Voltage Disturbance Classification for Transmission Grid Using Wavelets and Deep Learning
Power Quality (PQ) is directly associated with the quality of the voltage signal waveform.The classification of PQ issues becomes an important challenge, which uses pattern recognition tools to make a cause and effect relationship of the quality problems. This paper aims to develop a digital signal processing system for voltage signal pattern recognition, to classify voltages disturbances in a transmission grid. For this, Continuous Wavelets Transform (CWT) with filter banks is used to perform the signal feature extraction, through scaloragrams.Thus, obtaining the time-frequency analysis in the 2-D image, a technique based on deep learning is used for classification, it is the Convolutional Neural Network (CNN).The detailed simulation study of power quality disturbances has been carried out in MATLAB/Simulink environment. After generating the signals using the grid model, it was possible to convert the obtained signals into scalograms. CNN's training showed high performance. The proposed system was able to classify the events with great success, in addition to highlighting the efficiency in using advanced signal processing along with deep learning techniques. Using image classification also shows new options for solutions and visualization for power quality events. Finally, a dataset with the codes and programs are available for research community.
