Making use of already-built space through rooftop solar photovoltaic systems is one of the most effective strategies for accelerating the energy transition in consolidated urban environments. However, traditional planning models tend to systematically overestimate real renewable potential by relying on simple aggregations of available surface area or averaged production factors.
To overcome this methodological limitation, a research team from the Urban Energy Transition Chair at UPV (Universitat Politècnica de València) has developed a pioneering, fully automated methodology. The study integrates Artificial Intelligence (AI)-based computer vision algorithms, three-dimensional geometric modelling derived from LiDAR point clouds, and radiometric simulation to identify and optimise, at large scale, the placement of solar panels across the urban building stock.
📘 «Towards realistic urban rooftop solar photovoltaic large-scale planning: integrating artificial intelligence-based roof detection, 3D geometry, and energy modelling»
👥 Authors: Tomás Pino-Gallardo, Paula Bastida-Molina, Tomás Gómez-Navarro, Carla Montagud-Montalvá
📍 Journal: Engineering Science and Technology, an International Journal (JESTECH)
From 18.4 to 2.67 million square metres: the need for rigorous technical filtering
The methodology has been validated on virtually the entire municipality of Valencia, analysing a total built rooftop area of more than 18.4 million m².
Although the initial screening using the Segment Anything Model (SAM) AI model —previously fine-tuned and trained on orthophotographs of the Valencian region— identified 8.93 million m² free of permanent obstacles (such as antennas or air-conditioning units), applying construction and energy constraints narrowed the technically viable surface down to 2,670,079 m² (14.6% of the total urban area, or 32% of the obstacle-free surface).
Total rooftop surface
18.4 M m²
18.4 M m²
↓
Obstacle-free surface (AI)
8.93 M m²
8.93 M m²
↓
Technically viable surface
2.67 M m² (14.6% of total)
2.67 M m² (14.6% of total)
This automated technical screening excluded areas based on rigorous criteria:
- Radiometric filter: exclusion of roofs with solar irradiation below 1,000 kWh/m²/year.
- Geometric and structural filter: exclusion of slopes greater than 40°, unfavourable north-facing orientations, buildings registered as derelict in the land registry, and surfaces smaller than 10 m².
- Safety and maintenance: a mandatory 0.60-metre perimeter clearance around each installation.
270 GWh a year: the decisive role of flat roofs
Distributed processing through the PVLayout module made it possible to calculate the optimal panel design and layout to maximise annual electricity generation. The city-wide results for Valencia confirm a large-scale deployment capacity:
| Installable modules | 342,785 panels |
| Total installable power | 173 MW |
| Estimated generation | 270 GWh/year |
| Share of the residential sector | 69% of total potential |
| Modules on flat roofs | 78% of all panels |
The study reveals a critical technical finding for compact Mediterranean-type cities: 78% of the optimal modules are located on flat roofs. The optimisation algorithm showed that, in urban practice, favouring layouts with a higher number of panels —even adjusting tilt angles to suboptimal values to avoid row-to-row shading— generates more overall energy than trying to force the theoretically ideal 35° tilt on limited surfaces.
District and sector disparity: a geospatial snapshot of the city
The distribution of solar potential is not uniform across the urban fabric; it varies markedly depending on building typology and the age of construction:
- Residential dominance: the residential sector accounts for 69% of the city’s total potential, followed by public facilities and service buildings with 15% and the commercial sector with 6%.
- Best-performing neighbourhoods: Benimàmet (8.39 MW − 13.40 GWh/year), Benicalap (8.02 MW − 12.61 GWh/year) and La Carrasca (6.57 MW − 10.53 GWh/year) lead generation capacity due to their high proportion of flat roofs and modern buildings.
- Historic centre limitations: at the opposite end, historic neighbourhoods with dense layouts and heterogeneous sloped roofs, such as El Mercat (0.38 MW) or La Seu, show significantly lower suitability rates due to cross-shading and complex roof geometries.
A decision-support tool for the 2030 Climate Mission
In contrast to theoretical models disconnected from the physical reality of buildings, this Chair-led methodology provides an open-source operational platform based on Python and GIS.
For public officials working on a Sustainable Valencia, urban planners and citizens alike, this tool makes it possible to base local climate policies on real data intelligence. It allows identifying which districts to prioritise for the creation of Renewable Energy Communities (RECs), accurately estimating the impact of distributed generation on the local grid, and advancing with scientific rigour towards Valencia’s 2030 climate neutrality commitments.
