Sub-hourly clipping correction

Because the PVsyst simulation is in hourly steps, inverter clipping is primarily evaluated on the basis of the hourly average IV curves of DC arrays. However, in reality, sub-hourly fluctuations in irradiance may push the maximum power point of a DC array temporarily below or above the inverter clipping threshold. This sub-hourly behavior leads to extra losses, which are not captured by simply using the hourly IV curve.

PVsyst offers a model to evaluate these supplementary sub-hourly clipping losses. Starting with PVsyst v8.0.0, the weather import allows one to leverage sub-hourly data to extract relevant statistical information, which is then used in the simulation for a more accurate clipping loss evaluation.

How to apply the model

The only prerequisite to apply the model is to have a MET file that has been generated using sub-hourly irradiance data. Currently, these MET files can be generated using the custom file import. A checkbox allows one to choose to generate a compatible MET file.

The compatible MET file will then be marked as containing the statistics necessary to use the clipping correction.

Once the MET file exists, it can be used alongside any variant. The clipping correction will be automatically used whenever it is applicable.

Results

The sub-hourly clipping correction loss can be found among the simulation variables, and in the loss diagram.

It is represented among the simulation variables by "Sub-hourly inverter clipping correction". The variable tag is ILPmxSH. It is included in the Invert category for inverter losses.

In the loss diagram, the extra loss is grouped with IL_Pmax, the inverter loss over nominal power.

Model explanation

The model is based on two main observations:

1. The hourly clipping loss is always inferior to the sub-hourly clipping loss, and can always be evaluated via the surface above or below the clipping threshold in the MPP time series plot.
2. Most of the PV production processes do not depend on the ordering of time steps, i.e. the minutes in an hour can be reordered by increasing irradiance.

Based on these two points, it is possible to easily approximate the missing loss using few coefficients for each hour, extracted from the sub-hourly data.

The details of the model have been reported in a publication¹.

Missing clipping loss evaluation

There are two cases, where the hourly clipping loss evaluation is incorrect.

First, when the hourly average MPP is below the clipping threshold, but this threshold is passed at some times within the hour.

Second, when the hourly average MPP is above the clipping threshold, but this threshold is not reached at some times within the hour.

In the two figures above, the red line represents the equivalent clipping level in terms of irradiance, the black line represents the average hourly irradiance, and the orange area characterizes the missing clipping loss. In these two cases, the missing loss is characterized by the area above or below the threshold.

Reordering of minutes

Most of the physical processes in the simulation can be well represented by an instantaneous model. For example, the PV production primarily depends on the instantaneous irradiance. Therefore, the minutes within an hour can be rearranged without too much impact.

A counterexample is the temperature of the modules, which due to thermal inertia effects, should depend on previous time steps. However, in terms of PV production, temperature is secondary to the irradiance dependence.

Approximation

Once the minutes have been reordered, it is possible to approximate the missed clipping loss by various means. We choose to represent the irradiance reordered evolution with two linear segments (in green in the figure below), above and below the average irradiance (horizontal black line).

The clipping threshold (red line) can then be applied on the approximated sub-hourly irradiance reordered evolution. This allows to simply approximate the clipping correction.

The minimum information to be stored at each hour is, therefore, the maximum GHI irradiance, the minimum GHI irradiance, and the number of minutes above or below the threshold.

While the approximation for each hour may be rudimentary, it is more solid on the scale of a full simulation. The mean bias error in the clipping loss evaluation is indeed strongly suppressed. In the figure below, we compare the clipping evaluation without correction (top) to the one with correction (bottom).