Greenbyte Documentation

Aggregating data over time

The signal documentation describes how the signals are calculated for a single time step. To view the data on other time intervals, the data needs to be aggregated. Following are the two different ways this can be done:

Normal

For normal signals, aggregation is just an average or a sum of the value for each time step within the requested interval. For example, the average Availability for a month would simply be the average Availability of all time steps during the month. How this is done when the device and source resolutions don’t match up exactly is described on the Resolution conversion.

Resolution-dependent

For resolution-dependent signals or resolution signals, the calculation is made for the specified resolution. If data is requested for a month, the calculation will be done for that month.

A signal will be a resolution signal if:

For example, the formula for Performance Index is:

But the Performance Index is not calculated per time step, but for the full month with n time steps the formula looks like the following:

Photo-voltaic solar panels degrade in performance over time. This can be approximated by an annual degradation factor. The following formula is used in potential power calculations to calculate the effective degradation factor for a given timestamp, given the commissioning date and degradation factor.

D_Y is the metadata value (yearly) Degradation Factor set per site.

t_M(t,u) gives the absolute month difference between two timestamps.

t₀ is the Commissioning Date metadata value set per inverter.

The device resolution controls the resolution of the time series data for the device. Real-time data from the device will be aggregated to time steps according to the device resolution.

A device resolution of 10 min is the same as 1/6 hours.

For example, transforming Energy to Power: E = P / 6.

When the requested target resolution for data is different from the device resolution, a Resolution conversion will be made.

In some calculations, we make sure figures are defined, where we use 0 (zero) if undefined. This is marked as x^z and calculated as:

Helper signals are used as input to other signals, but are not available in the signal selector in Greenbyte and can therefore not be visualized on their own.

Irradiance data is often unreliable because of cheap sensors, night-time moonlight, etc. It is not uncommon to get false irradiance values around 0.

Therefore, it is advised to set a threshold to use in the calculations with irradiance cut-in, disregarding all irradiance values below the threshold.

I is the irradiance for each timestep.

I₀ is set as the Availability Irradiance Cut-In metadata value per inverter (W/m²)

The positive part of a value x is noted x⁺ and is calculated as:

When you look at data in a different resolution than the device resolution, the data is converted to the requested resolution. This resolution conversion is always done before any aggregation of data between different devices is performed.

When averaging signals, the aggregation for a target interval is done as a weighted average:

V is the set of values overlapping with the target interval I

W_v is the weight of the value V, which is the fraction of the value's interval overlapping with the target interval.

When summing signals, the aggregation is done as a weighted sum:

Example data conversion from 15-minute resolution to 10-minute resolution

A turbine has the following 15-minute wind speed data:

The conversion to 10-minute data is done using the weighted average.

Example data conversion from 10-minute resolution to 15-minute resolution

A turbine has the following 10-minute energy export data:

The conversion to 15-minute data is done using the weighted sum.

Example data conversion from 10-minute resolution to 5-minute resolution

A turbine has the following 10-minute wind speed data:

The conversion is done using the weighted average. However, since the device resolution is divisible by the target resolution, this results in each value being repeated.

Signals that are "resolution-dependent" are calculated for the specified resolution. If data is requested for a month, the calculation is done for that month.

For example, the formula for Performance Index is:

But the Performance Index is not calculated per time step. It is calculated for the full month with n time steps. Therefore, the formula is:

For most signals, the values will be calculated for every time step in the device resolution. In contrast, resolution signals are signals where the values will be calculated in the presentation resolution. See more details on Aggregating data.

A time step refers to a period based on the device’s resolution. Normally for wind turbine time series data is aggregated to 10-minute periods. But for other devices, it could be something else, for example, 5-minute, 15-minutes, hourly or daily data.