breakdown voltage reductions of up to multiple tens of percent are common within the first breakdowns. These breakdowns seem accepted in laboratory

investigations, but add experimental complications and additional uncertainty during evaluation. Reliable methods for preventing such a reduction have not been established, but would be highly welcome. Therefore, this work quantifies the impact of electrode surface contamination on the breakdown voltage and outlines a practical and transferable mitigation strategy. First, laser scanning microscopy is used to characterize possible contamination, such as environmental particles, cloth fibers, and dried solvent residues. Second, a three-step chemical cleaning protocol is described, which universally removes most of any contamination. It is measured that its application increases the mean breakdown voltage of the first twenty consecutive breakdowns and reduces the breakdown voltage scatter. Third, the effect of residues possibly resulting from the cleaning method itself and the electrode storage duration before experimentation on the breakdown voltage is shown. This is intended to improve the laboratory practice and improve the comparability of gas insulation breakdown measurement data.

Introduction

Power transformer alarm systems must carefully distinguish between chemical indicators of genuine failures and benign phenomena that can elevate gas readings and trigger unnecessary interventions. Accurate interpretation is essential for reliable operation. Unnecessary maintenance activity — and even premature, expensive transformer replacement due to improper interpretation — can be as costly as a real failure that actually damages the transformer.

Simplified Gas-in-Oil Testing Framework

This study introduces a streamlined approach for gas-in-oil analysis that maximises true alarms while minimising false positives. The method focuses only on gas species that most directly reflect dangerous energy release and are least influenced by stray gas formation. This targeted selection improves both diagnostic accuracy and efficiency.

Diagnostic Priority of Gases

Not all gases are equally important:

• Acetylene: The most decisive marker of high-energy electrical stress.

• Hydrogen: A sensitive early warning indicator, but it must be confirmed by other gases. It is also prone to false alarms caused by stray gas sources.

• Ethylene: Indicates sustained thermal stress.

• Oxygen: Reveals possible oxidation or air ingress and constrains interpretation by indicating when other readings may be biased.

• Other hydrocarbons: Provide contextual support for the overall diagnosis.

• Nitrogen: Helps assess air ingress and dilution.

• Carbon monoxide: Considered only as a last resort due to its low specificity under normal oil oxidation.

Decision Rules

The method integrates absolute thresholds, temperature-consistent rates of change, and co-occurrence patterns of gases, with modest hysteresis and hold-off logic to prevent unstable alarm chatter. This reduces unnecessary boundary crossings and improves alarm reliability.

Operational Efficiency

By monitoring fewer but more failure-relevant, high-value gases, the process requires fewer sampling steps and fewer measurement pathways. This reduces handling and calibration errors, shortens turnaround time, and improves pinpoint accuracy. When implemented in an online instrument that measures only these necessary species, the approach can also lower operational costs.

Validation and Outcomes

Field validation shows that selective monitoring reduces false alarms without sacrificing early sensitivity. It also improves prioritization by helping operators focus attention and resources on units that are genuinely at risk across large fleets.

Conclusion

By aligning alarm decisions with the gas species that are most diagnostic of real fault physics and least susceptible to stray gas phenomena, this methodology enables clearer, faster, and more trustworthy fault identification in power transformers.

The number of transformers filled with natural and synthetic ester liquids has been steadily increasing in recent years, reflecting what can be seen as an industry trend to move toward more environmentally friendly insulating fluids. This trend is more evident in sectors such as industrial applications, renewable energy systems, data center power infrastructure, and electricity distribution networks. In some of these segments, the market share of ester-filled transformers is already estimated to exceed 10% in the United States, according to industry sources. Despite this growing adoption, the application of dissolved gas analysis (DGA) as a diagnostic and condition-monitoring tool for ester-filled transformers has not yet achieved the same level of maturity as it has for traditional mineral oil-filled units. Current IEEE and IEC standards provide some guidance, highlighting both similarities and differences in the interpretation of DGA results between ester liquids and mineral oils. However, many of these differences remain under active investigation, and the assessment criteria are still evolving. This work presents a set of insights derived from the analysis of a database comprising approximately 55,000 laboratory test results of dissolved gas content in samples collected from transformers filled with natural from one large laboratory in the USA. The study explores several aspects of the assessment, including the significant differences in the ratio oxygen/nitrogen compared to mineral oil samples, the observation of a strongly bimodal distribution in nitrogen levels, and the contents to be expected for some of the gases under normal operating conditions. Furthermore, the paper discusses the estimated generation rates for individual gases and introduces a novel methodology for calculating a Fault Energy Index. These findings contribute to the efforts to advance diagnostic practices and improve the reliability of condition assessment for this growing class of transformer insulation systems.

Moisture measurements in transformer oil are often perceived as erratic, particularly when the transformer experiences significant load variations. This behavior is primarily driven by two factors: changes in the moisture saturation level of the oil and the migration of water from the paper insulation into the oil. In mineral oil-filled transformers, approximately 99% of the water resides in the paper insulation. While moisture in paper is expressed as a percentage, and a water content of 0.5% represents a dry insulation, the moisture in oil is measured in mg/kg or ppm, and its value for a new unit is in the range of 10ppm. Consequently, even minor changes in the water content of the paper can result in substantial fluctuations in the oil moisture readings.

This work presents an enhanced application of the Thermo-chemical Digital Twin, introduced at this conference in previous years. The model has undergone relevant improvements in its transient water diffusion calculations. For this application we also implemented some simplifications in the discretization of the active part (core and coils) to reduce the amount of required input data. By leveraging this model, it becomes possible to distinguish between moisture variations caused by water migration and saturation changes, versus actual increases or decreases in the transformer’s total water content. This approach effectively “denoises” moisture-in-oil measurements, isolating the underlying processes and enabling users to make better-informed decisions regarding transformer condition.