3D-PV-Locator: Large-scale detection of rooftop-mounted photovoltaic systems in 3D. (15th March 2022)
- Record Type:
- Journal Article
- Title:
- 3D-PV-Locator: Large-scale detection of rooftop-mounted photovoltaic systems in 3D. (15th March 2022)
- Main Title:
- 3D-PV-Locator: Large-scale detection of rooftop-mounted photovoltaic systems in 3D
- Authors:
- Mayer, Kevin
Rausch, Benjamin
Arlt, Marie-Louise
Gust, Gunther
Wang, Zhecheng
Neumann, Dirk
Rajagopal, Ram - Abstract:
- Abstract: While photovoltaic (PV) systems are being installed at an unprecedented rate, it is challenging to keep track of them due to their decentralized character and large number. In this paper, we present the 3D-PV-Locator for large-scale detection of roof-mounted PV systems in three dimensions (3D). The 3D-PV-Locator combines information extracted from aerial images and 3D building data by means of deep neural networks for image classification and segmentation, as well as 3D spatial data processing techniques. It thereby extends existing approaches for the automated detection of PV systems from aerial images by also providing their azimuth and tilt angles. We evaluate the 3D-PV-Locator using a large dataset gathered from the official German PV registry in a real-world study with more than one million buildings. In terms of azimuth and tilt angles, our evaluation shows that the 3D-PV-Locator and the official registry coincide for about two thirds of the observations and are within neighboring classes for 84 and 99 percent of the observations, respectively. In terms of detected PV system capacity, we show that the 3D-PV-Locator clearly outperforms existing approaches. It performs particularly well for the groups of small and medium-sized PV systems (3.6–33.1 percent error reduction) and PV systems tilted beyond 40° (25.6–38.1 percent error reduction). The 3D PV system data generated by the 3D-PV-Locator can inform several practical applications, such as improvedAbstract: While photovoltaic (PV) systems are being installed at an unprecedented rate, it is challenging to keep track of them due to their decentralized character and large number. In this paper, we present the 3D-PV-Locator for large-scale detection of roof-mounted PV systems in three dimensions (3D). The 3D-PV-Locator combines information extracted from aerial images and 3D building data by means of deep neural networks for image classification and segmentation, as well as 3D spatial data processing techniques. It thereby extends existing approaches for the automated detection of PV systems from aerial images by also providing their azimuth and tilt angles. We evaluate the 3D-PV-Locator using a large dataset gathered from the official German PV registry in a real-world study with more than one million buildings. In terms of azimuth and tilt angles, our evaluation shows that the 3D-PV-Locator and the official registry coincide for about two thirds of the observations and are within neighboring classes for 84 and 99 percent of the observations, respectively. In terms of detected PV system capacity, we show that the 3D-PV-Locator clearly outperforms existing approaches. It performs particularly well for the groups of small and medium-sized PV systems (3.6–33.1 percent error reduction) and PV systems tilted beyond 40° (25.6–38.1 percent error reduction). The 3D PV system data generated by the 3D-PV-Locator can inform several practical applications, such as improved forecasting of solar generation, the optimized planning and operation of distribution networks, improved integration of electric vehicles, and others. All datasets and pre-trained models associated with this paper are available online. Graphical abstract: Highlights: Methodology for large-scale detection of solar panels in three dimensions. Solar panel information is extracted from aerial images and 3D building data. Extension of existing PV detection approaches by providing azimuth and tilt angles. Improved solar panel area and capacity estimates, especially for residential units. All associated datasets, models, and code are publicly available. … (more)
- Is Part Of:
- Applied energy. Volume 310(2022)
- Journal:
- Applied energy
- Issue:
- Volume 310(2022)
- Issue Display:
- Volume 310, Issue 2022 (2022)
- Year:
- 2022
- Volume:
- 310
- Issue:
- 2022
- Issue Sort Value:
- 2022-0310-2022-0000
- Page Start:
- Page End:
- Publication Date:
- 2022-03-15
- Subjects:
- Solar panels -- Renewable energy -- Image recognition -- Deep learning -- Computer vision -- 3D building data -- Remote sensing -- Aerial imagery
Power (Mechanics) -- Periodicals
Energy conservation -- Periodicals
Energy conversion -- Periodicals
621.042 - Journal URLs:
- http://www.sciencedirect.com/science/journal/03062619 ↗
http://www.elsevier.com/journals ↗ - DOI:
- 10.1016/j.apenergy.2021.118469 ↗
- Languages:
- English
- ISSNs:
- 0306-2619
- Deposit Type:
- Legaldeposit
- View Content:
- Available online (eLD content is only available in our Reading Rooms) ↗
- Physical Locations:
- British Library DSC - 1572.300000
British Library DSC - BLDSS-3PM
British Library HMNTS - ELD Digital store - Ingest File:
- 21079.xml