Thermal modelling of transformers

Power transformers are an important part of the electrical grid. Transformers in today's grid have been installed over a period dating back to the 1950s, and replacement has been moderate. Therefore, the average age of transformers in the grid is continuously increasing. In fact, for many of the older transformers, their service life exceeds their designed service life. Although several of the transformers are old, it is important to point out that the condition of the transformer park is generally good. The long service life can be attributed to the fact that the transformers have in many cases been subjected to a moderate load, and that the design margins were conservative due to inaccurate design methods. The goal of the project is to obtain accurate temperature predictions based on load, load history and ambient conditions. The goal of the project is to obtain accurate temperature predictions based on load, load history and ambient conditions. To meet the goals of this project, the focus is on modelling temperature rise and cooling in a transformer. The heat sources in the transformer are the electrical losses due to the current in the copper windings, the electromagnetic losses in the transformer core and in structural parts such as the transformer housing. In the project we have developed models for losses under normal operating conditions and these are compared with measurement results. These models are now expanded to include e.g. temporary overload situations. The challenge has been to obtain accurate results within acceptable calculation times, which we believe we have now gained control over. The PHD work in the project has focused on the losses in the iron core. Here, a test setup has been set up that measures losses that occur at high loads. The research fellow will defend his dissertation in 2021. Three papers have been accepted in journals and 2 papers are out for revision. A test setup has been developed that measures the oil flow between windings. The layout consists of 70 thermal elements which are assembled as a transformer winding in an oil-filled tank. The oil should be filled with nanoparticles to measure movement using Laser Doppler Velocimetry (LDV). The production of this setup has taken more resources than expected. Basic tests have been performed to examine the functionality, but thorough tests have not been performed in the project. Thermal modelling is performed by different methods with a decreasing level of detail. The most detailed and computationally challenging method is Computational Fluid Dynamics (CFD). CFD is well suited for measuring the oil flow around - and its cooling of - the copper windings. In order to reduce the calculation time of the problem, a project has been developed in the project that reduces the number of elements modelled. In this method, the oil channels between the windings will be modelled as a porous material. This gives accurate results with rough element breakdown. This gives more accurate results with rough element division. A paper describing the method has been published. Thermal network modelling is used to describe a full transformer. A lot of work has been done to achieve satisfactory convergence when calculating a realistic winding. We have had challenges in making a model that can count on entire windings in a transformer. A severe thermal overload can cause moisture, which is pushed out of the paper insulation, to form bubbles in the transformer oil. In the worst case, such bubble formation can lead to breakdown of the insulation and consequently greater failure. An experimental apparatus for studying the phenomena has been built, and the first experiments have started. Water bubbles have been observed on the electrodes at high temperatures. It has taken more resources than planned to make this equipment and we have only made simple measurements that confirm the operation of the equipment. A conference paper has been written describing the equipment. This paper will be presented at the IEEE Electrical Insulation Conference 2021)

Dette prosjektet har gitt SINTEF og NTNU større kunnskap om termiske forhold i krafttransformatorer. Prosjektet har utviklet ny metode for å modellere oljestrømmer rundt viklinger. Dette vil gi fagmiljøet større fleksibilitet for termisk modellering i fremtiden. Prosjektet har utviklet to testrigger som vil være bærebjelker i fremtidige prosjekt på denne problematikken. En testrigg for målinger av oljestrøm rundt viklinger. Dette prosjektet har gitt miljøet i Trondheim nå mulighet til teste hvordan oljen flyter under kaldstart av transformatorene, slik at man i fremtiden kan unngå risikoen for lokale overoppheting. Den andre testriggen vil tilføre forskningen mulighet til å se nøyere på hvordan fuktighet kan skilles ut av papirisolasjonen ved en kraftig termisk overbelastning. Dette har ikke blir testet før av andre.

Good thermal models for transformers is a requirement for life assessment of insulation, emergency overloading, risk assessment in connection with low frequency loading during solar storms (GIC), and several other aspects of transformer operation and management. There is also a need to improve of knowledge of thermal performance elder transformers. Although thermal models such as IEC loading guides exist, there is a need to improve several aspects relating to e.g. oil viscosity changes with temperature, particularly important for arctic climate, calculation of losses and ensuring applicability to older transformers. The project will focus on models for operation over a wide ambient temperature range, transient behavior and managing heat source distribution during abnormal operation, e.g. core saturation during GIC. The modelling approach will be based on existing thermal-hydraulic network models complemented by computational fluid dynamics studies of subsystems and finite element modelling of stray magnetic fields. Experimental work on scale models and field work on instrumented transformers will support and verify the models developed. Improved knowledge of thermal issues of transformers will increase benefits from previous research into e.g. aging through lifetime extensions, provide immediate improvements to load thresholds during emergency overloading without increased risk and also form a foundation for future understanding, e.g. of the effects of intermittent operation. Improvement of models for operation of power equipment is in line with the issues raised in the European Technology Platform's Smartgrids SRA 2035 asking for advanced monitoring and models for prediction of life time.