Identification of structural parameters by using observation data with noise in time domain was studied. Autoregressive-moving-average (ARMA) time-series model of a vibrating structure was established, and the identification problem of structural parameters was transformed into the identification problem of parameters of ARMA mode1. Based on the assumption of the unknown but bounded noise, an interval algorithm for set-membership identification of parameters of linear time-invariant system was used to seek the minimal hyper-rectangle of parameters, which is compatible with the measurements and the bounded noise, so that the structural parameters can be obtained. Then the Taian pumped-storage power station engineering project illustrates its feasibility and effectiveness.