基于KPCA与LSSVM的网络控制系统时延预测方法

Networked control system time-delay prediction method based on KPCA and LSSVM

针对网络控制系统中随机时延很难精确预测的问题,首次将核主成分分析(kernel principal compo-nent analysis,KPCA)与最小二乘支持向量机(least squares support vector machine,LSSVM)结合对随机时延进行预测,KPCA对输入随机时延序列降维,消除重复性与噪声,减少LSSVM的运算量,降维后的时延序列通过LSSVM算法预测时延值。仿真结果表明,基于KPCA与LSSVM的时延预测方法的预测精度高于其他的预测方法。

The random delay of networked control system is difficult to predict accurately. Firstly the kernel prin- cipal component analysis （KPCA） and the least squares support vector machine （LSSVM） algorithm are combined to predict the random time-delay. The KPCA can reduce dimensionality of input random time-delay sequence, eliminate the noise and interference and reduce the amount of computation of the LSSVM algorithm. The time-delay is predicted by the LSSVM algorithm through the time-delay sequence after dimensionality reduction. The simulation results show that the prediction accuracy of the time-delay prediction based on the KPCA and LSSVM algorithm is higher than the other prediction methods.