dtarin

Total loss

The first section for AC loss should be zero. The MV AC losses are further below, they will not be zero. There is no good way to model it. Modeling with 1 overestimates loss, modeling with 2 underestimates loss. For an in-between figure you can simulate with 1 partition and again with 2, average those and calculate energy delta from your base case, then apply that delta in a new simulation in MQF. For example, base case A is 1 partition, case B is 2 partitions. The two partitions will likely have higher energy. Then calculate (EArrMpp_B - EArrMpp_A) / (2 * EArrMpp_A) and include this result in MQF in a new simulation for case A.

Are you running in fast mode?

For clarification I am not suggesting LID divides by 2. LID occurs on shorter timescale compared to annual degradation and should be taken in full. My overall comment is that taking 1st year degradation from the datasheet and subtracting (annual degradation/2) to get LID may be conservative and something to keep in mind. Different modules may test higher or lower in LID and it is recommended to have module-specific data to support LID assumptions.

Lid and annual power degradation are separate degradation modes. They also occur on different timescales. It is common to assume first year annual degradation divided by 2 (as is done in PVsyst). This may be conservative, which could be fine, but may not be ideal in other aspects. For example, most PERC modules are Ga-doped and certain suppliers test quite low on LID (~1%). For most PERC modules, first year degradation is 2%, 0.45% annual degradation thereafter. The method above would add 0.775% loss over what would be expected for the first year. For TOPCon modules if your LID is 0.5%, you'd have an additional 0.3% loss in year 1. Might be fine in some cases but not in others, something to keep in mind.

Night losses are everything the plant draws at night, including transformers. The utility does not distinguish between devices. Most utility-scale plants (in my opinion) do not have nighttime disconnects on the MV equipment. The MPT is always connected to the grid. It is best to assume (unless you have a detailed design stating otherwise) that the plant will not be disconnected from the grid at night. The caveat here is that depending on how you are using your energy model and if it is flowing into a financial model, you may want to separate these out in case they are considered independently in the financial model.

Transformer losses will come from the manufacturer. If you are modeling transformers in PVsyst, ensure the no-load losses are captured there, and it will reflect in the 8760 for night hours. For MV transformers, 0.10% - 0.16% is a good range for no load losses, for MPTs, 0.04% - 0.06% is a good range. Other items you will need to determine based on plant equipment. 0.2% of POI rating for total hourly night loss at the revenue meter is a good estimate overall (utility scale SAT). Use something higher for cold-weather plants.

If backtracking is enabled, GlobInc considers backtracking, as does GlobEff, and all other irradiance parameters.

Transforming fixed tilt tables to trackers changes their orientation. It looks like the parameters are being passed directly to the new tracker object like azimuth and tilt, when these are calculated/defined differently for each type of object, and should be recalculated so that the object remains in place.

v8.0.6

The first module is not "inferior", module efficiency is 20% vs 17%, making it the better module. There must also be a significant difference in the temperature coefficient.

Custom tracker is a better option than central. It reduces simulation time and can approximate very close to all trackers; it just depends on how good of a tracker you select. Simply looking at the scene visually for a location which has a good representation of the site generally works (for larger sites), but smaller, complex sites might take some trial and error. In any case we see <0.1% delta using custom tracker compared to all trackers.

Check that bifacial is turned on for both subarrays. Looks like it is not.

Slow can be used on any size system, but simulations will extend to multiple hours for large, complex sites. The main difference that I see between slow and fast (modeling projects in US) is that fast will have higher near shading and lower electrical loss, and for slow, it is the opposite. Not always, but I would say more often than not. In total, shading losses are typically comparable for most sites with fast being slightly higher. Sometimes difference is negligible, but depends on the site; less complex and utility scale will have closer figures. More complex sites with topography, smaller sites with severe shading, will benefit more from slow simulation and the increased accuracy with calculating electrical loss. As for when to use, it depends on your position, project, etc. For utility scale in development stages, fast makes sense, and then moving to slow (or at least benchmarking slow) later on when design is closer to final, prior to financing, construction, etc. If the project is small and simulation time is short, I dont see a reason not to run slow for single, one-off estimates. But if you're doing batch runs, running multiple designs, evaluating different components or weather files, in early stage and need indicative numbers, etc., fast is fine. What's more important is understanding the differences and impacts to simulation results (irradiance, shading, production, etc.), and deciding when to prioritize simulation speed over accuracy and vice versa (and knowing if there even exists material differences).

Shading working fine from what I see, 8.0.5.