Research On Estimation Algorithms Of Harmonic And Interharmonic In Power System

Harmonic and interharmonic are two critical indices of power quality, an accurate and effective estimation algorithm will be very important for the safe and stable operation of power system. The detailed work is as follows:Firstly, the Fourier transform algorithm is applied for the estimation of harmonic and interharmonic, and the spectrum leakage and picket fence effect caused by asynchronous sampling have been researched carefully, while an improved result can be got by using different window functions and two spectral line interpolation. Simulation results verify an interharmonic very close to the fundamental frequency is hard to detect accurately, besides, it can lead to a poor real-time by increasing the sampling points to raise the precision.For the low precision in asynchronous sampling and poor real-time of Fourier transform and its improved algorithm, an improved radial basis function neural network has been adopted, which can obtain the learning centers and standard deviation by using K-means clustering algorithm, the weights can be got by a least squares matrix inversion, when compared with the classical multiple layer feedforward neural network, the improved algorithm can solve the shortcomings of difficulty for electing the suitable learning rate and easy to fall into local minimal. Simulation results show the improved algorithm yields fast convergence speed and high accurate estimation.In addition, for the bad estimation result of special interharmonic and poor real-time of Fourier transform algorithm and its improved algorithm, the Prony algorithm is applied for accurately analyzing frequency and the number of harmonic and interharmonic firstly, then the adaptive linear neural network is used for getting the precise amplitude and phase. In this way, the hybrid algorithm can detect the harmonic and the interharmonic which is very close to fundamental frequency accurately in stable signal, besides, it can also obtain high estimation precision and good real-time in time-varying signal. Simulation results show the hybrid algorithm is very effective and accurate.