Comparison Daysim with Measured data

radiance-daysim
1

Dear readers,
Currently I'm using Daysim (version 3.1b (beta)) for the simulation of irradiance on a new type of BIPV-system. According to what I've read in several articles, this should be possible. I am comparing my simulation results to irradiance data measured at a full-scale mockup of the BIPV-system using on-site recorded solar radiation data as input for the simulations. For a large part of the year, Daysim predicts almost the exact same irradiance as measured. However, suddenly from September on, Daysim significantly under-predicts the results.
When looking at daily sun profiles and irradiance simulated and measured, it appears that Daysim uses only the diffuse irradiance instead of both direct and diffuse irradiance. Figure 1 shows a typical day, early in the year, where you can see that the agreement between measurement and simulation is quite good. Figure 2 is for 25 September, where you can see that the measured irradiance is much higher than the simulation results from Daysim and that irradiance simulated with Daysim follows almost exactly the diffuse irradiance.
Figure 1: https://www.dropbox.com/s/11xu2ivuyf866zg/Comparison%20Daysim%20%26%20Simulation%20-%205%20May.png?dl=0
Figure 2: https://www.dropbox.com/s/7wj8svs5nraex54/Comparison%20Daysim%20%26%20Simulation%20-%2025%20September.png?dl=0
I've checked the following things:

- Perhaps there was a flaw in the weather file. I'm using a weather file created from a measurement station (DNI and DIFHI) and validated that it is correctly implemented in Daysim. I simulated the system also with a weather file based on an EPW file. Similar results occur.

- Perhaps there is a flaw in the design. I've modelled the system in SketchUp and imported this in Daysim. To see if the model was corrupted I have built a simple cube in SketchUp with sensors on the same orientation as the PV-system. Again similar results occur.
Even though there are some slight changes during the year, one would expect a mirroring of the results around January (January to June show similar results as July to December, but then mirrored). However, the results in January from the simulation in Daysim show again almost perfect results with the measurements.
I've talked to several people about these deviations, but none of them can give a clear explanation why this happens.
I hope that some of you can give an explanation and help me with this problem.
With kind regards,
Stefan Koenders

2

Hi Stefan,

are you sure the weather file you got from the weather station is from the same year where your measurement took place ?

What kind of data you got from the weather station? (Direct-normal, Direct horiziontal, Diffuse horizontal, Global)?
Do you have an obstruction in the simulation (and in reality)? If yes, is it continuous (like a block?) or something with a pattern (like stripes or large leaves from a tree)?

Jan

···

On 09/02/17 13:54, Koenders, S.J.M. wrote:

Dear readers,

Currently I�m using Daysim (version 3.1b (beta)) for the simulation of irradiance on a new type of BIPV-system. According to what I�ve read in several articles, this should be possible. I am comparing my simulation results to irradiance data measured at a full-scale mockup of the BIPV-system using on-site recorded solar radiation data as input for the simulations. For a large part of the year, Daysim predicts almost the exact same irradiance as measured. However, suddenly from September on, Daysim significantly under-predicts the results.

When looking at daily sun profiles and irradiance simulated and measured, it appears that Daysim uses only the diffuse irradiance instead of both direct and diffuse irradiance. Figure 1 shows a typical day, early in the year, where you can see that the agreement between measurement and simulation is quite good. Figure 2 is for 25 September, where you can see that the measured irradiance is much higher than the simulation results from Daysim and that irradiance simulated with Daysim follows almost exactly the diffuse irradiance.

Figure 1: Dropbox - File Deleted - Simplify your life

Figure 2: Dropbox - File Deleted - Simplify your life

I�ve checked the following things:

-Perhaps there was a flaw in the weather file. I�m using a weather file created from a measurement station (DNI and DIFHI) and validated that it is correctly implemented in Daysim. I simulated the system also with a weather file based on an EPW file. Similar results occur.

