Abstract: The problem of prediction of soft soil settlement has always been a difficult problem in engineering, but the settlement prediction method based on the measured data is intuitive and effective. The traditional settlement prediction method has the limitation of low reliability and low accuracy, so a new settlement prediction model is proposed. The characteristics of the new settlement prediction model are analyzed, and the model parameters are solved by using the least square method. Based on the site settlement monitoring data of soft foundation treatment in Zhoushan Green Petrochemical Base, the surface subsidence law and settlement rate of monitoring points are analyzed, and the new settlement prediction model is verified theoretically. Finally, the following three conclusions are drawn. Using this model to predict the settlement of soft soil for a long time is more accurate than using some traditional prediction methods. The prediction of the middle and late periods of this method is relatively accurate, and the predicted results of the middle and late periods of actual engineering are more meaningful. The relative error of this method is smaller than that of the three-point method and the hyperbolic method. It is proposed that the whole settlement period of soft soil treatment should be graded and superimposed, and the corresponding parameters of each model are different, which is more in line with the law of actual soft soil settlement. The larger the thickness of soft soil, the greater the settlement.