There is unavoidable error in the monitoring process of lateral displacement at foundation pit underground diaphragm wall. The error of the monitoring data was neglected in the past when estimating the underground diaphragm wall bending moment by lateral displacement. In order to solve the problem caused by the error, wavelet denoising method is introduced to denoise the lateral displacement monitoring data and the underground diaphragm wall bending moment can be estimated based on the data after denoising. The lateral displacement monitoring data of underground diaphragm wall typical measuring points in Shangjie Station foundation pit of Fuzhou Metro Line 2 are studied. Db3 wavelet basis function, hard threshold function and heursure threshold are selected to perform 5layer wavelet denoising. The sextic polynomial is used to fit the denoising monitoring data and radius of curvature and deflection of the fitting curves are determined. The bending moment of underground diaphragm wall is estimated based on the theory of flexural members of material mechanics. The results show that: (1)Wavelet denoising can effectively remove the error of lateral displacement monitoring data which is mainly distributed in the range of -0.6~0.6 mm. (2)The polynomial fitting accuracy can be effectively improved by using the noise reduction monitoring data. (3)The maximum estimated bending moment of the underground diaphragm wall is 635.1 kN·m, and is about 40% of the design value, so the underground diaphragm wall is in good condition during construction.