/

August 13, 2020

Solar panels in Switzerland brought to the next level with Picterra

solar panels in Switzerland

In order to achieve the objectives of the Swiss Confederation’s 2050 strategy, it is necessary to drastically increase the number of photovoltaic installations deployed annually in Switzerland. However, relatively few individuals install solar panels in Switzerland for several reasons, such as lack of technical knowledge, fear of bad choices, lack of time. Based on these observations, the “GROUP-IT” project was born, with the aim of encouraging Swiss people to get involved in the energy transition. Thus, the Energy Management Lab of the HES-SO Valais-Wallis, offers property owners support in installing photovoltaic panels on their roof(s) – individual houses, residences, and small businesses.

For this purpose, the GROUP-IT team proposes an approach in 4 main steps to increase the number of solar panels in Switzerland:

  1. Approaching the municipalities, or other territorial entities, in order to propose them a GROUP-IT action. The administered owners can then register online.
  2. Providing the registered owners with an energy potential assessment of their building (estimate of producible kWh, investment, profitability, etc.) in the form of a personalized document.
  3. Organising a visit of the registered buildings in order to collect the necessary information permitting companies to submit bids.
  4. Launching a grouped call for tenders that allows each owner to receive the two best bids from national companies.

However, this document deals exclusively with the work carried out in the second step of this process. Indeed, it is at this stage that the majority of the technical work is carried out and that all the specific skills of each partner come into play.

Solar panels in Switzerland: a new marketplace for installers and individuals

Generally speaking, in order to compute an accurate estimate of the solar potential of a building, it is necessary to get some information about its roof, including the following data:

  • the topography of the building zone (elevation, horizon profile),
  • the geometry of the roof (area, slope, orientation),
  • the surface that is available on it.

This last point is used to calculate the number of solar panels that can be installed on the roof. To get the best estimate of the number of solar panels, a lot of objects that can prevent the installation of a panel such as chimneys, antennas, windows or even other panels (like thermic ones) have to be taken into account.

Admittedly, the Swiss Federal Office of Energy (based on the 3D models of the buildings generated by the Swiss Federal Office of Topography) provide an estimate of the solar potential of almost every Swiss roof. However, they take into account the entire area of the roofs for their calculation, even if only a small surface can be equipped with solar panels. If the roof contains many objects, it is the owner himself who must manually modify the usable surface by applying a ratio (50% or 75% of the available surface). The roofs’ geometry and their solar potential are available as an open web geoservice and in an online portal. In this project, these data are used as a basis for a more accurate estimate.

To reach this level of accuracy, the GROUP-IT assessment of the solar potential has four components: a web platform to interact with the citizens and gather social data, an AI detection of the roof objects preventing the installation of solar panels, a GIS database and workflow for the computation of the solar panels and finally an algorithm to estimate the solar potential based on the roofs and topography characteristics.

In this paper, we will detail the interdisciplinary GIS and AI workflow which lead to the computation of the accurate solar potential estimation. The GIS workflow is delivered as a set of QGIS projects and QGIS plugin functions. They provide a simple and automated workflow for GROUP-IT administrators who can, at the end of the process, easily process the resulting data in their algorithm.

Solution & technical aspects

The GIS workflow was developed in QGIS: an Open Source GIS which comes with a Python programming interface to develop plugins. The data are stored in a PostgreSQL database enabled with the GIS extension PostGIS. QGIS is used to visualize and edit the geographic layers stored in the database. A plugin provides the set of functionalities required by the GROUP-IT administrators to import data in the database, compute values, and export the results.

Step 1: Get information from the citizen

This first step is ensured by the GROUP-IT web platform, through which citizens can register the building they want to equip. For this purpose, the property owner has to answer several simple questions: annual electric consumption, heating, and domestic hot water systems, number of residents, average occupation, the existence of electric vehicles, fiscal information (optional). Questions are obviously adapted relating to the type of building that is registered (private house, building with rental apartments, company premises). In any case, the property owner is asked to click on a web map to indicate precisely the building on which he wants to install solar panels.

Step 2: Download geographic data from web services

The points clicked by the citizen are used to download the geometry of the roofs of the corresponding buildings. They are downloaded from the web service provided by the Swiss Federal Office for Energy (OFEN). Each roof collected from the web service has some attributes describing its solar potential. In addition, the horizon and elevation of the buildings are retrieved from a web service provided by the EU.

