Sagging of the UCTE mains frequency to 49.84 HZ

On 28.03.2023 from 16:00 (UTC), the [grid radar monitoring network](/en/wide-area-monitoring- system) recorded a strong and long-lasting underfrequency event in the UCTE network area. Several unusual phenomena occurred:

Background

Recently, we have been observing more frequently the departure from the +/- 100 mHz region around 50 Hz. This is certainly partly due to the changeover from the winter to the summer timetable. However, the deviations differ from what we observed in previous years. While in the past strong and short frequency deviations around hourly or quarter-hourly breaks could be observed, in the meantime longer lasting frequency deviations above 100 mHz over several minutes up to hours occur frequently.

The situation on 28.03.2023

On 28.03.2023, such a situation could be observed again. This time, however, the strength of the deviation and the duration of the deviation above 100 mHz is surprising. In the case of an underfrequency, generation power is missing from the grid. The frequency then drops faster or slower depending on the generation shortfall and available instantaneous reserve. In order to raise the frequency back into the deadband of 50 Hz +/- 10 mHz, the use of primary and especially secondary control power (FCR and aFRR) is necessary.

Frequenzverlauf des Unterfrequenzereignisses am 28.03.2023 um 1656 Uhr (UTC) Figure 1: Course of the mains frequency

As Figure 1 shows, there was a deviation to below 49.84 Hz at about 16:56 (UTC) in the observation period between 16:00 and 18:15 (UTC). The length of the underfrequency is striking, with a few exceptions of just under two hours between 16:20 and 18:10 (UTC). The intermediate frequency rise to over 50 Hz at shortly after 17:00 hrs can probably be explained solely by the activated aFRR as a result of the preceding event with the frequency drop to 49.84 Hz. This is because as soon as the maximum is reached at 17:07 (UTC), the frequency drops again relatively quickly below the controller deadband. This indicates that the retrieved aFRR was deactivated again.

The long duration of the frequency restoration to above 49.90 Hz until around 6:10 p.m. is probably due to a lack of marketed power plants in the overall system. Balancing group managers are obliged to replace missing power in the portfolio with a purchase on the spot market at the latest at the hourly change, in most countries already at the quarter-hourly change. Obviously, this has not been done here, or not sufficiently.

Looking in particular at the time window between 17:07 and 17:43, another problem becomes visible: In this period of almost half an hour, there were two quarter-hour changes. According to market rules, the lead time would have been sufficient twice to procure shortfalls via intraday. However, it is also strange that until 17:43, again no aFRR was activated to a sufficient extent. It is true that the control power activation apparently stabilises the grid frequency around 49.95 Hz in the quarter hour between 17:15 and 17:30. However, it oscillates more strongly again afterwards and only moves upwards again after 17:50. From a control point of view, there can be several causes for this observation, which we can only speculate about here. For example, several events could have caused the underfrequency. Therefore, there could have been a misjudgement of the expected power deficit in the corresponding control zone(s) or the control of two LFC controllers could have influenced each other.

Regional consideration

Against this background, the question arises as to the country of origin of the situation. Transparency ENTSO-E unfortunately does not allow any conclusions: Neither large unplanned line outages nor large unplanned power plant outages are published in the period under review. Only an unplanned power reduction between Portugal and Spain is indicated. Closer analyses of the grid frequency in Spain do not reveal any significant effects on the grid frequency in the overall system during this period.

Regionaler Frequenzvergleich des 49,84 Hz Ereignisses am 28.02.2023 Figure 2: Regional frequency comparison

If we look at the grid frequency at different locations in the grid (Figure 2), we see a very high overlap of the respective measured frequency curves -- even around the low point. This indicates that there was obviously very little grid-related interference in the centre of the continental European synchronous area. Only at the edges, in Malaga or on Crete, there were larger fluctuations. However, this corresponds to the usual frequency behaviour in Europe.

Phasenwinkeldifferenz am 28.03.2023 während und nach dem Erreichen der 49.84 Hz Referenz. Die Spreizung des Phasenwinkels zeigt das folgende Leistungsungleichgewicht der Reaktion auf das Ereignis Figure 3: Course of the normalised phase angle differences

More information is provided by looking at the normalised phase angle differences. These are sorted in ascending order in Figure 3 with Leszno in Poland as the reference measuring station.

