3D model discretization in assessing urban solar potential: the effect of grid spacing on predicted solar irradiation. (December 2018)

• Record Type: Journal Article
• Authors: Peronato, Giuseppe; Rey, Emmanuel; Andersen, Marilyne
• Is Part Of: Solar energy. Volume 176(2018)
• Page Start: 334
• Page End: 349
• Publication Date: 2018-12
• Subjects: Solar potential -- 3D city model -- Mesh generation -- Uncertainty analysis; Solar energy -- Periodicals; Solar engines -- Periodicals; 621.47
• DOI: 10.1016/j.solener.2018.10.011
• Languages: English
• ISSNs: 0038-092X

Abstract

Highlights: We assess sensor grid resolutions and algorithms in urban solar irradiation studies. Multiple sensor spacing intervals (1–4 m) are tested against a 0.5-m resolution. For structured grids, the RMSE linearly increases with the spacing interval. Maximum relative RMSE was found limited to 7%. Highly irradiated surfaces are less affected by the resolution. Abstract: The increasing interest in solar energy production in urban areas requires an accurate simulation of solar irradiation on building surfaces, including vertical surfaces. However, solar potential analyses are usually conducted on 2.5D models, which are limited to roof surfaces. Methods based on 3D models, instead, allow the simulation of solar radiation on all building surfaces also accounting for inter-reflections. 3D models are thus discretized by grids of sensor points on which the solar potential is calculated. This paper investigates the discretization error in the assessment of solar potential based on 3D models. To this end, we tested the sensitivity of simulated solar irradiation to the resolution of the sensor grid. We analyzed the impact of the grid resolution using typical discretization approaches affecting the spatial arrangement of the sensor points. The study was conducted in a dense area of the city of Geneva represented at level of detail (LOD) 2. The simulated solar irradiation on 109 buildings was analyzed at different spatial, i.e. per surface and per building, and temporal granularities, i.e. hourly and yearly. The results show that the error increases linearly for grids spaced at up to 4 m with maximum relative root mean square error lower than 7%. The impact of the grid resolution was found greater for structured grids than unstructured grids. The results also highlight that finer grid resolutions (i.e. smaller spacing) are necessary if the analysis is conducted at high spatial or temporal granularity, notably when analyzing roof surfaces with shading artifacts.