COMPUTATION OF STRAY LOSSES IN GSU TRANSFORMER USING SURFACE IMPEDANCE FORMULATION S. Saravanan*, Dr. N. K. Deshmukh** CG Global R&D Centre, Crompton Greaves Ltd, Mumbai 400042, India E-mail: * saravanan.selvaraj@cgl.co.in., ** narayan.deshmukh-cons@cgl.co.in Abstract - A complex problem, the eddy current field due to windings and high current carrying busbars in a generator transformer is analyzed by using surface impedance formulation in a commercial finite element package. To check the validity of the method TEAM problem 21B is solved and experimentally verified. This paper also describes the influence of busbar grouping on stray losses. Indexing terms : Eddy losses,Transformers,High current leads,Surface Impedance. I.INTRODUCTION In a three phase transformer the leakage flux emanating from in between the windings and surrounding the leads, causes considerable losses in the magnetic structures such as tank plates, frames, tie rods etc. When the losses due to leads carrying high ac currents are high, several options like moving the leads farther from the magnetic structures, placing a plate or shield of high conductivity and low permeability on the magnetic structures next to the leads are possible for reducing them. Another option is to place the leads from different phases, from a balanced three-phase network, near each other so that their magnetic fields can partially cancel each other out [1].The effectiveness of aforesaid possibilities can be explored with 3D nonlinear eddy current Finite Element Analysis which is not a trivial task for an analyst especially with the 3-D mesh generation, because a very fine mesh is required within the skin depth of the material. A very fine mesh is, however, very costly in the terms of computation time and memory space. Problems associated with the 3-D mesh generation can be avoided by using surface impedance formulation. The use of surface impedances requires that the skin depth is relatively small compared to the size of the conductor.The idea is to provide a boundary condition with which the ratio of electric to magnetic field, i.e., the surface impedance, at the surface of the work piece is specified to be equal to a complex number. It is assumed that the actual distribution of the field inside the material, replaced by this boundary, is not of interest[2]. In this paper the accuracy of the Finite Element approach has been checked by solving the TEAM problem 21 B formulated by IEEE [1]. The model is analyzed with linear, nonlinear Timeharmonic, nonlinear Transient, and linear surface impedance methods. Having validated the modeling and computation approach , it is applied to compute the stray losses in magnetic structures of a GSU transformer.Also the effectiveness of leads placement in stray loss control also discussed. II. TEAM PROBLEM 21B A. Describtion: TEAM 21 is an engineering oriented loss model, which is approved by International TEAM (Testing Electromagnetic Analysis Methods) board in 1993.A number of calculated results have been presented worldwide. TEAM 21 has two models: Model A and Model B .Model A consists of two exciting coils of the same specifications and two magnetic steel plates. In the center of one steel plate, there is a rectangular hole. Model B has the same exciting coils as Model A and only one steel plate without a hole. The direction of exciting current of one coil is different from that of another coil. Each coil having 300 turns and is carrying a current of 10 A (RMS,50 Hz). Fig.1 shows the structure with overall dimensions of the TEAM Problem 21 Model B. Figure 1: Analyzed Model B. Finite Element Analysis: The material properties and model details used in the analysis are similar to that mentioned in [3], [4]. By taking advatage of double symmetry, only one-fourth model is analyzed. The accuracy of linear analysis depends upon the selection of constant  r value for the magnetic steel plate hence linear analyses are carried out with a range of constant  r values and the results obtained is shown in Fig.2. Since it is assumed that the flux and the induced eddy currents are sinusoidally-varying in both linear and nonlinear