Echo State Networks for Determining Harmonic Contributions from Nonlinear Loads

Abstract

This paper investigates the application of a new kind of recurrent neural network called echo state networks (ESNs) for the problem of measuring the actual amount of harmonic current injected into a power network by a nonlinear load. The interaction between loads connected to a point of common coupling (PCC) is a highly dynamic process. The determination of true harmonic current injection by individual loads is further complicated by the fact that the supply voltage waveform at the PCC is distorted by other loads at the PCC or further upstream and is therefore rarely a pure sinusoid. Harmonics in a power system are classified as either load harmonics or as supply harmonics. The principles of ESN are based on the use of a recurrent neural network (RNN) as a dynamic reservoir. In order to compute the desired output dynamics, only the weights of connections from the reservoir to the output units are calculated. This is simply a linear regression problem. Experimental results presented in this paper confirm that attempting to predict the total harmonic distortion (THD) of a load by simply measuring the load's current may not be accurate. The main advantage of this new method is that only waveforms of voltages and currents at the PCC have to be measured. This method is applicable for both single and three phase loads.