-Perhaps there is a flaw in the design. I�ve modelled the system in SketchUp and imported this in Daysim. To see if the model was corrupted I have built a simple cube in SketchUp with sensors on the same orientation as the PV-system. Again similar results occur.

Even though there are some slight changes during the year, one would expect a mirroring of the results around January (January to June show similar results as July to December, but then mirrored). However, the results in January from the simulation in Daysim show again almost perfect results with the measurements.

I�ve talked to several people about these deviations, but none of them can give a clear explanation why this happens.

I hope that some of you can give an explanation and help me with this problem.

With kind regards,

Stefan Koenders

_______________________________________________
Radiance-daysim mailing list
[email protected]
http://www.radiance-online.org/mailman/listinfo/radiance-daysim

--
Dr.-Ing. Jan Wienold
Ecole Polytechnique F�d�rale de Lausanne (EPFL)
EPFL ENAC IA LIPID

LE 1 111 (Office)
Phone +41 21 69 30849

3

Hi Stefan,
Do you use gen_dc and ds_illum in regular mode (recommended) or with the "-dds" option?

Christoph

···

From: Jan Wienold [mailto:[email protected]]
Sent: Thursday, February 09, 2017 8:44 AM
To: DAYSIM discussion <[email protected]>
Subject: Re: [Radiance-daysim] Comparison Daysim with Measured data

Hi Stefan,

are you sure the weather file you got from the weather station is from the same year where your measurement took place ?

What kind of data you got from the weather station? (Direct-normal, Direct horiziontal, Diffuse horizontal, Global)?
Do you have an obstruction in the simulation (and in reality)? If yes, is it continuous (like a block?) or something with a pattern (like stripes or large leaves from a tree)?

Jan

On 09/02/17 13:54, Koenders, S.J.M. wrote:
Dear readers,
Currently I'm using Daysim (version 3.1b (beta)) for the simulation of irradiance on a new type of BIPV-system. According to what I've read in several articles, this should be possible. I am comparing my simulation results to irradiance data measured at a full-scale mockup of the BIPV-system using on-site recorded solar radiation data as input for the simulations. For a large part of the year, Daysim predicts almost the exact same irradiance as measured. However, suddenly from September on, Daysim significantly under-predicts the results.
When looking at daily sun profiles and irradiance simulated and measured, it appears that Daysim uses only the diffuse irradiance instead of both direct and diffuse irradiance. Figure 1 shows a typical day, early in the year, where you can see that the agreement between measurement and simulation is quite good. Figure 2 is for 25 September, where you can see that the measured irradiance is much higher than the simulation results from Daysim and that irradiance simulated with Daysim follows almost exactly the diffuse irradiance.
Figure 1: https://www.dropbox.com/s/11xu2ivuyf866zg/Comparison%20Daysim%20%26%20Simulation%20-%205%20May.png?dl=0
Figure 2: https://www.dropbox.com/s/7wj8svs5nraex54/Comparison%20Daysim%20%26%20Simulation%20-%2025%20September.png?dl=0
I've checked the following things:

- Perhaps there was a flaw in the weather file. I'm using a weather file created from a measurement station (DNI and DIFHI) and validated that it is correctly implemented in Daysim. I simulated the system also with a weather file based on an EPW file. Similar results occur.

- Perhaps there is a flaw in the design. I've modelled the system in SketchUp and imported this in Daysim. To see if the model was corrupted I have built a simple cube in SketchUp with sensors on the same orientation as the PV-system. Again similar results occur.
Even though there are some slight changes during the year, one would expect a mirroring of the results around January (January to June show similar results as July to December, but then mirrored). However, the results in January from the simulation in Daysim show again almost perfect results with the measurements.
I've talked to several people about these deviations, but none of them can give a clear explanation why this happens.
I hope that some of you can give an explanation and help me with this problem.
With kind regards,
Stefan Koenders

_______________________________________________

Radiance-daysim mailing list

[email protected]<mailto:[email protected]>

http://www.radiance-online.org/mailman/listinfo/radiance-daysim

--

Dr.-Ing. Jan Wienold

Ecole Polytechnique Fédérale de Lausanne (EPFL)

EPFL ENAC IA LIPID

http://people.epfl.ch/jan.wienold

LE 1 111 (Office)

Phone +41 21 69 30849

4

Dear Jan & Christoph,
Thank you for your fast response. To answer both of your questions:

- I'm using the gen_dc and ds_illum in the DDS option. This was chosen due to the "most accurate" solution. Christoph, can you explain why you recommend the regular mode instead of the DDS option?

