基于聚类和支持向量机的非线性时间序列故障预报

Nonlinear time series fault prediction based on clustering and support vector machines

针对非线性时间序列故障预报问题,提出了一种基于聚类和支持向量机的方法．将正常的时间序列按照K-均值聚类算法进行聚类学习,同时利用支持向量机回归的时间序列预测算法获得预测序列,然后通过比较聚类所得的正常原型和预测序列的相似性实现故障预报．仿真结果表明:本文提出的方法更能满足实时性的要求,也更为准确．

Based on clustering and support vector machines, a new method is proposed to solve the nonlinear time series fault prediction. The normal time series is clustered using K-means clustering algorithm to get the normal prototype. Meanwhile, the predicting series is obtained by time series predicting algorithm based on support vector regression. Fault prediction can also be implemented by calculating the similarity between the normal prototype and the predicting series. Finally, the simulation results indicate that the proposed method can predict the fault more quickly and more accurately.