Power Quality Disturbance

This paper proposes two new monitoring methods capable of detecting electrical disturbances in low voltage grids. Both approaches rely on machine learning techniques that classify voltage signals in the frequency domain. The first technique here proposed uses the Fourier Transform (FT) of the voltage waveform and classifies the corresponding complex coefficients through a Multilayer neural network with Multi-Valued Neurons (MLMVN). In this case, the structure of the classifier has three layers and a small number of neurons in the hidden layer. Therefore, the computational effort is very low, the learning time is short, and no coding operations are necessary because the neural network can process complex-valued inputs. The second technique involves the use of the Short Time Fourier Transform (STFT) and a Convolutional Neural Network (CNN) with 2-D convolutions in each layer for feature extraction and the reduction of dimensionality. The five disturbances considered in this paper are: voltage sag, voltage swell, harmonic distortion, voltage notch and interruption. The performances of the two classifiers are compared during the training phase using simulated data and subsequently through experimental measurements, obtained from an artificial generator of disturbances and a variable load.Both techniques represent an innovative approach to this type of problem and guarantee excellent classification results.