- I'm hundred percent certain that the measurement data is from the correct year. It was measured last year, January 2016 till December 2016, and I'm still getting the data.

- The data I get from the weather station is DNI, DHI, DiffHI and GHI. I built two weather files (DHI file and DNI file) and simulated with both of them. The first file is built from DHI & DiffHI and the second from DNI & DiffHI. One would expect similar results, but there was a strong deviation between the two results. The weather file based on DNI and DiffHI gave the best results.

- There is an obstruction in reality, but this object was placed and designed in such a way that 99% of the time there is no obstruction from this object. Several researches before showed that this obstruction could be neglected.

- The system is placed on a flat roof, so there is no other obstruction or strange pattern that I know of.
What I find particularly strange is the sudden change in behavior from September on. It would be strange if this was due to an obstruction or something, because this should also show in March/April (due to similar solar elevation and azimuth).
With kind regards,
Stefan Koenders

···

From: Christoph Reinhart [mailto:[email protected]]
Sent: donderdag 9 februari 2017 15:06
To: DAYSIM discussion <[email protected]>
Subject: Re: [Radiance-daysim] Comparison Daysim with Measured data

Hi Stefan,
Do you use gen_dc and ds_illum in regular mode (recommended) or with the "-dds" option?

Christoph

From: Jan Wienold [mailto:[email protected]]
Sent: Thursday, February 09, 2017 8:44 AM
To: DAYSIM discussion <[email protected]<mailto:[email protected]>>
Subject: Re: [Radiance-daysim] Comparison Daysim with Measured data

Hi Stefan,

are you sure the weather file you got from the weather station is from the same year where your measurement took place ?

What kind of data you got from the weather station? (Direct-normal, Direct horiziontal, Diffuse horizontal, Global)?
Do you have an obstruction in the simulation (and in reality)? If yes, is it continuous (like a block?) or something with a pattern (like stripes or large leaves from a tree)?

Jan

On 09/02/17 13:54, Koenders, S.J.M. wrote:
Dear readers,
Currently I'm using Daysim (version 3.1b (beta)) for the simulation of irradiance on a new type of BIPV-system. According to what I've read in several articles, this should be possible. I am comparing my simulation results to irradiance data measured at a full-scale mockup of the BIPV-system using on-site recorded solar radiation data as input for the simulations. For a large part of the year, Daysim predicts almost the exact same irradiance as measured. However, suddenly from September on, Daysim significantly under-predicts the results.
When looking at daily sun profiles and irradiance simulated and measured, it appears that Daysim uses only the diffuse irradiance instead of both direct and diffuse irradiance. Figure 1 shows a typical day, early in the year, where you can see that the agreement between measurement and simulation is quite good. Figure 2 is for 25 September, where you can see that the measured irradiance is much higher than the simulation results from Daysim and that irradiance simulated with Daysim follows almost exactly the diffuse irradiance.
Figure 1: https://www.dropbox.com/s/11xu2ivuyf866zg/Comparison%20Daysim%20%26%20Simulation%20-%205%20May.png?dl=0
Figure 2: https://www.dropbox.com/s/7wj8svs5nraex54/Comparison%20Daysim%20%26%20Simulation%20-%2025%20September.png?dl=0
I've checked the following things:

- Perhaps there was a flaw in the weather file. I'm using a weather file created from a measurement station (DNI and DIFHI) and validated that it is correctly implemented in Daysim. I simulated the system also with a weather file based on an EPW file. Similar results occur.

