Deep learning in the built environment: automatic detection of rooftop solar panels using Convolutional Neural Networks

Roberto Castello, Martin Esguerra, Adrian Guerra, Jean-Louis Scartezzini
IOP Science, 2019
Article de conférence

Résumé

Mapping the location and size of solar installations in urban areas can be a valuable input for policymakers and for investing in distributed energy infrastructures. Machine Learning techniques, combined with satellite and aerial imagery, allow to overcome the limitations of surveys and sparse databases in providing this mapping at large scale. In this paper we apply a supervised method based on convolutional neural networks to delineate rooftop solar panels and to detect their sizes by means of pixel-wise image segmentation. As input to the algorithm, we rely on high resolution aerial photos provided by the Swiss Federal Office of Topography. We explore different data augmentation and we vary network parameters in order to maximize model performance. Preliminary results show that we are able to automatically detect in test images the area of a set of solar panels at pixel level with an accuracy of about 0.94 and an Intersection over Union index of up to 0.64. The scalability of the trained model allows to predict the existing solar panels deployment at the Swiss national scale. The correlation with local environmental and socio-economic variables would allow to extract predictive models to foster future adoption of solar technology in urban areas.

Official source

https://infoscience.epfl.ch/record/272682?ln=fr

À propos de ce résultat

Cette page est générée automatiquement et peut contenir des informations qui ne sont pas correctes, complètes, à jour ou pertinentes par rapport à votre recherche. Il en va de même pour toutes les autres pages de ce site. Veillez à vérifier les informations auprès des sources officielles de l'EPFL.

Concepts associés

Chargement

Publications associées

Chargement

Quantification of the suitable rooftop area for solar panel installation from overhead imagery using Convolutional Neural Networks

Riccardo Cadei, Roberto Castello, Shasha Jiang, Jean-Louis Scartezzini, Alina Ursula Rosa Walch

The integration of solar technology in the built environment is realized mainly through rooftop-installed panels. In this paper, we leverage state-of-the-art Machine Learning and computer vision techniques applied on overhead images to provide a geo-localization of the available rooftop surfaces for solar panel installation. We further exploit a 3D building database to associate them to the corresponding roof geometries by means of a geospatial post-processing approach. The stand-alone Convolutional Neural Network used to segment suitable rooftop areas reaches an intersection over union of 64% and an accuracy of 93%, while a post-processing step using building database improves the rejection of false positives. The model is applied to a case study area in the canton of Geneva and the results are compared with another recent method used in the literature to derive the realistic available area.

IOP PUBLISHING LTD2021

Chargement

, ,

Solar energy is a prime way to reach net zero energy buildings. Addressing buildings’ energy needs is a priority because buildings are the largest demander of energy (followed by industry then transport). However, harvesting energy from solar farms, implies covering large areas with standard black or blue solar modules, which is generally not well accepted by the stakeholders. If architectural coherence is to be preserved, building envelopes cannot simply be covered with black panels with visible cells and contact bars. Today, panels of different colors are available to integrate solar energy smoothly into the built environment. However, while a choice of colors is good news for architecture, it should not cause excessive energy losses. The Kromatix panels presented in this paper are covered by glass that combines effects of diffuse surfaces and interference filters. With a minimal loss of efficiency, these panels pave the way to new considerations in solar architecture.

Elsevier2017

Chargement

What will the cities of tomorrow look like? This is a crucial question because in the four and a half billion years of its existence, the Earth has never been so threatened, as it is today. Since 2008 half of the world's population has lived in towns – more than 70% in rich nations. Three quarters of CO2 emissions are produced in towns, 75% of the world's energy is consumed in towns. The battle for our climate, for the continuation of our species and all other life, will be largely waged in cities. Will it be possible to create the ideal ecological town? The objective of this study is to compare actual and theoretical urban forms in order to explore the diverse effects of daylighting and solar potential on densely built-up sites. Indeed, the search for ideal solar orientations has been a persistent problem from the 19th century until the beginning of the 20th. Even today, for all our scientific knowledge and empirical experience of solar energy, making the correct choice has still not crossed the boundary from an enigmatic, tricky art into the field of an exact science. As a first step, six existing urban sites in Switzerland and Brazil, as well as two projects in the United Kingdom, a few generic models and a utopian city, were modelled and analyzed. The goal was not merely the construction of new, exemplary districts but also the transformation of existing towns. The findings have been used to assess the potential of façades and roofs located in urban areas for active and passive solar heating, photovoltaic electricity generation and daylighting. The results obtained from these different case studies revealed large variations of the potential for solar energy collection on the buildings' façades and roofs. Ideally, these investigations are expected to help architects and town planners understand how to optimise solar collection on buildings design. In addition, a study was made of the Contemporary City of Three Million Inhabitants created by Le Corbusier. According to the latter, the Contemporary City could increase the urban capacity and at the same time improve the urban environment and the efficiency of the city. Therefore the present study extracted the solar potential of this utopian city in order to assess Le Corbusier's propositions. Since 1922, Le Corbusier's studies have influenced architects and town-planners in their choice of orientations and urban forms: it was therefore crucial to discover if this reliance on Le Corbusier was justified. Solar irradiation and illuminance values obtained through numerical simulations form the core part of the method. The method adopted for the entire set of computer simulations consisted in modelling — via a computer-aided drawing piece of software, namely AutoCAD — the digital model of each building or group of buildings prior to exporting them to an urban performance analysis software (PPF/RADIANCE software package). Spatial distributions of solar irradiation and daylight fluxes over the overall building façades and roofs were calculated using ray-tracing simulation techniques to determine the appropriate placement of different solar technologies (passive and active solar, photovoltaic and daylighting). Performance indicators were used to assess the solar utilisation potential of the urban sites (e.g. statistics of sky view factors and daylight factors) in order to determine the optimal solar strategies for a given urban context. In addition, various works and sun-related recommendations that have appeared mainly since the end of the nineteenth century, together with the various scientific models and software tools pertaining to this topic, are presented in this study.

EPFL2010

Chargement

EPFL2010