“…ANNs are typically used in classification problems where each neuron of the output layer represents a category and is activated according to the respective inputs. In the problem of multiple PQDs and voltage sag classification, several types of shallow ANNs have been used taking advantage of their flexibility and adaptability to problems where labels are well identified, for instance, the learning vector quantization [37][38][39], probabilistic neural network [44,52,56,65,71,84,95,102,123,147,168], self-organizing learning array [46], radial basis function [47,83], multilayer perceptron [48,53,57,65,89,93,102,168,177,178], adaptive linear network [82], feedforward [82,85,102], backpropagation [90], random vector functional link [113], and modular ANN [143]. Most recently, a learning algorithm known as ELM has been gaining popularity because of its remarkable efficiency.…”