摘要
针对WSN流量预测,基于AR模型提出一种WSN流量双卡尔曼并行递推预测算法。该算法使用两个Kalman滤波器,交替进行AR模型参数的递推辨识与时变数据中真实值的最优估计,根据序列数据的最新信息实时修正AR模型参数进行动态预测。同时针对大步长的流量预测,引入滚动修正思想,克服动态预测算法存在间隔时间过长的缺点,降低多步预测误差。实验研究表明,利用研究的双卡尔曼并行递推算法使用AR模型进行多步预测,从原理设计和实现算法上,实现了WSN流量的准确预测。
Aimed to the prediction of WSN traffic,advanced a REPK dynamic prediction arithmetic of WSN traffic based on AR model. The arithmetic used two Kalman filter to recursively identify parameters of AR model and optimally estimate actual data in time-varying data,it can use fresh information to amend parameters of AR model real-timely and predict dynamically. Aimed to the prediction of time series signal with long delay,it applies scroll-amendment method,it solves great intervals problem of REPK arithmetic and decreases multi-step prediction error. The experiment results show the proposed REPK estimation arithmetic based on AR model realizes the accurate prediction of WSN traffic by principle design of prediction method and skillful arithmetic.
出处
《计算机技术与发展》
2012年第10期165-168,172,共5页
Computer Technology and Development
基金
中国博士后基金项目(20080440754)
东莞市科技计划项目(201010815400156)
广东轻工职业技术学院2010年科技基金(KJ201022)
关键词
AR模型
无线传感器网络
卡尔曼
预测
网络流量
AR model
wireless sensor networks （ WSN ）
Kalman
prediction
network traffic
