本文提出了以非线性回归拟合发展趋势、正弦函数逼近周期变动和马尔可夫链刻划随机扰动的集成预测模型，并应用于山东省农业自然灾害成灾面积的变动规律模拟，得到了较好的预测效果。理论和实践表明，集成预测模型优于传统的单模型预测，为预测具有复杂机制的自然灾害演变提供了一种新方法。 In this paper, an integrated prediction model of nonlinear regression fitting trend, approximation of sinusoidal function and Markov chain scratching random perturbation is put forward and applied to the simulation of the variation rule of agricultural natural disasters in Shandong Province. Good prediction effect. Theory and practice show that the integrated forecasting model is superior to the traditional single model forecasting and provides a new method for predicting the evolution of natural disasters with complex mechanisms.