Fast gridless algorithm for harmonics and interharmonics estimation

Abstract

The increase of electronic power devices connected to the electric power system may cause the addition of harmonics and interharmonics components in the system signals, which may cause several adverse effects. The study and development of algorithms to correctly estimate the components that compound the power system signals considering the robustness to noise and frequency deviation are essential to guarantee the power quality of the system. In this work, a new algorithm called Anti-Leakage Matching Pursuit (ALMP) is proposed for estimating the harmonics and interharmonics components, avoiding the Spectral Leakage effect, in contrast to some techniques in the literature such as the Discrete Fourier Transform (DFT) and Harmonics and Interharmonics components Estimation based on Signal Sparse Decomposition (HIESSD). The algorithm is based on the Matching Pursuit technique and aims to decompose any signal into a linear combination of previously known waveforms. The main stage of the ALMP is the solution of a non-linear least squares problem by an optimization algorithm, where the waveform that best matches the signal is determined. The study of optimization techniques is carried out to determine which one will be used in the algorithm, and the most recent algorithms for harmonics and interharmonics estimation are presented. The proposed method is compared with DFT, Matrix Pencil Method (MPM), Fast Matching Pursuit (FMP), and HIESSD techniques to assess its proper estimation. In terms of amplitude, the proposed algorithm obtained results equivalent to the existing algorithms presenting a Mean Absolute Error (MAE) of 1.6 × 10−3 pu for harmonics estimation, while the DFT, MPM, FMP and HIESSD presented average MAEs of 1.8×10−3 pu, 4.3×10−3 pu, 8.3×10−3 pu and 8.3×10−3 pu, respectively, considering the same conditions. It also proved to be robust to frequency deviation, in contrast to HIESSD and DFT and presented a reduced execution time compared to the other presented algorithms, up to 24 times less than the time presented by the FMP, up to 73 times less than the time presented by the MPM and up to 294 times less than the time presented by the HIESSD.