Harmonic detection method using APFFT and neural network

A new algorithm, which utilizes all-phase FFT (APFFT) and artificial neural network (ANN), is presented to detect integer harmonics and non-integer harmonics in power system. In order to improve the accuracy of harmonic parameter estimation, the method incorporates the property of phase invariant of APFFT with high speed sought optimization solution function of ANN. First, the sampled data is processed by windowed APFFT algorithm, and then some of harmonic parameters can be obtained, including the number of harmonics, accurate phases of harmonics, inaccurate magnitudes and frequencies of harmonics. Second, the number of neural nodes, the initial weights and the iterative initial parameters of base function of neural network are set according to the results analyzed with APFFT. Finally, accurate harmonic parameters can be obtained by training ANN. Simulation results demonstrate that the method can detect harmonic parameters with high precision.