An application of neural network in power system harmonic detection

Widely distributed power electronic loads are an increasingly important source of harmonics in power systems. The objective of the paper is to detect the components (magnitudes and phases) of harmonics in power distribution systems by means of the multilayer perceptron neural network. The detection of 3rd, 5th and 7th harmonic components is verified by means of computer simulation. This method of harmonic detection has the advantage that it can determine harmonic components in real time. It is seen that neural networks can determine the harmonic components with very low error.