Clear Sky irradiance and weather influence

[Marius Schmiedt]

Hi together,

I'm new to pvlib and have managed it to determine the solar position and the clear sky irradiance. But I have a question to the clear sky models Ineichen and simplified solis. As far as I understood is the Linke turbidity factor is a function of the aod and precipitable water. But aren't they also influenced by weather? The Linke turbidity stored in the data folder seems constant over every year. Doesn't it vary from year to year?

And how can the irradiance during cloudy and/or rainy days be determined? I know in a previous release there was the forecast class but it was deprecated. Are there any other methods to determine the irradiance in absence of clear sky?

Best regards

Marius 

[Will Hobbs]

Hi Marius.

I recommend you start with free satellite-based weather datasets. There are also free ground measurement datasets and commercial satellite datasets (some have limited free data, as well).

Using functions in 

https://pvlib-python.readthedocs.io/en/stable/reference/iotools.html is the easiest way to get data to use in pvlib. For example, get_pvgis_hourly to get one version of PVGIS data. 

Also see:

- 

https://solarstations.org/intro.html

- 

https://nsrdb.nrel.gov/ (source for “psm3” data)

- 

https://joint-research-centre.ec.europa.eu/photovoltaic-geographical-information-system-pvgis_en

Will

--
You received this message because you are subscribed to the Google Groups "pvlib-python" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pvlib-python...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/pvlib-python/47766714-87d0-4025-aeb5-18c711d26b67n%40googlegroups.com.

[cwh...@sandia.gov]

>  The Linke turbidity stored in the data folder seems constant over every year. Doesn't it vary from year to year?

Yes, it does, for the reason you state. The data distributed with pvlib are multi-year average values meant to be typical. This paper showed that the average values give clear-sky results quite similar to using measurements: https://ieeexplore.ieee.org/abstract/document/8366314

Cheers,

Cliff

[Will Hobbs]

Marius,

With satellite data (e.g., from PVGIS or NREL PSM3), you will not need to model the clear sky GHI with pvlib. Those data sources will include observed GHI, DNI, DHI for a given location and time ("observed" meaning via satellite, which is not as accurate as high quality ground measurements, but still used very commonly in industry). 

One clarification, though. You referenced "current ghi" - the free data sources typically have a time lag of several months or more. For example, I think the NREL PSM3 (technically psm v3.2.2) only has data through the end of 2022 right now. If you want near-real-time observations, you may need to look at commercial data providers. Solcast and SolarAnywhere are two commercial products that have pvlib iotools functions. I think they both have free tiers, but I'm not sure if "live" data are available. 

Will

On Tue, Apr 23, 2024 at 9:21 AM Marius Schmiedt <schmied...@googlemail.com> wrote:

Hi Will,

thanks for your reply. My expectation was that e.g. the clear sky ghi is determined using the models of pvlib (offline) and corrected (with e.g. a factor) to consider weather.

So if I understand you correctly. Instead the ghi for clear sky as well as the current ghi is retrieved from the API and used to determine the PV’s power? 

Best regards 

Marius

Am 23.04.2024 um 14:08 schrieb Will Hobbs <will....@gmail.com>:



[Marius Schmiedt]

Hi Will,

ah ok got it. Actually I thought of predicting the irradiance for the next or next few days. That's why I thought of the clear sky models and using data from a forecast (e.g. due to cloudy conditions the next day) to determine a fraction of the clear sky value.

But I think I'm not able to do this with past data.

Marius

[Marius Schmiedt]

Hi Cliff, 

thanks for the answer. Makes sense to me. What I want is to predict the irradiance for the next few days. I thought of using the cloud cover for this to make an estimation. I saw a forecast class in the 9.5 release to modify the clear sky ghi depending on the cloud cover.

Do you know any model to estimate the ghi weather dependent (or is there maybe one in the current pvlib release I haven't seen).