- Perhaps there is a flaw in the design. I've modelled the system in SketchUp and imported this in Daysim. To see if the model was corrupted I have built a simple cube in SketchUp with sensors on the same orientation as the PV-system. Again similar results occur.
Even though there are some slight changes during the year, one would expect a mirroring of the results around January (January to June show similar results as July to December, but then mirrored). However, the results in January from the simulation in Daysim show again almost perfect results with the measurements.
I've talked to several people about these deviations, but none of them can give a clear explanation why this happens.
I hope that some of you can give an explanation and help me with this problem.
With kind regards,
Stefan Koenders

_______________________________________________

Radiance-daysim mailing list

[email protected]<mailto:[email protected]>

http://www.radiance-online.org/mailman/listinfo/radiance-daysim

--

Dr.-Ing. Jan Wienold

Ecole Polytechnique Fédérale de Lausanne (EPFL)

EPFL ENAC IA LIPID

http://people.epfl.ch/jan.wienold

LE 1 111 (Office)

Phone +41 21 69 30849

5

Hi Stefan,

I don't think it is necessary to use DDS - the sun-resolution is not that much higher than in the standard version and the diffuse patches for the sky it does not matter so much. So I would try the "standard" version. I also encountered sever problems (core-dumps on the linux version) once with the DDS option, because I wanted to use another sun-interpolation mode. This was initially included in all daysim modes, but it "disappeared" once for the standard.

That's why I was asking about external obstructions with a pattern - in that case you can run into problems with the sun-discretisation. But this seems not the case here - I think the DDS-option might have caused this "September-problem".

For the difference in DHI and DNI-files - are you sure you set the the right parameter in the header-file (and weather-file). It should be

/weather_data_file_units//1 /for Direct Normal, Diffuse Horizontal Irradiance[W/m²]

/weather_data_file_units//2/ for Direct Horizontal, Diffuse Horizontal Irradiance[W/m²]

Have you checked, that DHI+DIFFHI=GHI ? Sometimes the measurements are also not that reliable.

Jan

···

On 09/02/17 16:41, Koenders, S.J.M. wrote:

Dear Jan & Christoph,

Thank you for your fast response. To answer both of your questions:

-I’m using the gen_dc and ds_illum in the DDS option. This was chosen due to the “most accurate” solution. Christoph, can you explain why you recommend the regular mode instead of the DDS option?

-I’m hundred percent certain that the measurement data is from the correct year. It was measured last year, January 2016 till December 2016, and I’m still getting the data.

-The data I get from the weather station is DNI, DHI, DiffHI and GHI. I built two weather files (DHI file and DNI file) and simulated with both of them. The first file is built from DHI & DiffHI and the second from DNI & DiffHI. One would expect similar results, but there was a strong deviation between the two results. The weather file based on DNI and DiffHI gave the best results.

-There is an obstruction in reality, but this object was placed and designed in such a way that 99% of the time there is no obstruction from this object. Several researches before showed that this obstruction could be neglected.

-The system is placed on a flat roof, so there is no other obstruction or strange pattern that I know of.

What I find particularly strange is the sudden change in behavior from September on. It would be strange if this was due to an obstruction or something, because this should also show in March/April (due to similar solar elevation and azimuth).

With kind regards,

Stefan Koenders

*From:*Christoph Reinhart [mailto:tito_@mit.edu]
*Sent:* donderdag 9 februari 2017 15:06
*To:* DAYSIM discussion <[email protected]>
*Subject:* Re: [Radiance-daysim] Comparison Daysim with Measured data

Hi Stefan,

Do you use gen_dc and ds_illum in regular mode (recommended) or with the “-dds” option?

