Deep Learning-Based Assessment of Distributed PV Installation Potentials

With the global energy structure transforming to clean and low-carbon, distributed photovoltaic (PV) power generation systems are playing an increasingly important role in urban energy supply due to their advantages of environmental friendliness and efficient use of space, etc. The aim of this paper is to explore deep learning-based techniques for assessing the potential of large-scale rooftop distributed photovoltaic (PV) installations. In order to accurately predict the electricity demand of users, a composite prediction model combining variational modal decomposition (VMD) and Crossformer network is adopted. The model first preprocesses the raw electricity consumption data through the VMD to decompose the modal components of different frequencies, thus effectively reducing noise interference and improving data quality; subsequently, the Crossformer network is used to capture the long-term dependencies in the time series and realize high-precision prediction of electricity demand. Based on the prediction model and utilizing the supply and demand relationship between installed capacity and electricity demand, this paper also takes into account a variety of factors such as available roof area, sunlight conditions, building type, and geographic location to construct a framework for evaluating the potential of PV installations. The framework is designed to estimate the installed rooftop distributed PV capacity.