Rooftop Solar Orientation Inferred From Power Data

The research-stage method could help operators audit large solar fleets remotely, though accuracy drops sharply under partial shading.

Updated on Nov. 5, 2026 in Energy

Vector editorial illustration of tilted solar panels on a flat rooftop with a mustard compass wedge pointing across the array, evoking remote orientation analysis.
Researchers in Australia tested a method that infers rooftop solar panel orientation from DC power output alone across 17 commercial systems, removing the need for installation records or site visits. AI Illustration. Upload story photo >

Researchers have developed a method that recovers the tilt and azimuth of rooftop solar arrays using only DC power output data and site coordinates, removing the need for installation records or site visits. The method was tested on 17 commercial rooftop systems in Australia and published in Solar Energy; it is research-stage, not a commercial product.

Why it matters

Accurate orientation data is essential for estimating electricity generation and detecting faults, yet installation records are often missing or wrong and inspections are impractical across large PV portfolios. A method that infers orientation from operational data could let operators audit large fleets remotely.

In a test on 17 commercial Australian rooftop systems with verified orientations, the method recovered tilt to a median error of about 2° and azimuth to about 4° on unshaded arrays, compared with 10° to 12° and 48° to 62° for clear-sky modeling. The shaded-system figures and the unresolved clipping problem limit how far these results generalize.

The players

University of New South Wales

An Australian research university whose researchers contributed to the study of solar array orientation recovery.

Chinese Academy of Sciences

A Chinese national research institution whose researchers also contributed to the orientation-recovery study.

The details

The method selects periods of high direct sunlight with enough morning and afternoon readings, then uses an optimization algorithm to find the panel orientation that best matches the measured power output. Power is recorded as DC output from each maximum power point tracker (MPPT, the circuit that extracts the most power from a panel string), sampled every five minutes. The algorithm also adjusts the effective system capacity to account for losses, so that the modeled output and the measured output can be compared directly.

Timeline

  1. Operational data was collected from the rooftop systems during 2025.

  2. Findings were published in Solar Energy on November 5, 2026.

The Tech Race

The work sits within a broader effort to replace modeled irradiance inputs with measured operational data for PV monitoring, using clear-sky modeling via pvlib and ERA5 as the prevailing baseline. The reported gains are over that baseline, and the authors' stated next steps target the shading and clipping problems that limit it.

The method is a research result and is not yet available as a commercial tool, so operators cannot use it today. If adopted, it would let owners and maintenance teams verify array orientation remotely, which matters most for large rooftop portfolios where site records are unreliable.

The takeaway

Remote orientation inference from power data could reduce the need for site inspections across large PV fleets, but its accuracy drops sharply under partial shading. Watch for the authors' follow-up work on shading robustness and inverter clipping, which the researchers say is the next step.

What happens next

The researchers plan to reduce the effect of partial shading on recovery accuracy and to make the method robust to inverter clipping and tripping for large-scale use.

Further reading

For more on how solar output is measured and modeled, see the Energy section.