Christoph

*From:*Jan Wienold [mailto:[email protected]]
*Sent:* Thursday, February 09, 2017 8:44 AM
*To:* DAYSIM discussion <[email protected] <mailto:[email protected]>>
*Subject:* Re: [Radiance-daysim] Comparison Daysim with Measured data

Hi Stefan,

are you sure the weather file you got from the weather station is from the same year where your measurement took place ?

What kind of data you got from the weather station? (Direct-normal, Direct horiziontal, Diffuse horizontal, Global)?
Do you have an obstruction in the simulation (and in reality)? If yes, is it continuous (like a block?) or something with a pattern (like stripes or large leaves from a tree)?

Jan

On 09/02/17 13:54, Koenders, S.J.M. wrote:

    Dear readers,

    Currently I’m using Daysim (version 3.1b (beta)) for the
    simulation of irradiance on a new type of BIPV-system. According
    to what I’ve read in several articles, this should be possible. I
    am comparing my simulation results to irradiance data measured at
    a full-scale mockup of the BIPV-system using on-site recorded
    solar radiation data as input for the simulations. For a large
    part of the year, Daysim predicts almost the exact same irradiance
    as measured. However, suddenly from September on, Daysim
    significantly under-predicts the results.

    When looking at daily sun profiles and irradiance simulated and
    measured, it appears that Daysim uses only the diffuse irradiance
    instead of both direct and diffuse irradiance. Figure 1 shows a
    typical day, early in the year, where you can see that the
    agreement between measurement and simulation is quite good. Figure
    2 is for 25 September, where you can see that the measured
    irradiance is much higher than the simulation results from Daysim
    and that irradiance simulated with Daysim follows almost exactly
    the diffuse irradiance.

    Figure 1:
    Dropbox - File Deleted - Simplify your life

    Figure 2:
    Dropbox - File Deleted - Simplify your life

    I’ve checked the following things:

    -Perhaps there was a flaw in the weather file. I’m using a weather
    file created from a measurement station (DNI and DIFHI) and
    validated that it is correctly implemented in Daysim. I simulated
    the system also with a weather file based on an EPW file. Similar
    results occur.

    -Perhaps there is a flaw in the design. I’ve modelled the system
    in SketchUp and imported this in Daysim. To see if the model was
    corrupted I have built a simple cube in SketchUp with sensors on
    the same orientation as the PV-system. Again similar results occur.

    Even though there are some slight changes during the year, one
    would expect a mirroring of the results around January (January to
    June show similar results as July to December, but then mirrored).
    However, the results in January from the simulation in Daysim show
    again almost perfect results with the measurements.

    I’ve talked to several people about these deviations, but none of
    them can give a clear explanation why this happens.

    I hope that some of you can give an explanation and help me with
    this problem.

    With kind regards,

    Stefan Koenders

    _______________________________________________

    Radiance-daysim mailing list

    [email protected]
    <mailto:[email protected]>

    http://www.radiance-online.org/mailman/listinfo/radiance-daysim

--
Dr.-Ing. Jan Wienold
Ecole Polytechnique Fédérale de Lausanne (EPFL)
EPFL ENAC IA LIPID
http://people.epfl.ch/jan.wienold
LE 1 111 (Office)
Phone +41 21 69 30849

_______________________________________________
Radiance-daysim mailing list
[email protected]
http://www.radiance-online.org/mailman/listinfo/radiance-daysim

--
Dr.-Ing. Jan Wienold
Ecole Polytechnique Fédérale de Lausanne (EPFL)
EPFL ENAC IA LIPID

LE 1 111 (Office)
Phone +41 21 69 30849

6

Hi Stefan,

can you test the monitored data against other data sources? For many locations, you can get (at least interpolated) irradiance data from other weather stations. If you find e.g. good accordance between your data and that of others until Sep, but not afterwards, you would have to check the instruments rather than the simulation.

