基于ENN广义预测控制算法的滑坡位移多步预测  

Generalized Predictive Control Algorithm for Landslide Displacement Multi-step Prediction

作　　者：刘勇[1] 秦志萌 余宏明[2] 胡宝丹 Liu Yonga Qin Zhimenga Yu Hongmingb Hu Baodana(a. Faculty of Mechanical ＆ Electronic Informatio b. Faculty of Engineering, China University of Geosciences（Wuhan）, Wuhan 430074, Chin)

机构地区：[1]中国地质大学(武汉)机械与电子信息学院,武汉430074 [2]中国地质大学(武汉)工程学院,武汉430074

出　　处：《地质科技情报》2017年第4期255-259,共5页Geological Science and Technology Information

基　　金：国家自然科学基金项目(41302278)

摘　　要：针对滑坡位移预测时变性和复杂性的特点,将具有动态反馈特性的Elman神经网络(ENN)引入到分析研究中,充分发挥ENN对非线性情况的适应性。同时将之与广义预测控制算法相结合,以ENN作为广义预测控制算法的预测模块,结合滚动优化和反馈校正,提出了基于ENN广义预测控制算法的滑坡位移多步预测模型。将该方法应用于重庆市巫山县龙井乡白泉村大水田滑坡的位移预测,证实多步预测能够对位移变化趋势进行良好的拟合,同时高精度地实现了邻近时步的预测;通过与ENN多步预测进行对比,验证了模型的有效性。China is one of the countries vulnerable to the most frequent landslides.Landslide prediction is an effective way to reduce landslide disasters.However,many factors may result in landslides,with complex relationship,high uncertainty and nonlinear characteristics,making it difficult to obtain accurate mathematical physical model for displacement multi-step prediction.Aiming at the time-varying character and complexity,this paper introduces Elman neural network（ENN）,which has dynamic feedback,into the analysis because of its adaptability to nonlinear condition.By using ENN as the prediction module of generalized predictive control algorithm（GPC）,combined with the rolling optimization and feedback correction,the paper proposes ENN based generalized predictive control algorithm for the landslide displacement multi-step prediction.The application of this method to the displacement prediction for Dashuitian slope proves a good fitting to the displacement change trend showing up in multi-step prediction.At the same time,high precision is obtained in the prediction for adjacent time step.Compared with ENN in multi-step prediction,the proposed method produces lower error and higher precision,verifying its validity and effectiveness.

关 键 词：滑坡位移预测 多步预测 ENN 广义预测控制算法 

分 类 号：P642.22[天文地球—工程地质学]