PVSyst vs pvlib tracking angle comparison

[kapetav]

Hello,

I'm comparing the tracking angles from a simple PVSyst simulation with the tracking angles given by pvlib.tracking.singleaxis(), and I expect a very similar result between them. I do get a near-perfect match during the middle of the day, however I see the angles deviate significantly during backtracking times. The angles match only when I slightly modify the GCR in the pvlib function (from 50% to 50.25%), which suggests to me that the backtracking GCR in PVSyst may not be what I set myself (50%). Am I setting up the PVSyst simulation properly to compare with the pvlib function? I understand that there are "hidden parameters", but I haven't found one that changes the backtracking angles.

My PVSyst setup is as follows:

- 5 single-axis trackers defined on flat ground, 150 modules total, 1P orientation

- module length: 2000mm, module width: 1000mm

- unlimited trackers, pitch: 4m, GCR: 50%

My pvlib inputs are:

pvlib_tracking = tracking.singleaxis(
            apparent_zenith=90 - df["HSol"],
            apparent_azimuth=df["AzSol"] + 180,
            axis_tilt=0,
            axis_azimuth=180,
            cross_axis_tilt=0,
            max_angle=60,
            backtrack=True,
            gcr=0.5,  
        )

I compare pvlib's "tracker_theta" output with PVSyst's "PhiAng" output. The plot below shows the yearly differences (up to 5-6 degrees during backtracking times):

When I change the pvlib GCR to 0.5025 (corresponds to a ~0.02m difference in the pitch), the discrepancy goes away:

To better match the pvlib function at GCR=50% I tried setting PVSyst hidden parameters such as "Shed field default frame margin" to zero, but I didn't see a change in the resulting angles.

 

[Michele Oliosi]

There is indeed a baked-in 1 cm addition to the effective width of the tracker (both sides) when using backtracking in PVsyst. This is meant to account for possible inaccuracies in placement. When this was implemented, it was expected that such precautions (leaving a margin of error) are applied in the field. Probably it is still the case?

[kapetav]

Hi Michele, thanks this helps! To confirm, how would I best account for this added width? By increasing the backtracking pitch like below by 2cm?

 

[Michele Oliosi]

To compare with baseline pvlib, I would rather remove 2 cm to the "Sensitive width". Might be 1 cm, I am not 100% sure if 2 is double counting.

[kapetav]

It seems I get the same result whether I add 2cm to the pitch or subtract 1cm from the sensitive width:

[Michele Oliosi]

yes this is because of the ratio of width and pitch (GCR) which ultimately controls the tracking algorithm. In your case the GCR is .5. which explains that these two modifications are equivalent (approximately). I would still suggest the remove 1 cm option as better since it mimics most correctly what pvsyst does and is more general to other GCR ratios.