With the use of high power devices, resulting in a large number of harmonics, threat the safety of equipment. This paper proposes the use of wavelet neural network (wave neural network WNN) algorithm to detect harmonics. Firstly, the initial value of neural network convergence set due to improper slow or even non convergence problems, put forward a method of optimal initial parameters correlation correction, improve the network performance. Secondly, using smoothing training algorithm with additional momentum item weights learning path, avoid network training into local minimum, to improve the precision of harmonic detection. Finally, through simulation and comparison with other detection methods, proved that the method presented in this paper has advantages of fast convergence speed, detection high precision.
