The ConCEPT project at the University of Southampton builds on completed elements of several projects, namely Resilient Networks (RESNET (NIA_NGET0053) – an IFI/NIA funded project at University of Manchester), FENCE (Facilitating Enhanced Network Capacity Evaluation, NIA funded project with University of Southampton) TeRMiTE (NIA_NGET0165) ( Transformer Rating Modelling Tool Enhancements, NIA project with University of Southampton) and Improved Transformer Thermal Monitoring (IFI funded project with Southampton Dielectric Consultants). The aim is to take the findings from several of these projects and make them readily implementable.The specific areas from these projects that should be advanced are shown in the attached detailed project proposal.
Benefits
This project will be successful if the work enables the following:
- proposals for seasonal ambient conditions for static ratings taking into account urban heat islanding
- an evaluation of the likely accuracy of predictive ratings
- an indication of reduction of thermal performance resulting from transformer degradation
- a model for assessing WTI performance and identification of defective instruments
- a review of the application of alternative algorithms for transformer condition based on thermal performance
- a view on the likely effects of climate change on static ratings for transformer until the end of the 21st century.
Learnings
Outcomes
The ConCEPT project had 3 overall objectives, as follows:
- to develop understanding of phenomena that could materially affect the operation of transformers in the coming decades
- to explore how greater value that can be extracted from existing data and potentially new sensors to optimise maintenance interventions, and
- ensure that we make the most of the thermal capability of ageing transformers without compromising their reliability.
The key outcomes of the project will be discussed with reference to these objectives in turn.
Climate change was identified as a phenomenon that could impact transformers in earlier NIA projects. ConCEPT used the latest climate projections from the Met Office to produce recommended changes to ratings seasons moving forwards. From calculations performed, a decrease in the continuous rating between 0.01 to 0.02 per-unit is estimated between decades at all locations considered if existing conservatism is to be maintained. Based on conservative ambient temperatures calculated from the climate projections the following changes were proposed to the ratings seasons:
- Moving September – September is currently considered to be part of the autumn season for ratings (and therefore assumed to have an ambient temperature of 20 °C) but should be moved to the summer season with an assumed ambient temperature of 30 °C.
- Increasing winter conservatism – The conservative ambient temperature for winter should be increased from 10°C to 15°C if it is to be considered as a temperature which is rarely exceeded for the warmest regions in the UK.
- Introducing low summer – a new season covering May and October with a conservative ambient temperature of 25°C could be introduced.
- Introducing high summer from 2050 – The conservative ambient temperature assumption of 30°C for summer is sufficient for the warmest months until the 2040s. After this it could be prudent to introduce a new season, high summer, with a conservative ambient temperature of 35°C.
The ratings seasons are currently applied uniformly across England and Wales and are intended to be appropriate at the most thermally onerous locations. ConCEPT demonstrated that there may be opportunities to reduce the ambient temperature assumptions in some regions thereby increasing ratings especially during the Summer and Spring/Autumn seasons. Taking account of climate change, investigations suggest that increases to transformer ratings would still be permissible in some seasons for the next few decades in much of Wales and the North of England.
Historical data were utilized in several ConCEPT deliverables. Algorithms were developed to automatically identify faulty winding temperatures indicators (WTIs) and identify poor performance in cooling systems. Machine learning techniques were also trained against historical measurements to develop alternative thermal models. The performance of these algorithms was typically superior to established thermal network models with tests performed over 23 WTI, load and ambient temperature datasets. Furthermore, the performance of the machine learning techniques was slightly improved by including sensor data of wind speed and solar radiation from the substation weather station. In addition to machine learning algorithms a new thermal network model has also been developed using the sensor data. This model can capture influence of solar radiation and wind speed which has been shown to impact hot spot temperatures for some units. In addition to thermal modelling, analysis of historical DGA data has found that for a particular unit the concentration of key gases, including ethylene and hydrogen, was found to increase during periods of high WTI temperatures.
