Adaptively Determination of Model Order of SVD-based Harmonics and Interharmonics Estimation

The singular value decomposition (SVD) is one of the most popular methods in harmonics and interharmonics estimation. However, its accuracy strongly depends on the correctness of the selected model order. To this purpose, this work aims at contributing to the correct estimation of the model order. This is achieved by exploiting the energy of the singular values (SVs). Firstly, the relationship between one frequency component and its corresponding SVs is theoretically investigated. Secondly, a new indicator is proposed for determining the model order, which denotes the energy of the k-th pair of consecutive SVs. Thirdly, an adaptive threshold is defined for separating signal components from noise. This way, the number of components can be obtained for unknown noise levels. Finally, the effectiveness and robustness of the proposed method has been validated by simulations. They have been run implementing typical signals designed according to the harmonics and interharmonics measurements standard, the IEC standard 61000-4-7.