Power Harmonic Analysis Based on Real-Valued Spectral MUSIC Algorithm

Tao Cai,Shanxu Duan,Fangrui Liu

DOI: https://doi.org/10.3321/j.issn:1000-6753.2009.12.024

2009-01-01

Abstract:In recent years, modern spectral estimation methods, especially the MUSIC (Multiple Signal Classification) algorithms, are gradually used for high-resolution power harmonic analysis. However, most of them are proposed to detect frequencies of complex signals so that any real-valued signal should be transformed into complex form. The preprocessing may lead to higher computation burden. In addition to this, the picket-fence effect also exists in searching spectrum peaks, which causes low analyzing precision. To overcome these drawbacks, a real-valued MUSIC algorithm for power harmonic analysis is proposed in the paper based on subspace decomposition theory. And the computing method of pseduospectra is also derived. Furthermore, to improve the measuring accuracy of harmonics, Newton-Raphson algorithm is adopted to optimize the harmonic frequencies significantly. Simulation results show that: the spectral peaks of true harmonic components are more distinct from false peaks caused by noise in the real-valued MUSIC,and the computational complexity is notably lower than that of the classic MUSIC,as well as the detecting precision is close to that of root-MUSIC algorithm which is quite time-consuming.