分析了传统时间序列分析法建立ARMA模型的不足,提出了一种利用模型阶数判断准则和长自回归法建模的新方法。对数据进行了预处理和模型识别,比较了不同模型阶数判断准则的特点,研究了用长自回归法估计模型参数的步骤和方法。最后对光纤陀螺随机漂移进行了建模实验,通过对模型残差的白噪声检验,证实了使用新方法建立的ARMA模型具有很好的适用性。 The deficiency of ARMA model established by traditional time series analysis is analyzed, and a new method of modeling using the order of model order and long autoregressive method is proposed. The preprocessing and model identification of the data were carried out. The characteristics of the judgment criteria of different order of the model were compared. The steps and methods of estimating the model parameters by the long autoregressive method were studied. Finally, the random drift of fiber optic gyroscope was modeled experimentally. The white noise test of the model residual proved that the ARMA model established by the new method has good applicability.