{"id":10682,"date":"2017-03-06T16:57:20","date_gmt":"2017-03-06T15:57:20","guid":{"rendered":"https:\/\/www.etalab.gouv.fr\/?p=10682"},"modified":"2019-06-18T10:40:11","modified_gmt":"2019-06-18T08:40:11","slug":"building-an-open-solar-power-map","status":"publish","type":"post","link":"https:\/\/preprod.etalab.gouv.fr\/building-an-open-solar-power-map","title":{"rendered":"Building an open solar power map"},"content":{"rendered":"\n

espite the introduction of financial incentives for developing \nproduction of photovoltaic (solar) power systems since 2000, France \nranks only 15th out of 28 in Europe for photovoltaic production per \ninhabitant.<\/p>\n\n\n\n

As a comparison, Germany sets an example with production per \ninhabitant five times higher. Concerns about the selectiveness of the \nsubsidies, and the increasing burden on finances, led to a reduction in \nthe incentives after 2010.<\/p>\n\n\n\n

While Germany and other countries developed solar cadastres (public \nregisters of property) to assess the potential of candidate roofs for \nsolar panel installations, such initiatives are still limited to a few \ncities in France, Brest and Paris being the most successful examples. \nThese cadastres often use a three-dimensional model of a city, requiring\n expensive data collection and treatments, and, consequently, are used \nmostly for highly populated areas.<\/p>\n\n\n\n

An open solar cadastre, assessing the potential of roofs for solar \npanels covering the whole territory, would not only benefit public \nauthorities but also a whole community comprising energy providers, \npanel installers, consulting companies and homeowners.<\/p>\n\n\n\n

The Etalab team used an innovative, cost-efficient approach combining\n open data and open algorithms, relying on external contributions to \nbuild a nationwide solar cadastre.<\/p>\n\n\n\n

The French land cadastre provides the contours of every structure. \nThe shape of the roof is still uncertain and visual analysis is required\n to distinguish a ridge going west to east (suitable for solar \ninstallations) from one going north to south. So, satellite and aerial \nimages covering the whole French territory with sufficient precision for\n most situations are used.<\/p>\n\n\n\n

Etalab took advantage of a hackathon to design and set up a \ncrowdsourcing platform with the help of enthusiastic developers. The \nplatform displays the image of a roof and the user is invited to provide\n its orientation. The platform, being fun and somewhat addictive, \nreceived 100,000 contributions in a three-week span by word of mouth. \nThis allowed us to classify 10,000 roofs with confidence. We identified \njust one case of vandalism, which was easily spotted and discarded.<\/p>\n\n\n\n

\"\"
opensolarmap.org : Crowdsourcing platform, displaying a roof image and 4 possible orientation choices <\/figcaption><\/figure>\n\n\n\n

<\/p>\n\n\n\n

This is a small sample compared to the 50 million buildings in \nFrance, but it is enough to programme an automated classifier. Using \nstandard techniques in image processing, namely logistic regression and \ndeep neural networks, we obtained a classifier that was correct 80% of \nthe time. Later, we found the automated results to be comparable \nto\u00e9human contributions in accuracy. Run on standard hardware, the \nclassifier takes one second to make a decision on an image.<\/p>\n\n\n\n

This classification challenge was later taken on by a hundred teams \nduring the Data Science Game, an international data science competition.\n The winning team, using newer neural networks and advanced techniques \nlike data augmentation, fine tuning and ensemble learning, achieved a \n30% lower error rate.<\/p>\n\n\n\n

Solution and action<\/strong><\/h2>\n\n\n\n

The nationwide map of roof orientation shows differences between \nregions affected unequally by wind and topography. For example, most \nroofs in Brittany have a favourable orientation, whereas the opposite is\n the case in the \u00e9Rh\u00e8ne Valley. This map is of great importance to \nassess the relevance of solar incentives at a local level. It is worth \ncomparing with the solar exposure to evaluate solar potential.<\/p>\n\n\n\n

\"\"<\/a>
Roof orientation map. Regions with a high number of favourably orientated roofs are displayed in red. <\/figcaption><\/figure>\n\n\n\n

The solar cadastre doesn\u2019t take into account shades or the angle of \ninclination of roofs, and it discards roofs with complex shapes. \nHowever, it is intended to be completed by more precise, possibly local \nand expensive, data to deliver a better result. All the data<\/a>, as well as the code<\/a>\u00e9used\n and produced in this project, is open and documented. Therefore, it is \neasy for \u00e9anyone working in this field, whether from the public or \nprivate sector, to quickly build an improved solar cadastre on top of \nours or to replicate it in another country.<\/p>\n\n\n\n

\"\"<\/a><\/figure>\n\n\n\n

\n Horizontal irradiation \u2013 France. Source: PVGIS \u00e9 European Union, 2001-2012 (Copyright: 2011 GeoModel Solar s.r.o.)\n <\/p>\n\n\n\n

Learnings<\/h2>\n\n\n\n

As a government service attached to the Prime Minister, Etalab \nprefers inclusive approaches to closed ones. Machine learning projects \nlike this one seem particularly suited to public participation. Every \ncitizen, regardless of their skills and expertise, can offer their help \nusing a crowdsourcing platform.<\/p>\n\n\n\n

Calling on public participation is also a natural way to communicate and advertise the goals of the project.<\/p>\n\n\n\n

