Accurate no-load behavior prediction of large five-legged transformers using topological transient models

Purpose: The work proposes an approach for reproducing in detail the no-load losses and currents of large five-legged transformers when using their transient circuital models. This purpose is achieved by using topological transformer models based on a dynamic hysteresis model (DHM) and taking into account transformer capacitances. Design/methodology/approach: It is proposed to replicate initially the measured total losses by using DHM means, and then to find an equivalent per phase capacitance which provides the best coincidence of calculated and measured waveforms of all three line currents. A further improvement in the calculated current waveforms is achieved by introducing variable gaps between the core legs and yokes and determining their Weber-ampere characteristic. Findings: The authors show the importance of accounting for transformer capacitances for accurate modeling of line currents feeding large transformers. The possibility of representing transformer capacitances by equivalent per-phase elements is demonstrated. The variable core gaps are found to be effective elements of the model. Practical implications: The presented models can be used either as independent tools or as a reference for subsequent developments. The method is supported by a close agreement between the calculated and measured data for two large transformers. An accurate reproduction of current waveforms at elevated voltages indicates the applicability of the models for predicting inrush events and transformer behavior under geomagnetically induced currents (GIC). Originality/value: The originality of the method proposed lies in using only the voltages and supplying line currents. Although the validity of the models is illustrated at no-load operations, their topological nature ensures the applicability of the models for studying transients occurring in a transformer under abnormal conditions such as inrush currents or in the presence of GIC.