Abstract: When the classical fast Fourier transform method is used to estimate the power harmonic parameters, the fence effect and the spectrum leakage due to non integer period truncation and asynchronous sampling can cause estimation error. To improve the accuracy of harmonic parameter estimation, combined with the advantages of self-multiplication and self-convolution window, a self-multiplication-convolution window was designed based on Rife-Vincent window. With good sidelobe characteristics, the proposed window could effectively suppress spectrum leakage. The estimation formulas of harmonic amplitude, frequency and phase were obtained based on three spectral line interpolation and least square fitting. The simulation experiments were performed under the conditions of complex harmonics, interharmonics and fundamental frequency fluctuations with weak amplitudes. The results show that the proposed algorithm has higher detection accuracy than the classical windowed algorithm.