Inertia is one of the important parameters which can ensure that a power system frequency is in a stable state. So it needs to be measured accurately online. Aiming at the low measurement accuracy of the current system identification method, the influence of the algorithm model in system identification on the measurement result is studied. First, the principles of inertia identification based on transfer function model, auto-regressive moving average model and subspace identification model are analyzed and compared. Secondly, a state space estimation model based on a PEM iterative algorithm is proposed from the perspective of data matching at the initial stage of the inertia response. Finally, a 10-machine and 39-bus power simulation system is built to verify the correctness of the proposed identification model. The applicability of the four identification models is analyzed under different power disturbance degrees and sampling time windows. It provides a reference for power system operators to determine the optimal identification model. This work is supported by the National Natural Science Foundation of China (No. 51977128).