Best regards 

Marius

[cwh...@sandia.gov]

Marius,

I've seen several authors propose a linear relationship between cloud cover and GHI, i.e., GHI = clear sky x (fraction of sky dome that is clear of clouds).  There was code for these translations in pvlib.forecast, which was deprecated and removed at v0.10.0. You can still find the code if you go to the documentation and select an older version (instead of "stable"). Here's an example https://pvlib-python.readthedocs.io/en/v0.9.0/_modules/pvlib/forecast.html#ForecastModel.cloud_cover_to_irradiance

Best,

Cliff

[Jurand Bien]

Hi all,

As I understood Marius is interested in forecasted data. Even he was mentioned the previous class for forecasting data. In actual weather data there is no irradiance data. In the sources some of you gave irradiance data can be achieved from previous weather data or typical meteorological year. For forecasted data usually cloud cover data is likely to be found in weather forecasts. So based on "Larson et. al. "Day-ahead forecasting of solar power output from photovoltaic plants in the American Southwest" Renewable Energy 91, 11-20 (2016)." a function

cloud_cover_to_irradiance was defined. Using it you can calculate all irradiance elements: ghi, dni and dhi.

Regards

Jurand

To view this discussion on the web visit https://groups.google.com/d/msgid/pvlib-python/f05030b2-caae-457a-b3b3-ab32649b172en%40googlegroups.com.

[Will Hobbs]

For forecasted weather, Open-Meteo.com's website and API, e.g., https://open-meteo.com/en/docs#latitude=51.2&longitude=10.4&hourly=temperature_2m,windspeed_10m,shortwave_radiation,diffuse_radiation,direct_normal_irradiance, might be of interest. Note that it requires a license for commercial use. 

I have been using https://github.com/blaylockbk/Herbie a lot. See https://github.com/blaylockbk/Herbie/issues/182#issuecomment-1792598996 for one example. 

I recently published this project/repo: https://github.com/williamhobbs/solar-fleet-forecast-probability-tool. It is much more complicated than necessary for this question, but the first two notebooks walk through producing solar power forecasts using the NOAA/NCEP HRRR model for a fleet of solar plants. 

Will

[Marius Schmiedt]

Hi Will,

thanks for your reply. My expectation was that e.g. the clear sky ghi is determined using the models of pvlib (offline) and corrected (with e.g. a factor) to consider weather.

So if I understand you correctly. Instead the ghi for clear sky as well as the current ghi is retrieved from the API and used to determine the PV’s power? 

Best regards 

Marius

Am 23.04.2024 um 14:08 schrieb Will Hobbs <will....@gmail.com>:



[Will Hobbs]

Hi Marius,

In my case, I retrieve a GHI forecast for the day-ahead (e.g., on Tuesday, get the forecasted hourly GHI for Wednesday), along with 2m temperature and 10m wind speed, and use that to model power. You could also get the 0-hour-ahead forecast for something like real-time power, but almost all forecast NWP models (NWP = numerical weather prediction) take an hour or more for the weather agencies to run, so there is some lag. 

And whether you want to get GHI directly from the NWP depends on the NWP and your choice. As Cliff and Jurand mentioned, some forecasters use cloud cover and an additional model to estimate GHI. There are also considerations around tie interval length that are NWP-specific. For example, many NWPs have a 3-hour interval, and forecasters will post-process them to interpolate to hourly using a clear-sky model. Open-meteo.com seems to do all of this for you, but you would have to look into their documentation/source code to see what is actually going on. 

And another point for clarification, I don't use clear sky GHI directly for modeling a power forecast. 

Will

[Michael Foley]

Hi Marius,

You can obtain forecast values of GHI, including weather influences, from now out to 14 days, from Solcast via pvlib.iotools.get_solcast_forecast.  There are free trial and commercial options - see solcast.com for more info.

Regards,

Michael

---

Dr Michael Foley

Data Scientist

Solcast, a DNV company