Comparing Approaches for Estimating Residential PV System Orientations

Accurate knowledge of photovoltaic (PV) system orientations—tilt and azimuth—is crucial for forecasting generation, assessing self-consumption, and managing distributed solar power. Yet, orientation data are often missing or unreliable in PV databases. This study compares three approaches for determining the orientation of residential PV systems in Sweden: (i) self-reported values from the national capital subsidy program, (ii) statistical orientation distributions based on PV penetration, and (iii) a LiDAR-based method using aerial imagery and linear regression. Validation was performed against a manually derived ground truth dataset of 104 PV polygons across three municipalities. Self-reported orientations proved highly inaccurate, primarily because the application form allowed only a single entry for a system’s azimuth and tilt and lacked clear guidance on azimuth conventions, compounded by imprecise or careless estimations from applicants. In contrast, the LiDAR-based method closely matched the ground truth, with R² values of 0.96 for azimuth and 0.82 for tilt and captured the diversity of real-world orientations more accurately than the statistical model. These results demonstrate that LiDAR-based remote sensing provides a reliable, scalable, and high-accuracy approach for determining PV system orientations, supporting improved generation forecasts, grid operation planning, and the integration of distributed solar power at high penetration levels.