Step 3: Check the data

The GROUP-IT administrators have to check all registered roofs. Indeed, a successful building registration depends on the quality of three data sources: data given by the user through the form, data given by the OFEN solar database, the recency of the orthoimages. Regarding this amount of data, several scenarios can occur:

  1. Multiple points on the same area: the owner can indicate two or three roofs in the same area (for instance, he may also have a barn or garage attached to his house). In this case, GROUP-IT administrators may have to group all these points and attach them to one installation ID (if the owner wants to use different roofs to power his house).
  2. False roof area: certain roof surfaces could have been automatically attached to another near building, that is not owned by the same user. Equally, roof sections could be missing on the registered roof. Administrators can delete the supernumerary roof section or add the missing sections.
  3. Missing the entire building:
  • The user did not click precisely on the building.
  • The administrators have to correct the click position and relaunch the download.
  • The building is not in the OFEN solar database (i.e. it was built recently), but is represented on the orthoimage. Administrators have to create the print of the roof, by drawing a form around the shape of the building illustrated on the orthoimage. Then, information related to the roof (slope and orientation) has to be added.

  • The building is not on the OFEN solar database, nor shown on the orthoimage. Administrators have to create the print of the building by drawing shapes, based on the architectural plans that are sent by the property owner. Then, information related to the roof (slope and orientation) has to be added too.

Step 4: Alignment of roofs and orthoimage

After the control step, the GROUP-IT administrators check that the vectorial roofs are correctly aligned with the orthoimage (Swiss Federal Office of Topography, swissimage at 10 or 25cm).

Indeed, the roofs in the orthoimage might be shifted. This shift is due to the use of a Digital Terrain Model (containing only the level of the ground) and not a Digital Surface Model (containing the building’s height) for its generation. If administrators see a shift, they have to draw a translation vector and the vectorial roof is automatically moved.

Step 5: Solar panels generation

Once all roof sections are correctly aligned, they are automatically filled with solar panels. These solar panels have a standard dimension (1m width and 1.60 m length) and are oriented perpendicularly or in parallel to the roof slope, according to an optimization allowing the installation of a maximum number of panels. In case of an orientation difference between the 3D building data and the OFEN data, the GROUP-IT administrators can correct the orientation thanks to the “roof orientation correction” routine.

Step 6: Detections of objects on the roofs

The roofs are merged to define some Area Of Interest. The AOI is sent to Picterra’s server via its API. A detector of roofs objects (chimneys, antennas, roof windows, existing solar panels, eaves, etc.) has been previously trained using the Picterra user interface (read this article on how to build a custom detector). The orthoimage used for detection are served via the Swisstopo WMS service and directly connected to the Picterra platform. The roof object detector is run over the AOI and the detected objects are sent back to QGIS and saved in the database.

Thanks to this process, an average of 70% of the objects are automatically detected, which already saves the GROUP-IT team a considerable amount of time. Moreover, this ratio will undoubtedly evolve over the course of the various projects that will be progressively created.

Step 7: Detections check

GROUP-IT administrators then review the detections, remove false positive or add new detections if they are missing. These manipulations are possible through the QGis interface, by drawing shapes on the roof.

Step 8: Intersection of solar panels with detections

If a solar panel generated in the step 6 touches a detected object. The corresponding solar panel is automatically removed from the roof.

Step 9: Roof image generation

An image of the roof showing the solar potential of each roof section, the detections and the maximum number of panels that can be installed taking in account the obstacles detected on the roof, is generated to be put on the pre-evaluation report.

Outlook and the future

This workflow was applied to two initial projects in Switzerland. Some limitations have appeared. The first unforeseen problem was the shift between the buildings in the orthoimage and the vector buildings, we solved it by allowing the administrator to drag the vector building on the orthoimage. In the future, we might develop a similar function to align automatically the vector data with the image. Second, many buildings have a flat roof. In this scenario, the solar panels have to be set either oriented to the south with a provided slope and distance between the rows to avoid shadowing or they have to be oriented West-East in a pyramidal geometry.

Finally, another limitation of increasing the number of solar panels in Switzerland is the heavy reliance on the geographic datasets of the Swiss Confederation. Indeed the update of the 3D buildings and estimation of the solar potential can take several years. Hence, for new buildings in Switzerland or if our approach would be extended to another country, it could be necessary to have a method to generate the 3D geometries of the roofs. Technically, the 3D geometry can be obtained automatically or semi-automatically by LIDAR or stereophotogrammetry acquisition. Depending on the area which needs to be covered, the choice of the aircraft is open: drone, helicopter, or airplane.

Authors:

Timothee Produit, IG group SA

Jean-Marie Laurent, HES-SO Valais-Wallis

Lucien Papilloud, HES-SO Valais-Wallis

Want to learn more?

sign up to our newsletter