An international research team has introduced a new method capable of determining the tilt and azimuth of rooftop photovoltaic (PV) arrays by analyzing operational power data. Accurate orientation information is critical for estimating electricity generation and detecting system faults, yet installation records are frequently missing or incorrect, while physical site inspections remain impractical for large PV portfolios.
Corresponding author Hanrong Huang explained that the study offers a practical solution to recover panel orientation without requiring a site visit, even when accounting for partial shading and geometric limits at low tilt angles. The team validated the approach using operational data collected in 2025 from 17 commercial rooftop systems in Australia, where orientations were independently verified. The dataset consisted of DC power measurements recorded at five-minute intervals for each maximum power point tracker (MPPT).
To estimate solar irradiance on panels at various angles, the researchers utilized pvlib clear-sky modeling alongside ERA5 weather data. By selecting periods with high direct irradiance and sufficient morning and afternoon measurements, an optimization algorithm identified the orientation that best matched the measured power output while adjusting for effective system capacity to account for power losses.
The study compared two fitting metrics, R² and Huber loss, assessing how factors such as shading, low tilt angles, data resolution, module temperature, and inverter clipping influenced accuracy. While R² measures the alignment between modeled and observed power, Huber loss minimizes the impact of outliers caused by shading. On unshaded systems, the method recovered tilt within a median error of about 2° and azimuth within 4° using only DC power data and site coordinates. This approach also allows for determining the orientation of individual arrays at the MPPT level.
Results indicated that clear-sky modeling consistently outperformed ERA5 data; relying on ERA5 led to significantly higher median errors of 10° to 12° for tilt and 48° to 62° for azimuth in unshaded systems. Shading was found to increase azimuth errors, raising them from 4° to 35° when using Huber loss and from 15° to 27° with R². Additionally, determining azimuth proved difficult at tilt angles below 5° because panel orientation has minimal effect on incident irradiance, though this limitation has a reduced impact on estimated electricity generation.
Huang noted that the primary next step is reducing the effect of partial shading on recovery accuracy, which remains the main limitation of the method. The team also aims to make the technique robust against other real-world operating conditions, such as inverter clipping and tripping, to ensure reliable application at a large scale. The findings were published in Solar Energy under the title "Recovering rooftop photovoltaic array tilt and azimuth from operational power data." The research collaboration included scientists from the Chinese Academy of Sciences, the University of New South Wales, Diagno Energy, and Global Sustainable Energy Solutions.
Sources: pv magazine.