摘要
为减少计算量,及时预报系统故障,提出了一种基于聚类粒子滤波器的故障预报方法.在粒子滤波中引入k均值聚类算法,将粒子集中的所有粒子聚为若干类,以每一类粒子的质心代表该类粒子参与每次的粒子更新,显著减少了参与迭代计算的粒子数目,从而减少了粒子滤波算法的计算量.仿真结果验证了基于聚类粒子滤波器的故障预报方法的有效性.
A fault prediction method based on clustering particle filter is put forward to reduce computing cost and predict system faults in time. The k-means clustering algorithm is introduced into the particle filtering, and all the particles are classified into some clusters. The centroid of each cluster regarded as the representative of the particle cluster is updated each time. The number of particles needed to be iteratively computed is decreased sharply, thus the computing cost of particle filtering is reduced. Simulation results verify the validity of the presented clustering-particle-filter-based fault prediction method.
出处
《信息与控制》
CSCD
北大核心
2009年第1期115-120,共6页
Information and Control
基金
国家自然科学基金资助项目(60736026)
"教育部新世纪优秀人才支持计划"资助项目.
关键词
粒子滤波
聚类
K均值
计算量
故障预报
particle filtering
clustering
k-means
computing cost
fault prediction
