Difference between Pvsyst and NREL's Solar Position Algorithm

[bertin]

Hello,

Does someone know where the difference between PVsyst and NREL's Solar Position Algorithm (SPA) comes from?

I get an important difference in the incident angle between those models.

SPA calculator : http://www.nrel.gov/midc/solpos/spa.html

The parameters used :

- Site location : latitude = 20.5°S ; longitude = 55.3°E ; elevation = 14 m ; time zone = UTC+4 ; No skyline nor shading

- Pannel position : tilt = 0° ; azimuth = 0° (from north, negative toward east)

- Simulation : start date = 01/01/1990 ,end date = 31/12/1990

- The meteorological data for PVsyst simulation comes from Meteonorm. Thus the label "12:00" should represent 12:30.

Difference observed between the models : from +1° to +3° at sunrise (NREL - PVsyst)

Changing the pannel position to : tilt = 2° ; azimuth = -135° (from north, negative toward east), I get :

+3° to + 6° at sunride / +2° to + 3° before 12:00 / -2° to - 3° after 12:00

Based on the Beaminc (incident beam irradiation in the collector plane) calculation (BeamInc = BeamHor * sin Hsoli / sin Hsol), I assume the two models give very different results for the sun heigth on horizontal plane.

Cheers,

Sébastien

[André Mermoud]

I don't know, I have never implemented the NREL algorithms.

The fact that the differences are rather "symmetrical" around 12H indicate that the problem is probably related to the time definition.

The sun progresses by 15°/hour (along its trajectory). Therefore a discrepancy of 3° in the morning may correspond to a very high time discrepancy (3°/15° corresponds to 12 minutes, to be divided by cos(slope of the trajectory at sunrise) ). However in subtropical geometry (trajectory almost "vertical"), I don't understand well how you can get 3 to 5° error at sunrise.

The problem of the accuracy will mainly arise if you perform a retro-transposition for getting the GHI and DHI from a measured DNI value. The morning DNI calculated from the GHI and DHI may have a significant error, but without consequence on the incident irradiance on your almost horizontal plane.

PVsyst provides the opportunity of specifying a time shift, already at the meteo data import time, which can avoid this problem.