Inter-Harmonics Identification in Frequency Spectra by the Application of Artificial Neural Networks

Harmonic analysis using the Discrete Fourier Transform is a relevant topic in electric power quality. However, such analysis may be impaired by the presence of inter-harmonics, which is ever-growing due to the insertion of power electronic converters in the grid, and causes spectral leakage; despite the efforts of the scientific community in proposing solutions for the mitigation of this problem, there is still no consensus. This work proposes a solution based on Multilayer Perceptron Artificial Neural Networks for inter-harmonic diagnosis. The solution can properly identify and estimate the frequency of inter-harmonics, and the results for synthetic signals indicate the effectiveness of the proposed technique, with a hundred percent identification rate and a less than 1.3 percent relative error.