A CNN-Based Deep Learning Method for Total Harmonic Distortion Estimation in Power Quality Signals

In recent years, the existing studies on deep learning methods applied in the field of power quality ggenerally deal with the classifying power quality problems. In the proposed study, the main harmonic analysis of a signal is aimed at performing the total harmonic distortion (THD) amount of the signal with a CNN deep learning based method. The neurons in the structure of CNN, which is specially designed for image processing, are structured in 3 dimensions, just like an image, as width, height and depth. In the preliminary simulation results of the study, THB analysis was performed to show the accuracy of the proposed method. The training of the deep neural network was done using the first 5,000 images out of 10,000 artificially created images and the THB values of these images. The training success was tested with the remaining 5 thousand images of the dataset. The test and estimated THB values were compared. When the estimated THB values were compared with the test values, it was seen that quite successful results were obtained.