Ecosystem health is the goal of ecological environment management. An accurate and objective evaluation of ecosystem health is necessary to improve the regional ecological environment. As a semi-arid ecological transition area, the Agro-Pastoral Ecotone of North China connecting the agricultural region in the east with the pastoral area in the west, the ecological environment is sensitive and fragile due to the influence of human activities. Exploring the impact of the national ecological restoration project of the Grain-for-Green program on the ecosystem health of the Agro-Pastoral Ecotone of North China, determining the strategies of ecological restoration and resource management, it will contribute to promote regional sustainable development. Therefore, we built a comprehensive system, based on the "pressure-state-response" (PSR) model, including natural, social and economic aspects to assess the ecological health index of the Agro-Pastoral Ecotone of North China in 2000, 2005, 2010 and 2015. We used Moran's I to analysis the spatial correlation between the Grain-for-Green program and the regional ecological health. The optimal regression model is selected from the ordinary least squares model (OLS), the geographically weighted regression model (GWR), the temporally weighted regression (TWR), the geographically and temporally weighted regression (GTWR), to reveal the influence of the Grain-for-Green program (△NDVI) on the change in ecological health (△EHI). The running result of optimization showed that:(1) from 2000 to 2015, the ecosystem health index of Agro-Pastoral Ecotone of North China shows a trend of "decreasing and then increasing", and the region is in subhealthy and general health, and the state of ecosystem health is gradually improving. (2) The results of the bivariate spatial autocorrelation analysis(the bivariate Moran's I)showed that the impact of the Grain-for-Green program on ecological health was positive correlation during the beginning and development stages of the project from 2000 to 2010, while from 2010 to 2015, the relationship of them became negative. (3) Compared with various spatial regression models, the geographically and temporally weighted regression (GTWR) models performed the best, with a significant positive spatial effect of the reforestation project on ecosystem health in the northeastern part of the study area between 2000 and 2010, with the most positive effect in the northeastern part of the study area, but a positive to negative effect in the central part of the study area between 2010 and 2015. In the central region, the positive effect was maintained, while the southwestern region showed a fluctuating pattern of "negative-positive-negative effect". The results of the study can provide a reference for the scientific promotion of ecosystem structure optimization and ecological civilization construction in the Agro-Pastoral Ecotone of North China.