In data processing of anomaly separation, the structure method of trend surface adopts the fitting of a polynomial. It has some shortcomings, for example, the degree of fitting cannot be adjusted and can even show distortion. This paper proposed the construction method of trend surface based on the wavelet transformation. This method can extract the information effectively from the signal. The approximate coefficient obtained after the wavelet transformation characterizes signal tendency, and approximate coefficients of different criteria represent tendencies of different criteria. The coefficient and the detail coefficient of the static wavelet do not experience low sampling, so they have the same length as the original signal. The method can construct trend surfaces of different fitting degrees according to the requirement and is of distortion degree.