An increased use of renewable energy and of energy efficiency measures in buildings is needed to face the urgency of climate change. Buildings are in fact among the highest worldwide consumers of primary energy, mostly of fossil fuel origin, while still making insufficient use of in-situ renewable energy sources. To find a solution to this situation, many municipalities have promoted the use of solar cadastres mapping the solar energy potential of the existing building stock. However, their implementation has limits from different points of view including assessment accuracy, representation methods, and decision-support. To overcome these limits, this thesis proposes a planning-support system based on the photovoltaic (PV) potential of buildings. The goal is to provide decision-makers and stakeholders with a robust method to assess the potential of photovoltaic electricity generation of existing buildings under uncertain environmental conditions. The developed methodology is based on an urban-scale modeling workflow that includes the simulation of the photovoltaic electricity production and a simplified estimation of the building energy retrofit potential. Existing state-of-the-art models for solar radiation, building energy and PV performance are coupled in the workflow, which relies on a vector 3D city model featuring an accurate representation of buildings, terrain, and vegetation. The proposed modeling workflow also includes an innovative approach for simulating the arrangement of PV modules on the building envelope, which influences both the energy yield and the acceptability of the system. The modeling workflow is in turn integrated into a planning-support system that provides a robust assessment of the photovoltaic potential through risk-averse scenarios. We consider here two crucial yet underestimated uncertainty factors: weather and vegetation. The results are aggregated at different scales and, for each scale, the spatial locations are ranked through pairwise comparisons according to relevant energy indicators. The results are finally displayed in a 3D-mapping tool featuring false-color overlays at the considered aggregation scales to address different objectives and inform decision-makers. We conducted sensitivity analyses towards different input data resolutions and modeling scenarios so as to achieve a good trade-off between accuracy and computational cost and define confidence intervals for the calculated values. The simulated PV yield was also compared against measured data from an existing PV installation. The proposed modeling workflow and planning-support system were tested in an urban district within the city of Neuchâtel (Switzerland). The analysis highlighted areas with the highest potential and provided a priority list of interventions. It also showed the impact of vegetation on absolute results and especially on the ranking of the spatial locations evaluated by their energy potential.