The application of laboratory studies to in-service transformers is often only indicative rather than leading to accurate assessment of remaining asset life. Challenges with interpretation of chemical markers measured through oil sampling of in-service transformer are uncertainties in production, partitioning and stability.
These uncertainties tend to make it harder to predict when transformers have reached end of life until they are already almost in that state. More predictability of degradation over a longer time horizon will enable more efficient planning.
This project aims to improve our understanding of insulation ageing for in-service assets and to make better assessments of when wound plant is likely to reach end of life.
Benefits
The project delivers greater understanding of how to determine the condition of transformers and other wound plant from ageing markers found in the oil.
Specific guidance can be developed for interpretation of ageing markers enabling transformer replacements to be planned more efficiently over longer periods of time.
Learnings
Outcomes
ADAPT
A review of the literature on studies of furans in oil and esters revealed many studies on laboratory-based correlations with DP of paper but little systematic study of samples from transformers evidencing the need for more work in this area.
Prolonged storage of oil samples leads to a reduction of furan concentration, reinforcing the need for prompt analysis and calling into question some historic results in National Grid’s database.
Recommendations for further work
None at this stage
Lessons Learnt
ADAPT
An overview of plausible mechanistic pathways for the degradation of cellulose identified new molecular species that could potentially be used for tracking the deterioration of paper insulation. GCxGC-MS shows potential for identifying these in practice, especially those lacking a UV chromophore which would not be found by HPLC-UV.
Direct sample injection of ester samples into the HPLC for quantification of furans is necessary as esters are miscible with the solvent used for liquid-liquid extraction of furans from mineral oil. This was successful but requires further optimisation due to overlap of the furans with matrix components.
It is clear that furans correlate with DP reduction as reported in the literature, but secondary reactions of furans influence the levels found and can cause an underestimation of the depolymerisation of cellulose.
As an alternative to HPLC-UV analysis of furans on a routine basis, UV/visible spectroscopy may be a more rapid option.
TREND
Classical statistical models, compared with more modern artificial intelligence (AI) modelling, perform well with the level of data available for understanding transformer lifetime. AI will have a place but the quantity of data required for training is not currently available.
While the kinetic based modelling may contain various uncertainties using a published standard thermal model and hot-spot factors derived through reversed engineering combined with temperature and water content dependent Arrhenius equation, the thermal lives of 106 National Grid in-service transformers have been estimated. As a result, the thermal life expectancy of the population is derived as 84 years. It has been identified that the modelled transformer’s load, winding-to oil gradient, and top-oil temperature rise are three possible error sources in the hotspot factor derivation and hence, the thermal model.
Dissemination
Two papers have been submitted to conferences based on the work in TREND:
“Development of a Steady-State 2-Furfural production and partitioning model” has been submitted for publication and presentation at the International Symposium on High Voltage Engineering (ISH) in Nagano, Japan in August 2025.
“Studies on the partitioning of Furfural between liquid and paper insulation” has been submitted for publication and presentation at IEEE International Conference on Dielectric Liquids (ICDL) in Lodz, Poland in May 2025
