针对一类输入激振力不可测的工程结构，论述了基于向量自回归（ＡＲＶ）时间序列模型的结构模态参数识别的基本原理。为考察ＡＲ系数估计的精度，利用中心极限定理建立了估计的自回归系数矩阵协方差的表达式。分别利用水轮机的随机响应信号和自由响应信号识别了模态参数。数值仿真结果表明本文给出的识别方法是鲁棒的。此外，给出了试验结果。 Aiming at a kind of engineering structure whose input excitation force is not measurable, the basic principle of structure modal parameter identification based on vector autoregression (ARV) time series model is discussed. In order to examine the accuracy of the AR coefficient estimation, an estimate of the covariance matrix of the autoregressive coefficients was established using the central limit theorem. Modal parameters were identified using the turbine’s random response signal and free response signal, respectively. Numerical simulation results show that the identification method given in this paper is robust. In addition, the experimental results are given.