Full metadata record
|dc.description.abstract||Investigation of suitable regions for solar photovoltaic plants highly depends on the accurate prediction of potential solar energy at candidate sites. However, regional effects caused by adjacent terrain and weather conditions that can affect the solar energy still have rarely been modeled in the machine learning-based estimation process. Also, it remains unknown to investigate the site selection for solar power generation using photovoltaic power output directly for energy management. Thus, this paper proposes computational methods (e.g., convolutional neural network) that are designed to predict the solar irradiation and solar power output with raster image-based meteorological and geographic maps representing regional effects. In addition, this paper also presents a geospatial analysis approach for infrastructure planning using building information model and geographic information system. The results show that the proposed models achieve a mean absolute percent error of 0.470% in the solar irradiation prediction model, and 8.639% in the solar power output forecasting model. In particular, the potential production of solar power generation calculates 6,783,725 MWh within a 200m radius in the vicinity of the highway network at a national scale. It implies that learning the spatial patterns of geographical and meteorological features may emerge to affect solar energy forecasting and that the trained model can be utilized to select suitable regions for the installation of solar power plants. Thus, this study may assist energy management and planning in infrastructure planning.||-|
|dc.title||Optimal site selection for solar photovoltaic power plants using machine learning techniques and GIS-BIM integration approaches||-|
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.