Explanation of the phase angle difference: The phase angle difference makes it possible to draw conclusions about the power difference at two points in the network if the measurements are precisely time-synchronised. For this purpose, the measuring instruments used are GNSS time- synchronised. The greater the phase angle difference, the greater the power difference between the two locations. For stationarity reasons, we do not look at the power difference per se, but at its change in relation to the starting time of each series of locations.

The disturbance and the resulting power imbalance become visible in the phase angle. If the phase angle differences were quite close together until approx. 16:57, i.e. approx. 2 minutes after the low point, they subsequently diverge. A closer look reveals the west-east gradient between the measuring stations both before and after the frequency low point. This indicates that the net power in the eastern grid area was already lower than in western Europe from 16:00. After the frequency event, there was a massive aFRR activation, which caused the frequency to rise again. However, it appears that the additional power came mainly from Western Europe, presumably France. This is because it is there that the phase angle difference increases most rapidly and also permanently until 18:00. As with the grid split on 08 January 2021, France again presumably took the leading role in stabilising the system.

The situation in Germany

How did Germany support the system? For this purpose, we looked at the TSO's grid frequency published every second and the aFRR setpoint published every second.

Explanation of the German aFRR Setpoint: The recorded grid frequency and the aFRR setpoint are made available on netztransparenz.de by the German TSOs on the following day. The aFRR setpoint represents the activation quantity of aFRR deemed necessary by the LFC controllers every second. However, significant adjustments of the value only take place in the 4-second interval (cf. Hirth & Mühlenpfodt, 2021). When using both data series, it should be noted that the published data are shifted by four hours. The frequency low point is therefore incorrectly found at 20:55 (UTC) in the data. The aFRR setpoint is determined on the basis of the German Area Correction Error (ACE). According to ENTSO-E Operation Handbook, Policy 1, the ACE takes into account both the national power deviation and the (Continental European) grid frequency.

According to the System Operation Guideline, TSOs are obliged to regulate the power frequency of their control area. Therefore, the aFRR setpoint in comparison with the measured frequency provides an indication of whether the cause of a frequency deviation is to be sought within Germany or outside Germany.

Von ÜNB gemessene Netzfrequenz vs. SRL-Sollwertvorgabe in Deutschland Figure 4: Frequency response and aFRR setpoint published on netztransparenz.de

Figure 4 compares the two (time corrected) series of the grid frequency and the aFRR setpoint. Essentially, the setpoint opposes the identified frequency deviation. Expected deviations only occur near the controller deadband. At 16:56, however, the (positive) aFRR setpoint is relatively low. This indicates that the cause of the frequency drop should not be sought in Germany. The phase after 17:07, however, suggests at least an involvement of the deviation cause within Germany. This is because from about 17:30 the setpoint rises from below 500 MW to over 1200 MW.

While the aFRR setpoint is a second-by-second indicator of the aFRR power to be activated, Figure 5 shows the actual activated aFRR in the period under consideration.

Die von deutschen ÜNB gemessene Netzfrequenz gegenüber dem aFRR Abruf in Deutschland Figure 5: Activated German aFRR

Until 16:15, the German aFRR supports the system. In the following quarter of an hour, the frequency then drops massively out of the controller deadband, while Germany still has almost 385 MW negatively activated. This phenomenon of regulating against the system is unfortunately not uncommon. We will soon publish another article in which we systematically analyse this problem.

In the following quarter hours, we then see again how the German control power counteracts the frequency deviation. As with the analysis of the aFRR setpoint, the actual activated control power also shows that the sharp drop in frequency before 17:00 is countered with lower German control power use. However, the long underfrequency phase after 17:00 leads to higher aFRR calls up to the extremely rarely called almost 1000 MW positive aFRR in the quarter hour from 17:45. Ultimately, this power surge in the German control block coupled with the 18:00 hourly break was able to bring the system back into the target range.

Summary

It remains to be said that the unusual frequency curve on 28.03.2023 is probably due to several causes. The strong frequency drop to 49.84 Hz probably originated outside Germany. The long underfrequency phase after 17:00, on the other hand, can be explained by more than one event, whereby this underfrequency must have been caused at least partially in Germany.

Status: 31 March 2023 22:00 hrs