摘要
通过对常规时间序列周期性判断方法的研究,提出运用变宽度移动数据窗和最小二乘平滑判断方法对时间序列的周期性进行判断。此方法易于从序列中剔除周期项获取随机项,并对随机项建立时间序列预测模型进行预测。通过此方法在某振动时间序列中的应用,证明其不仅可以实现对时间序列特性的在线判断与预测,而且有利于在预测过程中克服由于序列特性改变对结果的影响。
In this paper, through researching the measuring method of periodicity in the convention time series, the method which smooth the time series and judge it by using the mutative extent window function and least square has been proposed. Using this method, the stochastic term of the time series is easily gained by eliminating the periodic term, and model the time series models to forecast it. Applying this method to the some time series of vibration, the result proved that this method could not only carry out judging and forecasting on-line efficiently but overcome the influence which is brought by change of the sequence characteristic in forecasting.
出处
《东北电力大学学报》
2006年第1期51-55,共5页
Journal of Northeast Electric Power University