One point to remember is that the approach to reconstruct the sky luminance distribution from the two measured irradiance integrals works well for perfectly clear or perfectly covered skys. If you have a sky with distinguishable clouds, the distribution can not be reconstructed. So while you had a highly non-uniform sky distribution in real world, your sky model would average all couds/sky directions other than the sun. With annual simulations, the assumption is that this does not make a big difference. For point-in-time simulations, this may not apply. In such cases, monitoring of cloud coverage makes sense.

Cheers,
Lars.

···

Am 09.02.2017 um 16:41 schrieb Koenders, S.J.M. <[email protected]>:

Dear Jan & Christoph,

Thank you for your fast response. To answer both of your questions:

- I’m using the gen_dc and ds_illum in the DDS option. This was chosen due to the “most accurate” solution. Christoph, can you explain why you recommend the regular mode instead of the DDS option?
- I’m hundred percent certain that the measurement data is from the correct year. It was measured last year, January 2016 till December 2016, and I’m still getting the data.
- The data I get from the weather station is DNI, DHI, DiffHI and GHI. I built two weather files (DHI file and DNI file) and simulated with both of them. The first file is built from DHI & DiffHI and the second from DNI & DiffHI. One would expect similar results, but there was a strong deviation between the two results. The weather file based on DNI and DiffHI gave the best results.
- There is an obstruction in reality, but this object was placed and designed in such a way that 99% of the time there is no obstruction from this object. Several researches before showed that this obstruction could be neglected.
- The system is placed on a flat roof, so there is no other obstruction or strange pattern that I know of.

What I find particularly strange is the sudden change in behavior from September on. It would be strange if this was due to an obstruction or something, because this should also show in March/April (due to similar solar elevation and azimuth).

With kind regards,

Stefan Koenders

From: Christoph Reinhart [mailto:tito_@mit.edu <mailto:[email protected]>]
Sent: donderdag 9 februari 2017 15:06
To: DAYSIM discussion <[email protected] <mailto:[email protected]>>
Subject: Re: [Radiance-daysim] Comparison Daysim with Measured data

Hi Stefan,

Do you use gen_dc and ds_illum in regular mode (recommended) or with the “-dds” option?

Christoph

From: Jan Wienold [mailto:[email protected] <mailto:[email protected]>]
Sent: Thursday, February 09, 2017 8:44 AM
To: DAYSIM discussion <[email protected] <mailto:[email protected]>>
Subject: Re: [Radiance-daysim] Comparison Daysim with Measured data

Hi Stefan,

are you sure the weather file you got from the weather station is from the same year where your measurement took place ?

What kind of data you got from the weather station? (Direct-normal, Direct horiziontal, Diffuse horizontal, Global)?
Do you have an obstruction in the simulation (and in reality)? If yes, is it continuous (like a block?) or something with a pattern (like stripes or large leaves from a tree)?

Jan

On 09/02/17 13:54, Koenders, S.J.M. wrote:
Dear readers,

Currently I’m using Daysim (version 3.1b (beta)) for the simulation of irradiance on a new type of BIPV-system. According to what I’ve read in several articles, this should be possible. I am comparing my simulation results to irradiance data measured at a full-scale mockup of the BIPV-system using on-site recorded solar radiation data as input for the simulations. For a large part of the year, Daysim predicts almost the exact same irradiance as measured. However, suddenly from September on, Daysim significantly under-predicts the results.

When looking at daily sun profiles and irradiance simulated and measured, it appears that Daysim uses only the diffuse irradiance instead of both direct and diffuse irradiance. Figure 1 shows a typical day, early in the year, where you can see that the agreement between measurement and simulation is quite good. Figure 2 is for 25 September, where you can see that the measured irradiance is much higher than the simulation results from Daysim and that irradiance simulated with Daysim follows almost exactly the diffuse irradiance.

