Research On Power Measurement Algorithms Of Harmonics And Interharmonics In Power Grids

With more and more applications of nonline and impact load in the power sys-tem, that cases severe problems of harmonics and interharmonics. It’s a challenge for precise power measurements of harmonics and interharmonics with fluctuations on the frequency and amplitude in the power system. The paper mainly investigates accu-rate and rapid measurement algorithms on the parameters and power for power signals with Dc bias, harmonics and interharmonics, under fluctuations on the frequency and amplitude or stabilization in the power system.For a steady state power grid, a algorithm based on modern spectral estimation is proposed. The paper deep analyzes the algorithms of Multiple Signal Classifica-tion (MUSIC) and Estimating Signal Parameters Viarotational Invariance Techniques( ESPRIT). A algorithm is proposed which integrates ESPRIT algorithm based on cross spectrum (IMESPRIT). Cross-spectrum technique is used to calculate signal and noise subspaces with the independence of white noise at different time, calculate the ampli-tude and phase with the characteristic of zero. Finally, the power of Dc bias, harmonics and interharmonics are obtained. Matlab results show that the proposed algorithm has advantages of high frequency resolution, and can be applied on measurement of Dc bias, harmonics and interharmonics parameters and power in a stable power system.For a situation of little fluctuations on frequency and amplitude in the power sys-tem, a algorithm based on the adaptive neural network is proposed. The algorithm in-tegrates the IMESPRIT, adaptive neural network algorithm with frequency modulation link and normalized LMS, which is divided into two parts:filtering and power measure-ment. Improved ESPRIT algorithm is used to analyze the voltage and current in the power system, gets the frequency of harmonics, M adaptive neural network model are structured for extracting the harmonics, the frequency initial value of which is set as the estimated one by IMESPRIT algorithm. Normalized LMS is used to adjust the weight vectors of improved adaptive neural network and the each compent of power signals is extacted, and measure the harmonics and Dc bias power, respectively. Normalized LMS is used to adjust the weight vectors for better convergence and high accuracy. Matlab results show that the proposed algorithm has the advatages of quick convergency and high accuracy.For a situation of large fluctuations on frequency and amplitude in the power sys-tem, a algorithm based on the adaptive notch filter (ANF) is proposed, which can process Dc bias. The algorithm integrates IMESPRIT algorithm and adaptive notch filter, which is divided into two parts:filtering and power measurement. Considering the tradition ANF model can’t process sinusoidal signals with Dc bias, or remarkable error. The improved ANF model with an capacity of processing sinusoidal signals with Dc bias is proposed, which can extract or eliminate Dc bias, harmonic and interharmonic com-ponent of voltage and current in the power system. Normalization algorithm is used to adjust the improved ANF parameters for better convergence and high accuracy. Matlab results show that the proposed algorithm has the advatages of quick convergency and high accuracy.In the last, we have a review on this research and some problems are discussed in the future.