“…In one of the first ANN applications, [121][122][123] proposes neural networks in combination with the third harmonic magnitude and angle of the positive, negative, and zero-sequence signals. In [124], a back propagation neural network is trained with four 1/4 windows FFT outputs; once trained, the method can discriminate between HIF, load switching, capacitor switching, and arc furnace. In [125], FIR filters are used to obtain the signal for training the neural network to detect HIF's.…”