The use of climate projections to inform ratings seasons, and the development of novel transformer thermal models, both relate to optimising the thermal performance of transformers. In addition to the outcomes discussed previously, investigations undertaken as part of Work Package 4 examined the potential for change in transformer thermal parameters due to ageing. It was found that there was a significant change between the original parameters determined from the heat run test and the parameters fitted to operational data.
Recommendations for further work
Utilising Fiber Optic Hot Spot Measurements
An issue with using WTI measurements is that they are not a direct measurement of hot spot temperature on the winding. Faulty heating elements could lead to erroneous measurements which could contaminate analysis. Many transformer windings now have fibre optic cable installed which are currently only used during temperature rise tests of type-test units in a transformer family but would enable an alternative direct temperature measurement in service.
NGET has previously attempted to revisit transformers in service to measure winding temperatures using installed optical fibres and found readings difficult to obtain. It was intended to collect data from these transformers as part of this project, but with previous difficulties there was no reliance on this, due to the effect of COVID-19, it could not be explored so fibre optic temperature measurements remain elusive. Therefore, the long-term reliability of optical fibres as thermal sensors and the opportunity they may provide as an independent check of WTI temperatures in studies is still to be investigated.
Analysing Data from Thermally Stressful Events under OFAF (oil forced, air forced) Conditions
In order to use machine learning algorithms to calculate ratings it is vital to obtain appropriate training data. Much of the work conducted throughout ConCEPT has clearly demonstrated the improved performance of machine learning techniques over traditional thermal network models. If appropriate training data could be obtained, then such algorithms could be used to update transformer ratings.
Such datasets would be difficult to obtain, as rated loads are rarely used in operation, but they would deliver benefit if a method could be found for generating them in practice.
Developing First Principles Transformer Thermal Models
A new approach to developing ratings algorithms would be to construct a first principles model of heat transfer within the transformer in conjunction with targeted sensor installations. This would certainly require technical drawings of the windings, tank and cooler banks, in addition to multiple thermocouple arrangements. The intention is to construct several finite element models that can be validated against operational data. When there is a high degree of certainty in the finite element model it may then be used to calculate ratings. These models could also be used to inform a low computational cost thermal network model.
This work overlaps with another National Grid NIA project Economic Ageing of Transformers (NIA_NGTO038) with the development of electromagnetic models to challenge the 1.5 p.u. short term rating limit. A potential future project could focus on a single transformer, with the intention of developing models that match reality in detail, once this was achieved for a single transformer it is anticipated that translating the methodology to other units would be relatively straightforward.
Management and Storage of Historical Data
ConCEPT has demonstrated there is value in co-ordinating operational data for transformers within a system for tracking condition. Many of the algorithms to determine transformer thermal parameters by fitting to data, or to detect WTI issues, are easily automated and have a low computational cost. If a centralised database existed software could be developed to interrogate the data and flag key results to a user.
On a related note, transformer WTI and load data is only available in eDNA for the last 7-8 years. This project shows that deterioration of thermal performance of assets in service for more than 50 years occurs over an extended period of time and that there would be benefit in storing data from the point of installation where possible. This would allow long term trends in the thermal performance of the transformer to be identified.
On-site Measurements of Thermal Parameters from Aged Units
Many of the investigations performed in ConCEPT indicated that the thermal performance of transformers has changed significantly since the original heat run test, which for some units could be over 40 years ago. On-site measurements of transformer thermal parameters, by repeating heat run tests, would test whether these indications are correct.
Sensor Installations
The sensor installations undertaken as part of ConCEPT were successful as a whole, gathering data from both substation weather stations and thermocouple measurements installed on transformer units. An unresolved issue was the data recorded by the flow sensors, which was consistently spurious and could not be used.
For future sensor installations it would preferrable to undergo one installation in advance and provide scope for return visits to resolve potential issues, before undertaking similar installations at other sites.
Lessons Learnt
The following key points have arisen from the work conducted so far:
- Climate projections in UKCP09 show that the existing rating seasons may need to be revisited, as the assumptions used for the summer and autumn months are likely to cease to be conservative during the 2020s.
- Malfunctioning winding temperature indicators (WTI) have been successfully detected through an analysis routine which could be run online; further work is planned to determine how this could be embedded into routine processes.
