امكان سنجي بهبود توان تفكيك مكاني نقشه‌هاي پتانسيل انرژي خورشيدي حاصل از مدل‌هاي رقومي ارتفاعي جهاني با استفاده از روش يادگيري عميق

In response to the environmental issues caused by the burning of fossil fuels, there has been a significant focus on utilizing renewable energy sources such as solar energy. To identify suitable locations for installing solar panels, maps of solar energy potential with appropriate spatial resolution are required for various urban, non-urban, and highway fringe areas. Generating accurate maps of solar energy potential relies on models that estimate solar radiation levels, which necessitate high-resolution Digital Surface Models (DSMs). However, obtaining highly-resolved DSMs, such as LiDAR-derived digital elevation models, for all regions worldwide is not feasible, and even if available, it entails significant financial and computational costs.In this study, globally available DSMs with an approximate spatial resolution of 30 meters were used to generate maps of annual average solar energy potential using solar radiation estimation models in the study area of the Netherlands. Subsequently, the spatial resolution of the solar energy potential maps derived from the global DSM models was improved using deep learning-based super resolution techniques such as EDSR (Enhanced Deep Super-Resolution) and U-Net-based super resolution models with different backbones like Resnet 18, 34, 50, and 101. The spatial resolution of the solar energy potential maps was improved five times. After eva‎luating the accuracy of the solar energy potential maps derived from the global DSM models compared to the solar energy potential map obtained from the LiDAR-derived DSM model, the solar energy potential map from the Copernicus DSM was selected. The super resolution models were trained and tested based on the solar energy potential maps obtained from the Copernicus DSM. After eva‎luating the accuracy of the super resolution models, the EDSR model was chosen as the more accurate and stable model (Test-rRMSE = 0.1331 KWh/m2, Test-rMAE = 0.0283 KWh/m2) for improving the spatial resolution of the solar energy potential map derived from the Copernicus DSM.In the prediction stage of the EDSR model, the solar energy potential map derived from the Copernicus DSM was utilized. The eva‎luation of the results of the super resolution of the solar energy potential map derived from the Copernicus DSM using the EDSR model indicated that the solar energy potential maps with a spatial resolution of 6 meters not only had a higher spatial resolution compared to the solar energy potential map from the Copernicus DSM with a spatial resolution of 30 meters but also exhibited higher accuracy, particularly in urban areas and highway fringes, compared to the solar energy potential map derived from the LiDAR-derived DSM model. Furthermore, the solar panel placement mapping based on the solar energy potential map derived from the Copernicus DSM with a spatial resolution of 6 meters (predicted by the EDSR model) demonstrated higher accuracy and correlation compared to the mapping based on the solar energy potential map derived from the Copernicus DSM with a spatial resolution of 30 meters in comparison to the mapping based on the LiDAR-derived DSM model.

5