Volume 13: Proceedings Applied Energy Symposium: CUE2020, Part 2, Japan/Virtual, 2020

Estimation of Rooftop Solar Potential using Publicly Available Geodata and Deeping Learning Zhixin Zhang, Teng Zhong, Min Chen*, Zixuan Zhou, Yijie Wang, Kai Zhang

Abstract

Rooftop solar photovoltaic power generation provides a feasible solution for the sustainable development of the city. The estimation of rooftop solar potential is of great significance to the formulation of urban energy plans. Quantifying the rooftop area is the basis of estimating the rooftop solar potential, but how to extract the rooftop information quickly in large-scale is still a challenge. In this study, a scalable framework is used to estimate the rooftop solar potential based on Google Earth satellite images. This framework uses a deep learning semantic segmentation method to extract the rooftop, which provides support for estimating the solar potential of the rooftop. In order to reduce the labor cost invested in the training process of the rooftop extraction model, a training data acquisition strategy was developed based on prior knowledge of the urban and rural spatial layout and landuse. This paper takes Nanjing, China as an example to make an empirical analysis. The results show that the framework can achieve good rooftop extraction effect. It is also found that the solar potential of buildings in Nanjing is huge.

Keywords rooftop solar potential, geographic information systems (GIS), deep learning, sampling strategy, city scale

Copyright ©
Energy Proceedings