Automatic partial discharge and noise signals separation using arithmetic coding in time domain and magnitude distributions in frequency domain

A system for separating partial discharge and noise signals comprising:

• a data collecting module configured to recording and translating electromagnetic signals measured from a sensor to digital signals;
• and
• a data processing module comprising one or more computing processors executing instructions, wherein the instructions are configured to:
• receive digital signals from the data collecting module;
• generate a feature from amplitudes in a time domain for each of the digital signals, wherein generating the feature from the time domain for each of the digital signals comprises:
• determining histogram bin width and the number of bins for all measured amplitudes in the digital signals, wherein the histogram bin width and the number of bins are determined via Freedman-Diaconis (FD) rule;
• associating all measured amplitudes with their respective bin values;
• building a global probability table;
• and
• encoding information in the global probability table to generate the feature;
• generate a plurality of features from a frequency domain for each of the digital signals, wherein the plurality of features from the frequency domain are dependent on a number of measured amplitudes per digital signal;
• apply clustering algorithm on the generated features for all the digital signals to identify a plurality of distinct clusters;
• display each distinct cluster on a Phase-Resolved Partial Discharge (PRPD) chart, wherein at least one axis of the PRPD chart represents a voltage amplitude of each digital signal in the plurality of clusters;
• and
• separate, using the distinct clusters, the partial discharge and noise signals from the digital signals.