Deep Learning Based Power Quality Disturbance Recognition using Residual Network in Smart Grid

With the integration of advanced digital technologies and communication systems into traditional power grid infrastructure, the power grid is having more and more issues with power quality disturbance (PQD). Automatic classification of PQD can be considered a great remedy for the power grid. The purpose of this study is to categorize seventeen different electrical power quality disturbances using a technique based on deep learning. ResNet50 is employed to classify PQDs automatically from a simulated dataset of 17 classes. Standard metrics such as Precision, Recall, and F1-score measure the performance of the deep neural network architecture. Our proposed approach produces a very encouraging result with a test accuracy of 99.66%, Precision of 99.65%, Recall of 99.65%, and F1-score of 99.66%.