Utilization of Echo State Networks for Differentiating Source and Nonlinear Load Harmonics in the Utility Network

Abstract

Echo state networks (ESNs) are a special form of recurrent neural network (RNN). Complexities with existing algorithms have thus far limited supervised training techniques for RNNs from widespread use. When it comes to practical applications, multilayer perceptron neural networks (MLPNs) still dominate. This paper investigates the application of ESNs for the prediction of true current harmonics of a nonlinear load in the presence of distorted supply voltage. The determination of true harmonic current injection by individual loads is complicated by the fact that the supply voltage waveform at the point of common coupling (PCC) is distorted by other loads at the PCC or further upstream. Experimental results are presented with the proposed ESN algorithm applied to a laboratory test circuit. The difference between the measured current harmonics and the predicted ESN results are quantified in the form of a new parameter. The uniqueness of the proposed harmonic prediction method is that the results are obtained without disrupting the operation of any load and only require the acquisition of voltage and current waveforms. This method is applicable for both single- and three-phase loads.