Studying the reduction effect of green infrastructure on PM_2.5 can provide theoretical support for sustainable development strategies to address climate change in urban agglomeration. First, taking the built-up areas of urban agglomeration in the Yangtze River's middle reaches as an example, this study systematically examined the spatiotemporal evolution characteristics of PM_2.5 concentration in urban agglomeration from 2000 to 2020 using built-up area data, land cover data, and PM_2.5 data. Second, we took five types of green infrastructure, including woodland, grassland, cultivated land, wetland, and waters, as driving factors and employed the factor detection and interaction detection functions in the geographic detector model to explore the reduction effects of green infrastructure on PM_2.5 concentration in the built-up area of urban agglomeration. Third, this study further analyzed the constraint effect of urbanization level on green infrastructure in the built-up areas by combining nighttime lighting data and the constraint line method. The results show that:(1) during 2000-2020, the annual average PM_2.5 concentration of the urban agglomeration in the Yangtze River's middle reaches showed an "inverted U-shaped" trend of rising and then decreasing in time series, and the spatial divergent characteristics of decreasing from the northwest to the southeast. (2) During 2000-2020, the green infrastructure in the built-up areas of the urban agglomeration in the Yangtze River's middle reaches had a reduction effect on PM_2.5 concentration, and the reduction rate did not exceed 4%. The reduction effect of the expansion area was significantly higher than that of the old urban areas. (3) The results of factor detection showed that in the urban agglomeration in the Yangtze River's middle reaches, the explanatory power of each green infrastructure factor on PM_2.5 concentration showed a pattern of woodland>grassland>cultivated land>wetland and waters in the old urban area and a law of woodland>grassland>cultivated land and waters>wetland in the expansion area. The results of interaction detection showed that woodland and wetland, woodland and grassland, and woodland and waters were the most significant interaction combinations to reduce PM_2.5 concentration, and their interaction explanatory power reached above 0.5 in both the old urban area and the expansion area. (4) The reduction effect of green infrastructure on PM_2.5 concentration in the built-up areas of the urban agglomeration in the Yangtze River's middle reaches was constrained by the urbanization level. The shape of the constraint line was exponential in both the old urban area and the expansion area. This study can provide a basis for the implementation of Natural Climate Solutions (NCS) at the urban agglomeration scale.