Small Signal Internal Voltage Transfer Measurements and White-Box Transient Calculations for Non- Standard Test Conditions of a Shell-Form Power Transformer Bjørn Gustavsen 1 , Ariana Martins 2 , Carlos A. Sá 2 , Luis Braña 3 , Ricardo Castro Lopes 3 , Pedro Lima 3 , Andrea Soto 3 , and Mário Soares 4 1 SINTEF Energy Research, Postboks 4761 Sluppen, NO-7465 Trondheim, Norway 2 Universidade do Porto, Faculdade de Engenharia, Porto, Portugal 3 Efacec Energia - Máquinas e Equipamentos Eléctricos, S.A., Matosinhos, Portugal 4 REN - Rede Eléctrica Nacional, S.A., Lisboa, Portugal E-mail: bjorn.gustavsen@sintef.no Abstract—Research tests were performed on a shell-form autotransformer with a more rigorous impulse test methodology than the traditional recurrent surge oscillograph method. These tests included non-standard terminal connections with open terminals, differing from those tests defined by the international standards related with impulse testing of power transformers. The test voltage responses were obtained using voltage transfer frequency sweep measurements that were converted into time domain waveforms. The measurement results were compared against simulations by a white-box model, demonstrating satisfactory accuracy of the transient calculation tool. In addition, the sensitivity of the measurements to measuring probes length was experimentally evaluated. Reducing the lengths with 1.5 m only affected frequency components around 1 MHz. Index Terms—shell-form transformer, transient response measurements, simulations, white-box model. I. INTRODUCTION The dielectric withstand capability of the transformer windings against transient overvoltages is verified by the lightning impulse factory test [1]. In-service voltages impinging the transformer terminals are however very different from the standard impulse test voltage. The verification against internal stresses due to other voltage wave shapes and terminal conditions can only be calculated using the white-box transformer model available to the transformer designer. It has been found in CIGRE JWG A2/C4.39 [2] and A2/C4.52 (ongoing) that although the white-box models used by manufacturers are well suited for the standard lightning impulse response prediction, their accuracy still needs to be evaluated for the application of non-standard waveshapes and terminal connections. In order to assess the white-box accuracy limitations, we performed research tests with direct measurements of frequency domain voltage transfer functions, from external terminals to critical internal points in the regulating winding. The time domain voltage response to any time domain excitation could then be determined via rational approximation and convolutions and be compared against simulations by a white-box model. While a similar approach has already been applied to core-form transformers [3], we apply in this work the approach to a shell-form transformer. We also investigate the effect of measurement lead lengths on the measured results. II. SHELL-FORM TRANSFORMER A. Transformer main data The tested transformer is a seven leg, 800 MVA, shell-form, three-phase autotransformer. The nominal phase-to-phase voltages are 500 kV, 161 kV and 34.5 kV, for the high voltage, low voltage and tertiary voltage respectively. Each phase of the series winding was designed with 6 tap leads, for connection to a de-energized tap-changer and were available for testing. At the time of testing, the connections between phases were not done yet (neutral and delta), so all the leads were available as 3 single phase units. B. Transformer winding layout Each phase is built with 56 pancake coils (4 for tertiary winding, 24 for common winding, 28 for series winding), and 6 static shields as per Fig. 1. Minimum tap position was chosen for these tests, because it leads to a higher floating portion of tap winding, so higher voltages should be expected at tap leads P5 and P6.