Summary and outlook of the study
A fault-induced system separation of the interconnected system poses a challenge for maintaining frequency stability in the continental European interconnected system, as the separation causes the previous transits via the three-phase grid to occur in the form of power imbalances in the separated sub-grids. Using suitable scenarios, it was possible to show that very high power imbalances can potentially be expected in the future, which, in combination with a reduced flywheel mass, will lead to significantly higher and critical frequency gradients. Since such disturbances, which go beyond the design case but are still possible, must not lead to a system collapse, the current behavior and necessary countermeasures were analyzed for both the overfrequency and underfrequency ranges.
The overfrequency power reduction (OF-PR) of the generation units is used to control disturbances in the overfrequency range that go beyond the design case. Since the high transients in Germany in the north-south direction are caused in particular by the high feed-in from wind turbines, the focus was on the LR of wind turbines. On the one hand, the curtailable output of the LF-LR must be sufficient to absolutely compensate for the power imbalance. On the other hand, the curtailment must be sufficiently fast so that the frequency can be kept within the permissible frequency band. It has been shown that the performance of the frequency stabilizers currently used in wind turbines is not sufficient to maintain frequency stability for the scenarios under consideration. Delays for measurement and communication of up to one second on the one hand and the time response of the UF-LR on the other play a significant role. It was shown that scenarios with a power imbalance of up to 30 % in relation to the grid load and frequency gradients of up to 2 Hz/s could be controlled, provided that these delays could be limited to 100 ms and this frequency gradient is controlled by the inverter control. In order to be able to control scenarios with higher power imbalances and frequency gradients, the time response of the frequency converter control would also have to be significantly faster. The speed of the power reduction is limited in the pitch control concept typically used today for the ÜF-LR in order to protect the system from mechanical loads. Without extending existing concepts, optimization of the time response is therefore likely to reach its limits.
For this reason, the contribution of braking resistors to optimizing the time response of the ÜF-LR was also examined. It was shown that there is potential to significantly optimize the time response, but a large number of systems would have to provide a contribution to optimizing the time response with braking resistors or comparable concepts, which is not yet state of the art. In addition, the contribution of artificial flywheel masses at overfrequency was analyzed. Here, too, it was shown that the frequency behavior can be positively influenced.
The results clearly show that in future, the technical connection rules should define specific requirements for the time response of the LRFs for all generation units. However, for the reasons described, the focus should be on avoiding grid disconnection of synchronous generation. For this reason, non-synchronous generation will essentially have to contribute to ensuring frequency stability through sufficient time response of the LR UF.
Underfrequency load shedding (UF-LA) is used to control disturbances in the underfrequency range that go beyond the design case. The disconnectable load of the UF-LA must be sufficient to absolutely compensate for the power imbalance and the load must be switched off quickly enough to prevent unwanted overreaction. It has been shown that scenarios with a power imbalance of up to 20 % in relation to the grid load and frequency gradients of up to 2 Hz/s can be controlled with the current load shedding in 10 stages and a delay of 150 ms. In order to be able to control scenarios with higher power imbalances and frequency gradients, the delay would have to be reduced or the current load shedding concept would have to be supplemented with additional functionalities. Therefore, the UF-LA was extended by a Δf/Δt functionality. (The real delay of the frequency-dependent load shedding can also be around 200 ms. In this case, a Δf/Δt functionality is already required for frequency gradients from around 1 Hz/s).
Under the assumed boundary conditions, it could be shown that scenarios with a power imbalance of up to 40 % in relation to the grid load and frequency gradients of up to 4 Hz/s could be mastered. Another problem with the UF-LR is that the frequency can "get stuck" at frequencies below 49 Hz under certain boundary conditions despite the UF-LA functioning as designed. A possible solution to this problem was identified by adding delayed load shedding to the UF-LA if the frequency is below 49 Hz.
The present investigations show necessary measures to control high power imbalances in combination with a low inertia due to grid disconnections at over- and underfrequency. However, the investigations are based on a summary grid model and are therefore limited to the power-frequency behavior. Further aspects must therefore be investigated in order to ensure that sub-grids formed as a result of grid disconnections can stabilize themselves. The following aspects should therefore be the subject of further investigations. Today, inverter-based generation units typically use current injection methods and require, for example, a PLL (phased locked loop) or similar methods to synchronize the inverter control with the grid voltage. These methods require sufficient short-circuit power or a sufficient proportion of voltage-inducing equipment in order to ensure robust operation and thus a stable supply. If a sub-grid with a very low proportion of synchronous machines (with or without a high proportion of inverter-based generation) is formed as a result of a grid disconnection, stable operation of such a disconnected sub-grid is a prerequisite for it to be able to safely stabilize itself as a result of a grid disconnection. It is therefore necessary to analyze where the limits of the current current-injecting control methods used today lie and, if necessary, derive a minimum necessary proportion of voltage-injecting grid elements. With a view to long-term development, it should also be examined what contribution the use of voltage-inducing methods for converter-based generation units and loads can make both to the stable operation of systems in weak grids and to grid support in dynamic processes following grid disturbances, in particular for the provision of instantaneous reserve.
In addition, it would appear to make sense to define reference scenarios for network disconnection scenarios to be controlled for the overfrequency and underfrequency range in the future. Based on these reference scenarios, requirements for the automatic final measures can then be defined and checked as part of the system protection plan (defense plan). As grid disconnections typically occur across national borders, these reference scenarios must be coordinated at ENTSO-E level for the entire continental European interconnected system. Specifically, the first step would be to define the minimum size above which a disconnected sub-grid must be safely intercepted. Based on this, the maximum power imbalance to be controlled and the maximum average frequency gradient to be controlled would also have to be defined. If necessary, it makes sense to define this reference disturbance (according to the predicted expected boundary conditions) in stages for different time horizons, i.e. to tighten it. Based on the analyses carried out, it seems sensible to manage scenarios with 2 Hz/s and power imbalances of ≥ 40% (in relation to the remaining grid load) in the long term.