Zhuo Sun

Urban expansion and fossil fuel dependence have led to energy and environmental concerns, highlighting the need for sustainable solutions. Rooftop photovoltaic (RPV) systems offer a viable solution for urban energy transition by utilizing idle rooftop space and meeting decentralized energy needs. However, due to limited information on building function attributes, detailed assessments of RPV potential at the city scale are still complicated. This study introduces a cost-effective framework that combines big Earth data and deep learning to evaluate RPV potential for various investment entities. We introduced a sample construction strategy that considers built environment and location awareness to improve the effectiveness and generalizability of the framework. Applied to Shanghai, our building function recognition model achieved 88.67%, 88.51%, and 67.18% for accuracy, weighted-F1, and kappa, respectively. We identified a potential installed capacity of 42 GW with annual electricity generation of 17 TWh for industrial and commercial, 16 TWh for residential, and 10 TWh for public RPVs. The levelized cost of electricity ranges from 0.32 to 0.41 CNY/kWh, demonstrating that both user-side and plant-side grid parity was achieved. This study supports sustainable development by providing detailed urban energy assessments and guiding local energy planning. The methods and findings may offer insights for similar studies globally.

Accurate and timely access to building rooftop information is very important for urban management. The era of big data brings new opportunities for rooftop extraction based on deep learning and high-resolution satellite imagery. However, collecting representative datasets from such big data to train deep learning models efficiently is an essential problem that still needs to be explored. In this study, geospatial stratified and optimized sampling (GSOS) based on geographical priori information and optimization of sample spatial location distribution is proposed to acquire representative samples. Specifically, the study area is stratified based on land cover to divide the rooftop-dense stratum and the rooftop-sparse stratum. Within each stratum, an equal amount of samples is collected and their spatial locations are optimized. To evaluate the effectiveness of the proposed strategy, several qualitive and quantitative experiments are conducted. As a result, compared with other common sampling approaches (e.g., random sampling, stratified random sampling, and optimized sampling), GSOS is superior in terms of the abundance and types of collected samples. Furthermore, two quantitative metrics, the F₁-score and Intersection over Union (IoU), are reported for rooftop extraction based on deep learning methods and different sampling methods, in which the results based on GSOS are on average 9.88% and 13.20% higher than those based on the other sampling methods, respectively. Moreover, the proposed sampling strategy is able to obtain representative training samples for the task of building rooftop extractions and may serve as a viable method to alleviate the labour-intensive problem in the construction of rooftop benchmark datasets.