AbstractNonlinear loads used in industrial systems cause harmonic pollution in power system voltages and currents. Therefore, it is necessary to estimate properly the propagation of voltage and current harmonic distortions at power system buses in order to eliminate the harmonics. On the other hand, power system harmonic analysis methods can be used to avoid direct measurement of the harmonics. In this paper, a neural network (NN) based algorithm for harmonic analysis of industrial systems (i.e., systems whose topology rarely changes during the year) is proposed. In the meantime, an asynchronous simulation of the system in time and frequency domains is used to avoid the complications of some hybrid methods. Due to the use of time domain simulation to obtain the harmonic currents injected by nonlinear devices, in addition to giving acceptable precision, the method is completely independent of the measurement system. Application of NN for modeling nonlinear loads along with the use of an appropriate fast harmonic analysis procedure in frequency domain has resulted in acceptable execution time in the presented algorithm. The results obtained by modeling and simulation of the Sn‐coating unit of the Mobarakeh integrated steel making plant demonstrate the high accuracy and acceptable execution time of the proposed method. Copyright © 2010 Institute of Electrical Engineers of Japan. Published by John Wiley & Sons, Inc.