混合系统二维离散与连续状态同步估计研究

Synchronous Estimation of Two Dimensional Discrete States and Continuous States for Hybrid Systems

在混合系统中,需要同时估计出系统的离散状态与连续状态。针对混合系统出现二维离散状态下的混合状态估计问题进行研究,根据系统特性,采用跳变马尔可夫线性系统建模,并应用Rao-Blackwellised粒子滤波算法对二维离散状态与连续状态进行同步估计。由于算法一定程度上缓解了粒子滤波算法在高维状态空间估计中的失效问题,并对离散状态单独采样,能提高系统状态的估计精度。仿真试验证明,方法能有效地同步估计出系统的二维离散状态与连续状态,其中,二维离散状态的估计准确率达到了96%。

Discrete states and continuous states need to be estimated synchronously in the fault diagnosis of hybrid systems. To estimate two - dimensional discrete states and continuous states, jump Markov linear Gaussian model and an efficient Monte Carlo simulation known as Rao - Blackwellised particle filtering are used. This algorithm overcomes some shortcomings of particle filter such as inefficiency in high dimensional state spaces, and increases the accuracy of fault diagnosis. Computer simulations prove that both two - dimensional discrete states and continuous states can be estimated synchronously and effectively, and the accuracy ratio of the discrete states can reach 96%.