本文介绍了在具有时变均值和时变方差的随机干扰序列作用下,ARMA模型的在线辨识与自校正预报。用简单的ELS方法实现模型的在线辨识。对指数平滑预报器一步预报误差序列进行实时建模和实时修正的基础上,实现对未来值的自校正预报。应用于雷达测量飞机低空飞行数据的短期预报,仿真和对实测数据的处理结果表明,此方法简单且精度较高,适用于微型机对一类非平稳过程的实时建模和短期预报。 This paper introduces the on-line identification and self-tuning prediction of ARMA model under the random disturbance sequence with time-varying mean and time-varying variance. Use Simple ELS Method to Realize Online Identification of Model. Based on the real-time modeling and real-time correction of one-step prediction error sequence of exponential smoothing predictor, the self-correction forecast of future value is realized. The short-term forecast, simulation and data processing of the measured data of low-altitude flight radar used in radar measurement show that the method is simple and accurate and suitable for real-time modeling and short-term forecasting of a type of non-stationary process.