新息自适应混合卡尔曼滤波算法构建地表沉降预测模型  

A land subsidence prediction model by adaptive hybrid Kalman filtering algorithm

作　　者：曾令权[1] 熊鑫 陈竹安[2,3,4] Zeng Lingquan;Xiong Xin;Chen Zhu’an(Department of Architectural Engineering,Guangzhou Panyu Polytechnic,Guangzhou 511483,China;Faculty of Geomatics,East China University of Technology,Nanchang 330013,China;Key Laboratory of Watershed Ecology and Geographical Environment Monitoring,NASG,Nanchang 330013,China;Jiangxi Province Key Laboratory of Digital Land,Nanchang 330013,China)

机构地区：[1]广州番禺职业技术学院建筑工程学院,广州511483 [2]东华理工大学测绘工程学院,南昌330013 [3]流域生态与地理环境监测国家测绘地理信息局重点实验室,南昌330013 [4]江西省数字国土重点实验室,南昌330013

出　　处：《工程勘察》2020年第4期55-61,共7页Geotechnical Investigation & Surveying

基　　金：广州番禺职业技术学院“十三五”(第二批)科研项目(编号:2018KJ010)。

摘　　要：为解决矿区地表沉降变形预测的问题,提高预测模型的精度,提出了基于自回归综合移动平均模型(Autoregressive Integrated Moving Average model,ARIMA)的新息自适应卡尔曼滤波(Innovation Adaptive Kalman Filter,IAKF)与组合神经网络相结合的混合预测模型。首先,针对沉降变形监测序列的非平稳性与复杂性等特点,ARIMA模型能够将原始数列平稳化,以此构建地表下沉的预测模型,并作为新息自适应卡尔曼滤波的状态方程。然后,将集合经验模态分解(Ensemble Empirical Mode Decomposition,EEMD)、粒子群优化算法(Particle Swarm Optimization,PSO)和BP神经网络结合,构建EEMD-PSO-BP神经网络的组合网络模型,将组合神经网络的沉降预测结果作为观测值引入到卡尔曼滤波观测方程中,以建立混合预测模型。最后针对噪声方差Q与R选取的问题,利用新息自适应卡尔曼滤波估计出噪声方差的协方差阵。混合预测模型能有效减小单一预测机制造成的同一性质误差的累积,将基于ARIMA的新息自适应卡尔曼滤波、EEMD-PSO-BP神经网络模型与混合滤波模型的精度进行对比,新息自适应混合卡尔曼滤波预测模型的均方根误差降低至0.3194mm,相对百分误差降到1.42%。实验结果表明,混合滤波模型的各项预测结果要优于传统预测模型,精度相比较传统的预测模型有较大的改善。In order to solve the problem of surface subsidence deformation prediction in mining area and improve the accuracy of prediction model,a hybrid prediction method based on ARIMA adaptive Kalman filter and combined neural network is proposed.Firstly,in view of the non-stationarity and complexity of the subsidence deformation monitoring sequence,ARIMA model is able to stabilize the original sequence,so as to construct the prediction model of surface subsidence and serve as the state equation of the new adaptive Kalman filter.Then,PSO,EEMD and BP neural network are combined to build a combined network model of EEMD-PSO-BP neural network.Finally,for the selection of noise variance Q and R,the covariance matrix of noise variance is estimated by using the new adaptive Kalman filter.The hybrid predictive model can effectively reduce the accumulation of the same properties of the single predictive mechanism,based on the new interest of ARIMA to adapt to the Kalman filter,the EEMD-PSO-BP neural network model and the accuracy of the hybrid filter model,the balance of the hybrid prediction model is 0.3194 mm,and the relative error is reduced to 1.42%.The experimental results show that the prediction results of the hybrid filter model are better than that of the traditional model with significant improvements on the accuracy.

关 键 词：沉降预测 集合经验模态分解 新息自适应卡尔曼滤波 粒子群优化算法 BP神经网络 混合预测 

分 类 号：O141.4[理学—数学]