According to the characteristics of the commonly used wind measuring ultrasonic sensor NU150E25TR in the market, the main reasons for the error are analyzed. The characteristics of several main nonlinear compensation algorithms such as linear regression method, least square method and fold line method are compared, and the measured data of single chip microcomputer are used as samples, and the conditions of order and mode are fitted according to different curves. The error of these algorithms is compared and analyzed. The test results show that the piecewise fold fitting method is both simple and effective to reduce the compensation error. Under the same conditions, the accuracy is higher than the linear regression method and the least square method.