Hi, I'm the Data to Policy Chatbot
BETA: This site is work in progress
INFO
CLOSE
BETA: You are looking at the early access version of the Data to Policy Navigator – Let us know your feedbackto help us improve.
Want to hear from us? Subscribe here
Country
PortugalRegion
EuropeProJECT YEAR
TYPe OF DATA
AI Capability
Computer vision
STAKEHOLDERs
SECTOR
DATA SOURCE
Local governments all over the world are formally signing the Global Covenant of Mayors for Climate and Energy, the world's largest movement for local climate and energy action. Signatories to the Covenant must commit to adopting an integrated approach to climate change mitigation and adaptation. Most are aiming for carbon neutrality by 2050, in alignment with the goals of the Paris Agreement. Within the first two years of signing up to the Covenant, signatories are required to develop a Sustainable Energy and Climate Action Plan (SECAP). The aim of a SECAP is to cut CO2 emissions by at least 40 per cent by 2030 and increase resilience to climate change.
Lisbon is a signatory of the Covenant and has developed a SECAP which is being used to drive climate action in the city. Lisbon is also taking part in 'AI4PublicPolicy', an initiative from the European Union that harnesses AI for citizen-centred policymaking. Lisbon is exploring how AI can be applied across a range of uses, from energy management policies to data analysis, and tracking progress on its Climate Action Plan (CAP) 2030. Lisbon’s approach is to accelerate policymaking by integrating AI-generated outputs and information from local stakeholders. The city is piloting a test that can provide information about the status of photovoltaic (PV) installations in the city. This is an important focus given Lisbon’s commitment to expanding solar energy as a strategy for reducing CO2 emissions. An AI model was used to precisely identify and geolocate PV installations in Lisbon which enabled policymakers to make informed decisions about PV deployment in the city. This data can also be leveraged to roll out energy-related campaigns to deprived or energy-poor districts.
The workflow proposed by the city of Lisbon will enhance municipal policymaking for its CAP 2030 goals. Manually inspecting all the satellite images of the PV installations would take months to complete and would quickly become outdated. The automated pilot will help city managers save time, enable Lisbon to make headway on its sustainability goals, and speed up the mapping and assessment process of the city’s PV installations.
In its CAP 2030 plan, developed under C40 Cities (a global network of cities committed to reducing climate change), Lisbon has set an ambitious goal to cut its greenhouse gas emissions (GHG) by 70 per cent in 2030 relative to 2002. According to historical GHG data, almost half of the city’s emissions come from the residential and services sector—its building stock. Measures have been put in place to increase the energy efficiency of buildings and enable on-site energy generation using PV technology. This supports Lisbon’s goal to transform the city’s building stock into zero-emission buildings by 2050 and strengthens its bid to be among the world’s 100 climate-neutral cities by 2030.
Despite the huge potential for producing solar energy in Lisbon, current capacity is still very low. In 2021, according to data from the Portuguese Directorate-General for Energy and Geology (DGEG), the city only had 551 photovoltaic systems producing 8 megawatts (MW) of energy. This corresponds to just 0.4 percent of the city’s annual electricity consumption. To accelerate capacity and unlock Lisbon’s solar potential, the city’s CAP 2030 project set a goal of achieving 103 MW of photovoltaic capacity. The city subsequently rolled out its Lisboa Cidade Solar® (Lisbon Solar City) strategy, encouraging the uptake of PV installations by providing local citizens and businesses with information and technical support. To gain an accurate picture of PV deployment, the city used geolocation tools to pinpoint existing PV assets and evaluate how PV technology was being deployed, citywide.
The Lisbon pilot revolved around a main use case and its corresponding pilot policy about mapping PV systems. A test policy used datasets showing the administrative boundaries of Lisbon and its 24 parishes together with a set of satellite maps of the city. These were made accessible via an open data platform. The test indicators for the AI model were based on the number of PV installations across the whole city and the installed PV capacity at each location. To identify the PV installations from aerial images, the pilot used an AI-powered computer vision model called 'Yolov8', a high-performing open-source image detection tool developed by AI solutions company, Ultralytics. After experimenting with different combinations of datasets, the best option combined crowdsourced data with existing Lisbon city images.
To evaluate the images, the pilot team used a graph comparing the number of PV panels detected to the actual PV panels that were present.
The second step in the pilot used an image segmentation model to identify the surface area of the PV installations in the Yolov8 images. The model was trained to detect the surface area of the PV panels and to separate these from the background cityscape. The model calculated the percentage of each image that showed a PV installation. Results from the models were shared on the AI4PublicPolicy virtual policy management environment, a cloud platform. The Lisbon pilot case dashboards displayed all the PV installations in Lisbon, broken down by parish, with the top parishes ranked. Users could see the location of the PV installations, with overall distribution shown at lower zoom levels and exact locations at higher zoom levels.
The methods deployed within the Lisbon use case helped the city to advance on two goals: accelerating technical planning for PV installations and conducting a socioeconomic impact assessment of PV needs. The project also generated funding insights that could be applied across the city. Decision-making was made easier thanks to more detailed knowledge about the geographical distribution of PV installations across Lisbon and its parishes. This will enable the city to carry out targeted campaigns promoting PV uptake and support the deployment of PV in energy-poor neighbourhoods with good solar exposure.
The AI model validated that the PV systems were correctly detected, and the validated set of images will be used to update an existing but incomplete public map available on SOLIS, Lisbon’s solar panel mapping platform. The aim is to use AI to generate continuous updates of this map and build on the lessons learnt in the pilot. As this knowledge becomes more widely available, businesses and citizens may become better informed and more interested in adopting PV technology. This could kickstart citizen-led initiatives like Lisbon’s various 'community solar' projects and will help deliver a just energy transition for all across the city.