- Analysis conducted using UKCP18 reinforced earlier findings using UKCP09. It was also demonstrated that introducing regional ratings seasons could be used to offer ratings enhancements in several locales.
- Alternative ratings algorithms were investigated, including neural networks, which showed improved performance against traditional thermal network models in predicting hot spot temperatures during thermally stressful events. An issue is the lack of data, used to train such algorithms, where the rated load is used.
- A range of investigations using machine learning algorithms demonstrated that they have the capability to detect anomalous temperature measurements. This suggests they could be utilised as a low-cost condition monitoring tool.
- The sensor data gathered in WP1 (sensor installation), has clearly demonstrated that the hot spot temperatures of some units are dependent on wind speed and solar radiation. The data collected has been used to construct novel transformer thermal models.
- Thermal simulations have demonstrated that online reclamation is feasible which is as expected based on existing industry practice. The investigations highlighted the need to understand the oil flow field in the tank during reclamation.
The following deliverable reports have been prepared, at the end of the project a summary of these will be available for GB licensed network operators to access upon request:
- Technical note on UKCP09 predictions and impact on transformer ratings
- Report on existing rating seasons used for transformers, including recommendations for future changes.
- Report on the extent to which transformer cooling and WTI issues can be identified via data analytics.
- Report on alternative algorithms which could be used to calculate transformer ratings.
- Technical note on the development of predictive ratings for transformers, identifying key enablers and resulting benefits
- Report on urban heat islanding and its impact upon transformers
- Technical note on UKCP18 predictions and impact on transformer ratings
- Report on site data collected, identifying further analysis to be conducted in other WPs, lessons learned, value derived from sensor data and key recommendations for future sensor installations
- Report on the creation of transformer rating algorithms from recorded data collected under WP1
- Report on the ability of alternative algorithms to calculate ratings, identify WTI and cooling problems and correlations with DGA data
- Report on the revalidation of the transformer thermal model using site data
- Report on the impact of online reclamation on the thermal behaviour of the transformer
Dissemination
The following conference papers have been published:
A. Doolgindachbaporn, N. H. Nik Ali, G. Callender, J. Pilgrim, P. Lewin and G. Wilson, "Detection of Forced Cooling Faults in Power Transformers based on Winding Temperature Indicator and Load Data," 2019 IEEE Electrical Insulation Conference (EIC), Calgary, AB, Canada, 2019, pp. 111-114. https://doi.org/10.1109/EIC43217.2019.9046583
A. Doolgindachbaporn, G. Callender, J. Pilgrim, P. Lewin and G. Wilson, "The use of thermal and load data to identify large autotransformers that have aged and degraded electrical insulation," 2020 IEEE Electrical Insulation Conference (EIC), VIRTUAL, pp. 313 316. https://doi.org/10.1109/EIC47619.2020.9158731
A. Doolgindachbaporn, G. Callender, P. Lewin, E. Simonson and G. Wilson, "A Top-Oil Thermal Model for Power Transformers that Considers Weather Factors," IEEE Transactions on Power Delivery (submitted).
G. Callender, A. Doolgindachbaporn, P. Lewin and G. Wilson, “Impact of Climate Change on the Power Flow Capacity of Transformers,” CIGRE Symposium 2021 (abstract accepted, paper submitted).
A. Doolgindachbaporn, G. Callender, P. Lewin and G. Wilson, "Estimation on Degradation Rate of Insulating Paper in Power Transformers Using Historical Load and Thermal Data," 2021 IEEE Electrical Insulation Conference (EIC), VIRTUAL (abstract accepted, paper submitted).
A. Doolgindachbaporn, G. Callender, P. Lewin, E. Simonson and G. Wilson. "Data Driven Transformer Thermal Model for Condition Monitoring" IEEE Transactions on Power Delivery (submitted).
A summary of the project was presented to an invited audience of transformer specialists across the industry at a transformer research dissemination event organised by the University of Manchester in December 2019 and at the EuroDoble 2020 colloquium in October 2020.