Real Values vs. PVSYST Simulations Mismatch

[Xavien]

Dear all,

I'm using PVsyst to calculate the expected output of a power plant.

I do 3 calculations based on 3 different meteorological datasets

- Meteonorm

- NASA

- PVGIS

I extract hourly calculations from CSV file and relate them to real field values. I believe I am doing something wrong. I use V5.73 due to license.

Here's the issue.

So this is the PVsyst report. I grab the CSV values from it and use it in an excel file

Here you can download the file that contains our measurements and PVsyst CSV files (merged) from 3 different Meteo sources.

This is the excel style that I use for manual tracking of the values.

Here in this month; you can see that our irradiation is 110,5% above the average of MN-PVGIS-NASA. However the real production is 78,4% of the average of the PVsyst simulations. The system is overloaded but it had never been a difference of %20. What might be the issue of this since this is not minot but major.

BR,

[André Mermoud]

Yes of course.

It is quite impossible to compare the instantaneous values with measured ones when you have meteo data specified in monthly values, which is the case for Meteonorm, PVGIS and NASA.

With such monthly data, the Daily and Hourly meteo data are randomly generated (synthetic hourly data generator), and cannot correspond to any real instantaneous data.

The only way for analyzing measured data of a power plant is to measure the Meteo (irradiance and temperature) on the site (or very eventually to get simultaneous measurements from a nearby station, or Satellite with a very narrow resolution).

And even the Monthly values cannot be compared accurately, as the data of the Historical sources are usually averages over several years, which cannot correspond to the weather of your particular month.

[Xavien]

Dear Andre,

You are right however I'm not comparing daily values; instead I'm comparing the monthly sum.

If you check the latest image in my post; at the bottom you will see monthy sum of irradiation is 101,5 kWh/m2. My corresponding production is 75MWh. In the simulation; it gives 95MWh for 104,3 kWh/m2. This I believe has to be inline with the irradiation.

Difference between real measured irradiance (monthly sum) and simulated one ise 104,3kWh/m2 / 101,5 kWh/m2 = %2

However corresponding montly production is 95MWh / 75 MWh = %26.

So that's my point. Input irradiance value is almost the same to the system however the output is very different. I believe PVsyst is overestimating it due to some errors in my config. I can share the meteo data file If you prefer to do the simulation by yourself.

BR,

[Xavien]

Dear Mhamon,

In order to be %100 precise; you are very much right. However my assumption is there shouldn't be a %26 difference in production given we have a %2 difference in irradiation. Somewhere between %5 difference is quite normal but %26 is unacceptable.

Attached you can see real measured values from site in 5 mins intervals. Measured values are ambient temp, module temp, irradiation and corresponding production for that interval. In addition; I'm providing the meteonorm data for the location but I'm afraid we will again have the same difference (%26).

I believe there is something I do wrong so the results are not correct. (a deviation of %5 from real values is acceptable for me)

BR,

RealtimeValuesPart1.rar

Realtime values for February Part1 (5mins Interval)

RealtimeValuesPart2.rar

Realtime values for February Part2 (5mins Interval)

MeteonormData.rar

Meteonorm data for the selected location

[Xavien]

Hello mhamon,

I'm not comparing daily values. I'm comparing monthly sum.

My monthly irradiation is 101,5 kWh/m2. My corresponsing monthly production is 75MWh (given the config in my first post).

Meteonorms montly irradiation is 104,5 kWh/m2. PVsyst's corresponding monthly production is 95MWh (again in my config).

There is a %26 difference in monthly production given %2 difference in monthly radiation.

So I believe the issue is not with Meteonorm data; It's with PVsyst simulation variables.

Andre, I can share the project file as well if you want.

BR,

[André Mermoud]

I don't understand well what you are doing.

Please take your project. Run the simulation with one meteo data file (for example what you call "My monthly meteo is 101.5 kWh", but don't define frome where it is taken).

Then just change the input meteo data file in your project (for example meteonorm). Run it again without changing anything.

The resulting yield should be fairly proportionnal to the input irradiance.

Now if there are big discrepancies these may be due to a very bad hourly meteo file (for example a time shift).

Please carefully check your meteo file using "Databases > Meteo tables and graphs > Check data quality", and here press F1 for detailed explanations.

[Xavien]

I don't understand well what you are doing.

Please take your project. Run the simulation with one meteo data file (for example what you call "My monthly meteo is 101.5 kWh", but don't define frome where it is taken).

Then just change the input meteo data file in your project (for example meteonorm). Run it again without changing anything.

The resulting yield should be fairly proportionnal to the input irradiance.

 

That's correct. I've done the same thing with data from NASA and PVGIS as well. PVsyst's simulation results are proportional to the data provided.

 

Now if there are big discrepancies these may be due to a very bad hourly meteo file (for example a time shift).

Please carefully check your meteo file using "Databases > Meteo tables and graphs > Check data quality", and here press F1 for detailed explanations.

 

My issue is the PVsyst's simulations v.s. real production. PVsyst estimates that there should be 95MWh per month production given 101,5kWh/m2/month irradiation. However real production is 75MWh per month instead. That's a huge difference.

The question is; is PVsyst over estimating or is my system under performing ?

[André Mermoud]

The discrepancy between Simulated and Measured data may be due to:

- An error in the measurements,

- A misrunning of the system,

- An error in the simulation program

- An error in the design parameters used in this simulation,

- An error in the input meteo data (either measurement or importing in PVsyst).

Nobody can anwer your question without more information. You have to check all these possibilities.

The better way is to analyze hourly data discrepancies.

However PVsyst is used by tausends of people. I don't think that a simulation error is very probable...