【文章內(nèi)容簡(jiǎn)介】 each junction, as shown in Fig. 2. Basically model equations of the hydraulic network are formulated by using the conservation of mass for flow into and out of each junction and by specifying that the algebraic sum of the pressure drops around each closed loop is zero [35,36]. An experimental study has been conducted for the validation of the present hydraulic network model [37]. The improved model is in good agreement with the experimental and numerical results.The temperature distribution in the winding disks is simulated with a heat conduction model using the finite volume method, and the boundary condition for it is the convection heat transfer at the disk surfaces, related to bulk temperature of surrounding oil and local heat transfer coefficient. Due to the axisymmetric geometry involved, the analysis is reduced to 2D cylindrical section. The temperature and location dependent heat generation rate, and the thermal contact resistance are taken into account, and the model can be applied to various disk configurations, made of Single, Twins and CTC conductors as shown in Fig. 1.With coupling between the heat conduction model and the hydraulic network model, a prehensive thermal model is developed to simultaneously investigate a 2D temperature field in the winding disks and hydraulic and thermal fields in the cooling oil.In the horizontal duct , the temperature rise in the ith fluid control volume is dependent on the heat absorbed within, whose value should be equal to the heat transfer from the neighboring solid control volume in disk I and in disk , the energy balance equation therefore bees（6）In the vertical duct, the temperature rise in the th fluid control volume depends on the heat transfer from the neighboring solid control volume in disk, only, since the inner wall on the other side is assumed as a thermal insulator due to negligible heat transfer there. Therefore, energy balance yields（7）The fluid properties in the control volume are determined at the bulk mean temperature,（8）At duct junctions, mixing among the fluids from different ducts occurs, and the fluid properties are determined by a bulk mean temperature. The specific heat at constant pressure is reasonably constant for the cooling oil, thus the thermal equation for the branch junction bees, （9）For the confluence junction ,以及 （10）Clearly, the junction temperatures are already discretized. At the solid/fluid (disk/oil) interface, the heat flux at the disk surfaces is treated as thermal boundary condition input into the hydraulic model, as given in (6) and (7), and interactively the convection boundary condition executed to the heat conduction model can be calculated from the outputs of the hydraulic model. Therefore, an iterative method is used to tackle the coupling between the heat conduction and hydraulic models. The correlation from Shah and London [30] is chosen in the thermal model to determine heat transfer coefficient. It is validated against experimental data, and generally the predictive accuracy is within [38,39].3. Experimental facility and measurementsAn experimental study is conducted to investigate the disk temperature and the oil flow and temperature distribution in the windings at steady state, as affected by heat generation rate in the conductors, total flow rate across the winding ducts and various winding duct geometries. The experimental results are employed to reduce the uncertainty associated with the heat transfer coefficient, and to further improve the thermal simulation model in accuracy and reliability. The experimental facility and method are described below.. Oil circulation systemThe oil circulation system of experimental facility is shown in Fig. 4. Mineral oil VOLTESSO 35 is adopted as experimental coolant. The tested winding model is setup in the test tank, and its position is fixed with four locating pins at the tank base. The two exits of the test tank, the lower and upper exits, are designed to ensure that the tested model is always pletely submerged in the coolant during the experiment. The top tank is employed to supply a steady oil flow by elevation head into the tested model from the bottom of the test tank. The total flow rate is controlled with a globe valve and measured using a flow meter, OMEGA Model No. FL6102A. It has a flow range of with an accuracy of 177。2% full scale (177。). Other flow operations, such as running backward, filtering oil, emptying oil tank and adding more oil to the system, are also necessary during the experiment period, and they can be executed with an on–off arrangement of the various valves. No specific radiator is installed in the oil circulation system. During the thermal performance testing, the oil tank, top tank and pipes are working as radiator to dissipate the heat generated from the winding model to the ambient air.. Tested winding modelsIn disktype transformer winding, spacers and sticks are placed circumferentially to evenly divide the horizontal and vertical ducts and the winding disks in the tangential direction. As a result, taking the domain between two adjacent spacers and sticks as one part, the thermal and hydraulic phenomena in each part are similar and equivalent. Based on this consideration, one part of transformer winding is facsimiled in the present experimental study, and the model scale is designed to be 1:1.In the developed simulation model, a variety of disk configurations, such as Single, Twins, CTC and Semisolid, etc., has been taken into account. Since the experimental validation in the present study is mainly focused on the oil flow and temperature distribution and the determination of local heat transfer coefficient as discussed above, some simplifications and modifications are made for easy model buildup and with no loss of generality. In disk construction, paring to the actual singleconductor disk configuration shown in Fig. 1, a model disk is illustrated in