Containing a lot of random noise in the partial discharge signal, it is difficult to separate partial discharge signal and random noise. A partial discharge signal extraction method based on dynamic measurement uncertainty is proposed in this paper. The deterministic component is separated by Gaussian, and the random component of the remaining residual after the separation is estimated using autoregressive (AR) model；The true value estimation and dynamic measurement uncertainty of noisy signal are obtained by the deterministic component and random component；According to the 3σ rule，the threshold is selected, and based on dynamic measurement uncertainty, the poor quality of the data is removed to improve the filtering effect. Because of the discontinuity, wavelet transform is used to improve the accuracy of the extraction. Finally, the effectiveness of the method is verified by MATLAB simulation and experimental noisy PD signal extraction.