基于SCKF的Elman递归神经网络在软测量建模中的应用

发布时间：2018-03-21 14:46

本文选题：平方根容积卡尔曼滤波　切入点：递归神经网络　出处：《信息与控制》2017年03期 　论文类型：期刊论文

【摘要】：针对具有强非线性、复杂的化工过程软测量建模,提出一种基于平方根容积卡尔曼滤波(SCKF)的递归神经网络方法.基于Elman递归神经网络,首先构建状态空间模型,然后应用SCKF算法进行训练,所有网络的权值将作为系统的状态进行更新.容积卡尔曼滤波(CKF)通过三阶Spherical-Radial容积准则生成容积点,利用容积点逼近状态的后验分布,使得高维非线性滤波中的多变量积分数值求解成为可能.在CKF的基础上,SCKF采用预测及后验误差协方差矩阵的平方根因子进行递推运算,进一步改进了算法的数值稳定性.将该方法应用于脱丁烷塔底部丁烷组分含量以及硫回收装置尾气中SO2和H2S含量的软测量动态建模实例中,在同等条件下,还与基于EKF、SCKF的前馈神经网络,基于EKF的递归神经网络等其它方法对比.结果表明,本文的方法能够获得很好的建模精度,显示出其有效性.
[Abstract]:In view of the soft sensor modeling of chemical process with strong nonlinear and complex characteristics, a recursive neural network method based on square root volume Kalman filter (SCKF) is proposed. Based on Elman recurrent neural network, the state space model is first constructed. Then the SCKF algorithm is used to train, and the weights of all networks will be updated as the state of the system. The volume Kalman filter (CKF) generates the volume points through the third-order Spherical-Radial volume criterion, and uses the volume points to approximate the posteriori distribution of the state. It is possible to solve multivariable integral numerically in high dimensional nonlinear filtering. On the basis of CKF, the square root factor of prediction and posteriori error covariance matrix is used for recursive operation. The numerical stability of the algorithm is further improved. The method is applied to the soft sensing dynamic modeling examples of the content of butane at the bottom of the debutane tower and the content of SO2 and H2S in the tail gas of the sulfur recovery unit, under the same conditions, It is also compared with other methods such as feedforward neural network based on EKF and recurrent neural network based on EKF. The results show that the proposed method can achieve good modeling accuracy and show its effectiveness.
【作者单位】：兰州交通大学自动化与电气工程学院;
【基金】：国家自然科学基金资助项目(51467008) 光电技术与智能控制教育部重点实验室(兰州交通大学)开放课题(KFKT2016-3)
【分类号】：TP183;TQ02

【相似文献】