New algorithms to investigate and search for deviations in
signals

Abstract: This Master thesis describes a partly new procedure developed to find
deviations in measured voltage and current signals. The work is an
excellent example of combining theory knowledge from the university with a
problem identified in the Swedish industry. A combination which stimulate
technical development.
The three phase signals used in this thesis work were recorded at a
substation located in Stackbo, Sweden. The purpose with sorting out
interesting deviations is to find connections between deviations and
problems in the equipment, which can be used to detect such problems
already during the measurements in connection to the substation. To sort
out interesting deviations, the signals have been analyzed using Matlab and
two transform methods, the Fourier transform and the wavelet transform.
The new Matlab algorithms developed to analyze the signals make it possible
to efficiently search through a large number of files containing signals in
accordance to some efficient criteria. Setting the values for the chosen
criteria is done based on typical amplitude and the noise level of the
signals.
The criteria can be adapted to the deviations of interest. Different types
of deviations and their locations in the time scale are possible to find.
For instance, notches in a signal, amplitude changes, significant deviation
in the power frequency (50/60 Hz) and small regular deviations lasting for
a relatively long time, can be found.
Some information about how to classify the detected waveform files by the
algorithms, based on the type of deviations, is also included in this
thesis. This can make further development easier. Suggestions for
improvements of the algorithms are mentioned at the Appendix of this Master
thesis.