• Autor: Jacek Smołka
  • Tytuł pracy: Thermal process analysis in an electrical transformer dipped into polymerized resin
  • Opiekunowie: Prof. dr hab. inż. Andrzej J. Nowak, Prof. Luiz C. Wrobel (Brunel University, West London, U.K.)

The electrical transformers are nowadays commonly used to provide an appropriate electrical supply for many machines used in industry. In many cases these transformers operate in quite rigorous conditions which result in difficulties to cool their coils bellow acceptable level of temperature. For instance, in mining transformers may specifically be used to power the combined cutter loaders. However, in the mining environment many strict conditions need to be fulfilled. Typically all devices must be fire resistant and also impervious to high humidity. For this reason transformers have to be placed in hermetic containers, which make their cooling very challenging. Number of companies began investigation of this cooling process using a scaled down transformer. Additionally, in order to fulfil these difficult operating requirements the model transformer has been put into a metal container and dipped into polymerized resin. It was expected that after these endeavours, the transformers might be less efficient at heat dissipation and also the maximum temperature of the winding insulation could be exceeded. Hence, in order to achieve an acceptable working temperature inside the transformer, a cooling system consisting of a steel cooling coil is planned to be attached to the bottom of the transformer. Detailed heat transfer analysis, based on the mathematical model of heat transfer processes within the transformer, is required to design properly cooling system, at the same time minimizing the number of prototypes.

In order to obtain all required data to build up the mathematical model and to be able then to verify it, the number of electrical and thermal operational measurements have been performed. The measurements of the temperature field were performed in order to be verified in a simulation of the transformer under normal operation, and the measurements of the electrical parameters were performed in order to determine the values of the heat generated in the windings and the core. During these experiments, the transformer was cooled using both natural convection (in ambient air) and forced convection (a water cooling system attached to the bottom wall). As the result, analysis of the transformer without water cooling system (Model 1) and with water cooling system (Model 2) in both steady-state and transient state, was carried out.

Additionally to the thermal and electrical measurements performed within the project, exact material properties should also be known. Unfortunately, only the thermal properties of the resin were provided by the supplier, whereas other properties had to be obtained from the standard literature.

In this research two different three-dimensional geometric models were created. In the case when the transformer had no water cooling system it was possible to generate a mesh of very high quality. Almost the whole grid was structured and only at 'the corners' of the windings a little irregularities of the grid cells occurred. In Model 1 there are two vertical planes of symmetry; therefore, this model was reduced and only its one-fourth was considered. However, attaching a water cooling system to the bottom wall meant that a different mesh should be generated. The complicated construction of the cooling system required a lot of changes in subvolumes of the transformer's geometry. Therefore, it was necessary to create many kinds of mesh. That is why two kinds of mesh were considered. These two models were made based on two different criteria. In Model 2 there is a high degree of regular elements, but the others are of poor quality (more degenerated). In Model 3 there are less regular elements, but the overall mesh quality is improved.

Model 3 was carried out because of two main goals. The first was to be confident that during the mesh generation process it is important to limit the number of the most degenerated mesh elements. Second goal was to check what the results will be obtained if the number of the mesh nodes was dramatically limited (mesh-dependence).

In order to build the mathematical model a commercial Finite Volume Method (FVM) package, known as Fluent, was used. This CFD code is a very useful tool when analysing the temperature, pressure and velocity fields of many industrial apparatuses and machines, such as transformers. Software solves energy equations for both, solids and fluids, as well as momentum and continuity equations for fluids.

Additionally to determine the temperature distribution within the transformer, it was necessary to complete the heat equation with appropriate boundary and initial conditions. On the boundary surfaces, heat transfer was governed by free convection because experiments were carried out inside the building without any air motion while forced convection took place within the cooling coil. A constant temperature was prescribed in some surfaces and also some of them were insulated.

During the thermal measurements, it was noticed that connection between an aluminium cooler and the steel container was not ideal. Therefore, it was decided to take into account, a thin air layer between these elements. As a consequence, two following cases were calculated:

  • contact resistance is neglected,
  • contact resistance is considered.

That is why for the bottom wall of a steel container, an equivalent thermal conductivity, which accounts also for the contact thermal resistance, was calculated. This procedure allowed to obtain quite encouraging results.

Presented results generally confirm satisfactory agreement of numerical calculations with experiments. The largest inaccuracies occurred if the heat dissipation process was controlled only by the free convection. This is not surprising, since relevant formulas to calculate heat transfer coefficients for free convection are not too accurate. However, when forced convection took place and heat transfer coefficients are estimated much better, comparison of calculations with measurements is fairly good. Moreover, analysis in unsteady state adequately represented the heating process and also gave information about its duration until a steady state was reached. 


Figure 1. Geometry of transformer partially immersed in resin.


Figure 2. Temperature field in two cross-sectional planes of transformer tank