Obtaining the nadir point of the frequency response trajectory is particularly important in low-inertia power systems, which can be used as an important indicator of whether the stable operation of the power grid can be maintained. Therefore, an online prediction method of frequency nadir is proposed in this paper. Firstly, the reasons for the generation and characterization of the nadir features of the frequency response trajectory are revealed, and then the analytical expression for the frequency nadir is derived by using a generalized frequency response model. For the unknown parameters in the expression, a sparse-measurement-based frequency modeling method is utilized for construct the dataset for parameter identification. Since the proposed method utilizes the initial frequency trajectory after perturbation to predict the frequency nadir point, the system’s own response data can improve the accuracy of the prediction results. Case tests demonstrate the effectiveness of the proposed method.