Abstract

Predictive maintenance of rectifier transformers in the aluminum smelting industry has become a major area of interest in planning for a replacement or refurbishment of these assets before a failure event occurs. The end of life of a transformer is linked to the rate of degradation of the winding paper insulation which is mainly due to heating processes. Rectifier transformers are subject to high thermal stress due to harmonic currents flowing through them. The need of monitoring and regulation of the hotspot temperature on the rectifier transformer winding is of great importance to keep the temperatures within safe limits as to preserve its life span. In this thesis, existing thermal models; the IEC model, the improved IEEE model, the G. Swift model and the D. Susa model used for hotspot temperature estimation in regulating power transformers has been adapted to account for increased heating due to harmonic currents flowing in the rectifier transformers. Extrapolation techniques, nonlinear least square optimization and genetic algorithm optimization are used for obtaining the rectifier transformer thermal model parameters using online measurements. The thermal model parameters are obtained in two different cooling fan operation conditions; OFAF mode 1 (one fan operation) and OFAF mode 2 (three fans operation) as the transformers under case study are utilized in these cooling modes. A predictive maintenance technique is implemented using typical loading profiles of the transformers and forecasted ambient temperatures to estimate and regulate future hotspot temperatures within safe temperature limits as derived using an industry accepted end of life equation.