Obstacle-incorporated Spatiotemporal Characterization of Building Solar-energy-potential and Holistic Design Optimization of Large-scale Distributed PV Systems in High-density Cities

Description

Renewable energy applications are important for the decarbonization and sustainable development of high-density cities. To achieve its decarbonization goals, Hong Kong has proactively promoted solar energy applications by implementing a superhigh Feed-in-Tariff, i.e., 300% to 500% of the regular-electricity-price. Hong Kong has over 45,000 buildings but accurate characterization of their diverse solar-energy-potentials remains challenging and still lacking.Building solar-energy-potentials mainly come from rooftops and façades. Existing studies have estimated the respective potentials from their total surface areas. However, common obstacles (such as rooftop cooling towers and façade decorative structures), which can substantially reduce the usable areas and thus the realizable solar-energy-potentials, have either been neglected or oversimplified, thereby leading to low-accuracy characterization. In fact, city-scale obstacle identification (including detection, positioning and sizing) is challenging and cannot be realized via traditional methods because of the obstacles’ complex features, such as irregular shapes, random distributions and time-varying numbers. On the other hand, due to the lack of accurate characterization, existing design methods cannot fast and precisely identify buildings with richer  solar-energy-potentials or other desirable features, and thus cannot identify the most suitable buildings for PV applications, leading to unsatisfactory overall performance at district-level or above.To achieve accurate characterization, we will develop a deep learning and 3D-GIS (Geographic Information System) integrated method. Given its fast advancements and impressive performance in image recognition, deep learning will be adopted for obstacle identifications. However, deep learning alone may wrongly identify obstacles outside of rooftops and façades. Thus, 3D-GIS data, providing precise building geometries, locations and orientations, will be integrated to prevent such misidentifications. The identified obstacles will then be incorporated for accurate characterization of building solar-energy-potentials in both spatial and temporal perspectives. To achieve large-scale design optimization, we will base on our accurate characterization to develop holistic design optimization, enabling the consideration of massive buildings and distributed PV systems. The challenge of the excessive computational complexities caused by the massive buildings and systems considered, will be addressed by two approaches, i.e., decomposition-based serial optimization and approximation-based parallel optimization. For the computational complexity reduction, the first one decomposes a big problem into smaller ones while the second one approximates a computation-intensive problem as a simplified one. By leveraging state-of-the-art technologies to improve characterization accuracy and enhance large-scale design performance, this project will help accelerate renewable energy developments  and applications in high-density cities and thus contribute to associated urban decarbonization and  sustainable developments.