Abstract: The field data acquisition of transient electromagnetic method is prone to jump due to various random disturbances. Wavelet denoising realizes data denoising by multi-scale decomposition, threshold setting and reconstruction of signal. In order to improve the data quality of TEM after wavelet denoising, a new wavelet denoising method based on the optimal threshold reference is proposed to improve the threshold selection method. Combined with the application of transient electromagnetic method in water-rich engineering, the noise reduction results of data under different threshold conditions are compared and analyzed, and the practical application effect of the proposed algorithm is discussed. The results show that the proposed method can further reduce the value of the induced electromotive force curve in the range of 1-4 ms, and make the whole period attenuation characteristics of the data more obvious. The smoke ring inversion results of the denoised data were analyzed, and three locations with relatively large electrical differences were explained, and the inversion results of the denoised data based on the optimal threshold reference were verified to be more accurate.