The Etalab team benefited from voluntary contributions during the \ndevelopment of the crowdsourcing platform, the construction of the \ntraining set and the building of automated classifiers. Additionally, we\n rooted our project in open data and open source tools. In this way, a \nteam of two people managed to develop this project within a few months, \nfor negligible hardware costs. Although projects using machine learning \nare often termed \u00e9?Big Data\u2019, it would be a misnomer in this case, since\n we systematically favoured small-scale, quick and cost-effective \nmethods, involving manageable amounts of data.<\/p>\n\n\n\n

This approach lays the foundation for tools facilitating the work of \npublic decision-makers involved in energy policies. It could be easily \nreplicated for similar issues; for example, detection of bus lanes and \npedestrian crossings, for land-use classification, and so on.<\/p>\n\n\n\n

We demonstrated the application of state-of-the-art, free, \nopen-source, well-packaged machine learning solutions outside of a \nresearch context. Engineers and developers can now extract value from \nimages without having to be specialists in image processing or deep \nlearning. It is probable that such tools will become increasingly \nwidespread and eventually find their way into the general IT engineer\u2019s \ntoolbox.<\/p>\n\n\n\n

This blogpost was originally published on the British Civil Service Quarterly.<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"

espite the introduction of financial incentives for developing production of photovoltaic (solar) power systems since 2000, France ranks only 15th out of 28 in Europe for photovoltaic production per inhabitant. As a comparison, Germany sets an example with production per inhabitant five times higher. Concerns about the selectiveness of the subsidies, and the increasing burden …<\/p>\n

Building an open solar power map<\/span> Lire la suite\u00a0\u00bb<\/a><\/p>\n","protected":false},"author":27,"featured_media":10683,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_uag_custom_page_level_css":"","site-sidebar-layout":"default","site-content-layout":"default","ast-global-header-display":"","ast-banner-title-visibility":"","ast-main-header-display":"","ast-hfb-above-header-display":"","ast-hfb-below-header-display":"","ast-hfb-mobile-header-display":"","site-post-title":"","ast-breadcrumbs-content":"","ast-featured-img":"","footer-sml-layout":"","theme-transparent-header-meta":"default","adv-header-id-meta":"","stick-header-meta":"","header-above-stick-meta":"","header-main-stick-meta":"","header-below-stick-meta":""},"categories":[271],"tags":[],"uagb_featured_image_src":{"full":["https:\/\/preprod.etalab.gouv.fr\/wp-content\/uploads\/2019\/06\/image-agd11.png",1000,577,false],"thumbnail":["https:\/\/preprod.etalab.gouv.fr\/wp-content\/uploads\/2019\/06\/image-agd11-150x150.png",150,150,true],"medium":["https:\/\/preprod.etalab.gouv.fr\/wp-content\/uploads\/2019\/06\/image-agd11-300x173.png",300,173,true],"medium_large":["https:\/\/preprod.etalab.gouv.fr\/wp-content\/uploads\/2019\/06\/image-agd11-768x443.png",768,443,true],"large":["https:\/\/preprod.etalab.gouv.fr\/wp-content\/uploads\/2019\/06\/image-agd11.png",1000,577,false],"1536x1536":["https:\/\/preprod.etalab.gouv.fr\/wp-content\/uploads\/2019\/06\/image-agd11.png",1000,577,false],"2048x2048":["https:\/\/preprod.etalab.gouv.fr\/wp-content\/uploads\/2019\/06\/image-agd11.png",1000,577,false],"rpwe-thumbnail":["https:\/\/preprod.etalab.gouv.fr\/wp-content\/uploads\/2019\/06\/image-agd11-45x45.png",45,45,true]},"uagb_author_info":{"display_name":"Michel Blancard","author_link":"https:\/\/preprod.etalab.gouv.fr\/author\/michel"},"uagb_comment_info":0,"uagb_excerpt":"espite the introduction of financial incentives for developing production of photovoltaic (solar) power systems since 2000, France ranks only 15th out of 28 in Europe for photovoltaic production per inhabitant. As a comparison, Germany sets an example with production per inhabitant five times higher. Concerns about the selectiveness of the subsidies, and the increasing burden\u2026","_links":{"self":[{"href":"https:\/\/preprod.etalab.gouv.fr\/wp-json\/wp\/v2\/posts\/10682"}],"collection":[{"href":"https:\/\/preprod.etalab.gouv.fr\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/preprod.etalab.gouv.fr\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/preprod.etalab.gouv.fr\/wp-json\/wp\/v2\/users\/27"}],"replies":[{"embeddable":true,"href":"https:\/\/preprod.etalab.gouv.fr\/wp-json\/wp\/v2\/comments?post=10682"}],"version-history":[{"count":1,"href":"https:\/\/preprod.etalab.gouv.fr\/wp-json\/wp\/v2\/posts\/10682\/revisions"}],"predecessor-version":[{"id":10684,"href":"https:\/\/preprod.etalab.gouv.fr\/wp-json\/wp\/v2\/posts\/10682\/revisions\/10684"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/preprod.etalab.gouv.fr\/wp-json\/wp\/v2\/media\/10683"}],"wp:attachment":[{"href":"https:\/\/preprod.etalab.gouv.fr\/wp-json\/wp\/v2\/media?parent=10682"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/preprod.etalab.gouv.fr\/wp-json\/wp\/v2\/categories?post=10682"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/preprod.etalab.gouv.fr\/wp-json\/wp\/v2\/tags?post=10682"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}