With the continuous increase of wind power penetration, the inertia level of power systems decreases significantly, posing new challenges to system frequency stability. To effectively evaluate the change of power system node inertia under wind power integration, this paper proposes an improved maximum likelihood estimation (MLE) based on autoregressive moving average with exogenous variable (ARMAX) to evaluate the inertia at nodes directly connected to generation units. First, an ARMAX model is constructed to represent the dynamic characteristics of the nodes directly connected to generation units. The improved MLE method is then applied to identify model parameters and estimate the corresponding node inertia. Then, based on the k-means clustering algorithm, the generator nodes are partitioned according to their inertia, allowing for the calculation of the inertia and center frequency of the system region. Furthermore, adaptive polynomial fitting is employed to estimate the node inertia of non-generator nodes based on their frequency behavior. Finally, the IEEE39-node system including wind turbine generator nodes is modeled, and the heatmap is drawn to visually display the inertia distribution of the power system nodes and regions, which verifies the effectiveness of the improved method in this paper. This approach enables accurate identification of dynamic responses at various nodes and provides strong support for the analysis and planning of wind-integrated power systems.