The use of deep learning and 2-D wavelet scalograms for power quality disturbances classification

This work investigates the use of advanced signal processing and deep Learning for pattern recognition and classification of signals with power quality disturbances. For this purpose, the continuous wavelet transform is used to generate 2-D images with the time-frequency representation from signals with voltage disturbances. The work aims to use convolutional neural networks to classify this data according to the images' distortion. In this implementation of artificial intelligence, specific stages of design, training, validation, and testing were carried out for a model elaborated from the scratch and a transfer learning technique with the pre-trained networks SqueezeNet, GoogleNet, and ResNet-50. The work was developed in the MATLAB/Simulink software, all signal processing stages, CNN design, simulation, and the investigated data generation. All steps have their objectives fulfilled, culminating in the excellent execution and development of the research. The results sought high precision for the model from scratch and ResNet-50 in classify the test set. The other two models obtained not-so-high accuracy, and the results are consistent when compared with different methodologies. The main contributions of the paper are: (i) developing a methodology to use DL and transfer learning on the classification of voltage disturbances; (ii) using a 2-D representation that incorporates time and frequency information that characterizes several PQ issues; (iii) conducting a study case that shows the suitability of CNN as a tool for voltage disturbance classification, with specific application for 2-D images. Considerations about the results were pointed out.