Design versus as-built variance check

Design file

Modules

Tilt (degrees)

Azimuth (degrees, 0 to 360)

Shading loss (%)

Modelled energy (kWh a year)

As built (site measurements and a re-run of the model)

Modules

Tilt (degrees)

Azimuth (degrees, 0 to 360)

Shading loss (%)

Model re-run with as-built values (kWh a year)

Meter and weather

Metered energy, first year (kWh)

Measured irradiance as a share of the model's typical year (%)

Your tolerances

Modules (count)

Tilt (degrees)

Azimuth (degrees)

Shading (percentage points)

Modelled effect of the geometry changes (%)

Unexplained production residual (%)

All starting values are invented, including the tolerances and the weather share. The check does not find causes: the residual can hold soiling, inverter faults, degradation, a wrong irradiance share or a metering error. Enter 14600 or 14,600; 2.5 and 2,5 are read as decimals.

How many test jobs does a "within tolerance" claim need?

Jobs tested

Jobs outside tolerance

One-sided 95% upper bound on the share of jobs that fall outside tolerance, from the exact binomial (Clopper-Pearson) bound. It assumes each job is an independent random draw. Several measurements on one roof are not independent, so the real bound is wider.