论文部分内容阅读
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。
In this paper, an iterative algorithm based on evidence theory is proposed to estimate the dynamic state of a system under noise boundedness and to apply it to the dynamic level estimation of industrial liquid level meters. The algorithm takes the state equation, observation equation and actual observation of the system as three information sources that provide evidence, and uses the random set representation of evidence and the extension rule of random set to construct the evidence about the state and observation of the system from the information source: Fusion method: The constructed evidence is fused and Pignistic expectations are used to calculate the state estimates from the fusion results. Compared with the pre-backpropagation algorithm based on confidence function, the fusion process in the proposed algorithm increases the degree of focus of the estimation and makes the estimation more accurate. The application of the new algorithm in the estimation of liquid level can make the estimated relative error increase about 0.3%.