How to Apply Subhourly Clipping Loss With Hourly Meteo Data

[kjs55]

Hello,

Would you please point me to the PVsyst v8 Help Documentation for how to apply the subhourly clipping loss when all you have is hourly meteorological (aka solar resource) data for a given site (which is almost always the case)?

I was just reading this PVsyst Help Documentation <https://www.pvsyst.com/help/physical-models-used/grid-inverter/subhourly-clipping-correction.html>, and I noticed this sentence: "The only prerequisite to apply the model is to have a MET file that has been generated using sub-hourly irradiance data."

Are you aware that this "only" prerequisite is almost never available, in the standard PVsyst modeling use case? Therefore, I think I must be missing or misreading something. I'm hoping you can please explain this to me here and point me to the relevant PVsyst Help Documentation that I'm presently unable to easily find.

Thank you.

[Linda Thoren]

Dear kjs55,

Thank you for your input. PVsyst is not a weather provider, but a simulation software. Several data providers supply sub-hourly values that you are welcome to import as a custom file and apply the sub-hourly clipping correction. 

Kind regards

[Li-ESS]

On 2/4/2025 at 1:03 PM, Linda Thoren said:

Dear kjs55,

Thank you for your input. PVsyst is not a weather provider, but a simulation software. Several data providers supply sub-hourly values that you are welcome to import as a custom file and apply the sub-hourly clipping correction. 

Kind regards

Hi Linda,

a general question, can PVsyst 8 also generate sub-hourly energy generation time series, if the custom file is also sub-hourly data? 

Kind regards

 

[kjs55]

@Linda Thoren Scientific literature (e.g., Clean Power Research CPR, UL, Townsend, PVsyst, several others) indicates that synthetic 1-min. satellite data exhibit only ~60% of natural radiation's variability as compared to ground-based data, thus defeating the purpose of performing subhourly corrections. For example, below I include an excerpt from a PVsyst publication on this point [1]. Furthermore, it remains unclear how to then accurately use subhourly correction results derived from a 1-min. dataset, obtained via PVsyst V8, together with the ultimate hourly, typical meteorological year TMY dataset employed ubiquitously in the simulation used for project finance. Thank you.

4.3 Synthetic data

The correction model relies on sub-hourly data, ideally 1-minute data, or at least a sampling frequency that matches the variations in power of the DC array. Arguably the most reliable source for high frequency data are ground measurements. However, in practice, it can be difficult to obtain this data, especially during preliminary design phases for the PV system. One possibility is to rely on a synthetic generation procedure. This type of approach allows to generate sub-hourly fluctuations from a lower sampling frequency input. For example, one can use typical statistics for sub-hourly data to complement satellite data, which is found generally in 15-minute time steps. The result of this procedure is a high frequency time series that matches the original measurements up to a certain time scale (e.g. 15 minutes for satellite data). We refer the reader to [8], e.g., for a more thorough discussion on this subject.

Since the clipping correction can utilize any sub-hourly time series, as a proof of concept, we use here the hourly averages for the data obtained from Table 1 and utilize such a synthetic procedure to generate a 1-minute time series. The results are summarized in Figure 13, compared against those from Figure 11, where the original time series was used.

From these preliminary results we can extrapolate that the synthetic generation methodologies may induce further biases. Indeed, in the example we have studied, the clipping correction extracted from the synthetic data is lower for all four sites and for all DC:AC ratios.

<see attached image>

Figure 13: Comparison of the yearly sub-hourly clipping correction, using the measured 1-minute data from the four sites shown in Table 1. Synthetic data was obtained by first summarizing the 1-minute data and then using a synthetic generation algorithm to recover a 1-minute time series. The synthetic generation does not always fully reproduce the statistical characteristics of the irradiance fluctuations, and in this case the corresponding sub-hourly clipping correction is under-reported.

[1] M. Oliosi, B. Wittmer, A. Mermoud, A. Bridel-Bertomeu, and R. Vincent, “Implementation of a Sub-Hourly Clipping Correction in PVsyst.” WIP-Munich, 2024. doi: 10.4229/EUPVSEC2024/4DV.4.7.