Practical fault simulation on an earthing transformer using SFRA: A unique analysis approach towards simplifying SFRA results to assist with deformation diagnosis in Earthing Transformers

Earthing Transformers are an integral part of power and distribution systems around the world, although, little consideration is given to their ongoing monitoring and maintenance. The failure of an earthing transformer can cause a multitude of issues including compromised stability and safety of the electrical network. The necessity to maintain both safety and stability of electrical networks highlights valuable real world applications for an SFRA earthing transformer testing toolkit. As a starting point, the project adopted a review of existing research along with an analysis of earthing transformer design principles. Research found that because of an inherent design strategy, many ZN wound earthing transformers have a unique failure type in common; axial displacement of the inner and outer windings. The second project stage involved physical simulation of an earthing transformer’s axial windings displacement using SFRA as a diagnosis tool. Simulation results provided evidence that (for the given test subject) defect detection is possible using SFRA benchmarked comparisons. Analysis of benchmarked comparisons found deviation only at select resonances with general spectral shape retention for all other points along the SFRA trace. Spectral consistency of benchmarked comparisons allowed the implementation of a speech processing technique known as Mel-Frequency Cepstral Coefficients (MFCC). An adaption of the MFCC process introduced a way of encoding and distilling the SFRA trace data while exaggerating critical points of deviation. The third major project stage involved the development of code using Mathworks MATLAB as a platform to the fulfil data management and computational requirements of the adapted MFCC. Select variables were isolated throughout the code to ensure that the process was tune-able on multiple levels for future optimisation. By selecting and mapping the appropriate resultant cepstral coefficients against each other, it was found that a meaningful representation of the SFRA trace can be graphically presented as a single point on a two-dimensional plot. Simulated transformer defect scenarios had notable deviation on both the x and y axis when processed and plotted together. Analysis, processing and comparison of 28 different earthing transformer SFRA traces found possible real world applications for a single point spectrum classifier. The spectrum classifier was proposed as a substitution for pre-existing subjective analysis techniques, potentially building on the communal engineering toolbox for SFRA analysis