基于证据理论的状态估计方法及其在液位估计中的应用

State estimation based on DS theory:application to liquid level estimation

本文给出了一种基于证据理论的迭代算法估计噪声有界下的动态系统状态,并将其应用于工业液位仪的液位动态估计当中。该算法将系统的状态方程、观测方程及实际观测看作提供证据的3个信息源,利用证据的随机集表示及随机集扩展准则从信息源中构造关于系统状态和观测的证据:通过相关证据融合方法:将所构造的证据进行融合,并利用Pignistic期望从融合结果中计算出状态估计值。与基于置信函数的前反向传播算法相比,所提算法中的融合过程增加了估计的聚焦程度,使得估计结果更加精确。通过在液位估计中的应用说明新算法可以使得估计的相对误差提高约0.3个百分点。

This paper presents an iterative algorithm based on evidence theory to solve dynamic system state estimation problem under the bounded noise. This algorithm regards the state equation of the system, the observation equation and actual observation as three sources of information for providing evidences. It uses evidence description of random set and extension principles of random set to construct state evidence and observation evidence from sources of information. Furthermore, in this algorithm, the proposed evidence combination of dependence evidence is used to fuse those constructed evidence and Pignistic expectation of fusion results is calculated as value of state estimation. Compared with the estimation algorithm based on belief function and forward-backward propagation, our method increases the focus of state estimation, and makes the estimation results more accurate. Finally, through the application in the liquid level estimation that the new algorithm can make the relative error of the estimation increases about 0.3%.