This study investigates the distribution and impact of photovoltaic (PV) stations in the arid Northwest China, a crucial area for regional economic cooperation. A hierarchical extraction method combining RegNet and SAM models achieved 91.89% accuracy in PV station identification, while also showing a broader extraction coverage compared to existing datasets. In 2023, we identified 688 centralized PV stations, covering a total area of 719.28 km(2). Using the Continuous Change Detection and Classification (CCDC) algorithm along with Global Moran's I, we observed significant development in PV installations between 2013 and 2021, with smaller stations being more spatially dispersed. Ecological analysis revealed that PV stations were predominantly situated within the Gobi area/desert, with a minor proportion located in low-coverage grasslands. The impacts of PV stations on local temperature exhibited both locational and seasonal variations. Temperature variations between PV stations and their 1 km buffer zones were significant, with over half of the PV stations contributing a cooling effect on their surroundings (50.78% in summers and 58.79% in winters). Additionally, vegetation coverage increased with distance from the PV stations, which indicated a substantial ecological interaction, underscoring the potential advantages and complexities associated with PV deployment in arid ecosystems.