Abstract: In recent years,due to the fact that wind farms are often built around newly built wind farms and old wind farms that require technological upgrades,wind farms not only need to avoid national restricted construction areas such as ecological protection red lines,permanent basic farmland,nature reserves,scenic spots,etc.,meanwhile need to consider the wake effects between wind turbines on site,but also need to focus on the wake impact of surrounding wind farms on newly built or technological upgrade wind farms.When calculating the impact of wake flow,engineers using local remote sensing images to manually mark the surrounding wind turbines of the existing wind farms one by one can easily result in missing statistics of wind turbine positions and significant gross error. This paper proposes a method for identifying wind farm avoidance objects based on deep learning models. The deep learning model is used to identify and label sensitive factors such as built wind turbines in the wind farms and residential buildings in remote sensing images. OpenWind software is used to avoid these sensitive factors while quickly arranging wind turbines. The research results show that the use of computers to analyze and label avoidance objects is accurate,less prone to omissions,and can effectively eliminate errors caused by negligence. At the same time,it significantly reduces the workload of lable avoidance objects,effectively reducing a lot of complex,inefficient,and highly repetitive work,and greatly improving the working efficiency and accuracy of identifying avoidance objects.