基于时变自回归模型的非平稳数据预测方法研究  

Prediction Research of Non-stationary Data Based on Time-Varying Autoregressive Model

作　　者：王莉娜 姜相夺 芦玉华[2] 

机构地区：[1]陕西法士特汽车传动研究院 [2]西安华为技术有限公司

出　　处：《微型电脑应用》2014年第8期34-36,共3页Microcomputer Applications

摘　　要：针对现有的自回归(Autoregressive,AR)模型对非平稳数据预测效果不佳的问题,提出了基于时变自回归(Time-Varying Autoregressive,TVAR)模型的时序预测方法。针对某型国产飞机发动机的低压转速信号,使用TVAR模型分别进行点预测和区间预测,并与AR模型的点预测结果进行对比。研究结果表明,TVAR模型能够很好地反映非平稳数据的变化趋势。在给定置信水平下,TVAR预测区间能够包含真实数据,因此TVAR模型在时序预测中具有更好的预测效果。As the AR（Autoregressive） model has poor prediction effect on non-stationary data, a Time-Varying Autoregressivemodel is established for timing prediction. Point and interval predictions of the TVAR Model are conducted based on NL data ofan aero-engine, and are compared to point prediction of the AR model. The result shows that the TVAR model can reflect the realtrend of the data, and the prediction errors are in the allowable range. At a given confidence level, the TVAR prediction intervalcontains real data, so the TVAR model has better prediction results in the timing projections.

关 键 词：时间序列 时变自回归模型 预测 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]