Figure 1: Dropbox - File Deleted - Simplify your life <Dropbox - File Deleted - Simplify your life;
Figure 2: Dropbox - File Deleted - Simplify your life <Dropbox - File Deleted - Simplify your life;
I’ve checked the following things:

- Perhaps there was a flaw in the weather file. I’m using a weather file created from a measurement station (DNI and DIFHI) and validated that it is correctly implemented in Daysim. I simulated the system also with a weather file based on an EPW file. Similar results occur.
- Perhaps there is a flaw in the design. I’ve modelled the system in SketchUp and imported this in Daysim. To see if the model was corrupted I have built a simple cube in SketchUp with sensors on the same orientation as the PV-system. Again similar results occur.

Even though there are some slight changes during the year, one would expect a mirroring of the results around January (January to June show similar results as July to December, but then mirrored). However, the results in January from the simulation in Daysim show again almost perfect results with the measurements.

I’ve talked to several people about these deviations, but none of them can give a clear explanation why this happens.

I hope that some of you can give an explanation and help me with this problem.

With kind regards,

Stefan Koenders

_______________________________________________
Radiance-daysim mailing list
[email protected] <mailto:[email protected]>
http://www.radiance-online.org/mailman/listinfo/radiance-daysim

--
Dr.-Ing. Jan Wienold
Ecole Polytechnique Fédérale de Lausanne (EPFL)
EPFL ENAC IA LIPID

http://people.epfl.ch/jan.wienold
LE 1 111 (Office)
Phone +41 21 69 30849
_______________________________________________
Radiance-daysim mailing list
[email protected] <mailto:[email protected]>
http://www.radiance-online.org/mailman/listinfo/radiance-daysim

7

Hi Stefan,
Have you done any graphical or numerical checks of the weather data values to see if there's something not right in the data, that could give incorrect results in both the measured and the EPW weather file? If you can view the data visually you might see a pattern that points out any flaws in the data. I'd recommend Dview as one example to check for these things. https://beopt.nrel.gov/downloadDView
I can't remember which one, but in the past I've seen EPW files with one whole month of illuminance values roughly 1/100th of what they should be. It looked like somehow the data file probably had the irradiance values incorrectly copied into the illuminance columns.
-Chris

- Perhaps there was a flaw in the weather file. I'm using a weather file created from a measurement station (DNI and DIFHI) and validated that it is correctly implemented in Daysim. I simulated the system also with a weather file based on an EPW file. Similar results occur.

···

From: Koenders, S.J.M. [mailto:[email protected]]

8

Dear all,
Thank you for the fast response and the given information.
I’ve checked my weather files both visually and manually (data-wise) and no strange things showed. After this check I changed from DDS to the regular calculation option . Jan was indeed correct, the DDS-option caused the problem. I ran the simulation again and there is now a much better fit with the measurement results (very high correlation) and the “September-problem” does not occur anymore.
For now it seems that this was the problem, so thank you all for helping me!
With kind regards,
Stefan Koenders

···

From: Jan Wienold [mailto:[email protected]]
Sent: donderdag 9 februari 2017 17:18
To: DAYSIM discussion <[email protected]<mailto:[email protected]>>
Subject: Re: [Radiance-daysim] Comparison Daysim with Measured data

Hi Stefan,

I don't think it is necessary to use DDS - the sun-resolution is not that much higher than in the standard version and the diffuse patches for the sky it does not matter so much. So I would try the "standard" version. I also encountered sever problems (core-dumps on the linux version) once with the DDS option, because I wanted to use another sun-interpolation mode. This was initially included in all daysim modes, but it "disappeared" once for the standard.

That's why I was asking about external obstructions with a pattern - in that case you can run into problems with the sun-discretisation. But this seems not the case here - I think the DDS-option might have caused this "September-problem".

For the difference in DHI and DNI-files - are you sure you set the the right parameter in the header-file (and weather-file). It should be

weather_data_file_units 1 for Direct Normal, Diffuse Horizontal Irradiance [W/m²]

weather_data_file_units 2 for Direct Horizontal, Diffuse Horizontal Irradiance [W/m²]

Have you checked, that DHI+DIFFHI=GHI ? Sometimes the measurements are also not that reliable.
Jan