In a Virtual Power Plant, thousands of systems provide balancing services every day. But how exactly are the individual units selected for the provision of balancing energy? Which factors are taken into consideration?
The VPP algorithm decides, which technical units to include in the provision of balancing energy, taking into account individual technical parameters of each power plant, such as:
- Reaction time to signals from the VPP operator for changing power generation resp. power consumption (this also results in the selection of the fitting products of balancing services that the unit subsequently will bid into)
- Ramping speed of changing power generation or power consumption
- (Flexible) capacity of the plant
- Size of the gas storage or heat storage tank, if applicable
- Geographical location of the plant
Apart from these parameters, the VPP algorithm takes into account restrictions given by the owner of the technical unit, who can communicate that the power plant is not available due to planned maintenance or unplanned failure of the system. If there is an explicit deregistration for the corresponding day, the plant will, of course, not participate in the activation.
Having analysed these parameters, the VPP algorithm decides which assets are chosen to provide balancing services and form a common response from the Virtual Power Plant. From this pool, the amount of flexibility that can be offered in the individual balancing energy auctions is calculated. Among other things, this pool is also formed using data from the operation of the individual plants and their availability during the previous days.
Composition of the Pool
Contractual, economic and technical aspects play an equal role in the composition of the pool response from available plants.
The balancing power products are interdependent both in terms of their timing and the amount of potential revenue, so the composition of the balancing power quantity is created in correlation to the balancing power products. Next Kraftwerke anticipates the market before the auctions take place and derives the quantities for the individual auctions. One possible approach would be to first compile the plants for the FCR, then for the aFRR and in the last step for the mFRR in order to finally know the quantities that can be bid in the auctions for each balancing power product. If the aFRR auction is awarded for a smaller quantity than the quantity of aFRR-eligible installations in the pool, the "surplus" installations could be transferred to the mFRR pool, for instance.
In order to enable detailed planning for the provision, a forecast for the coming week is made from the current operation mode and past operational schedules. Various factors such as weather or electricity market forecasts are taken into account.
Of course, redundancies are also planned for the provision of balancing services in order to compensate for short-term downtimes of individual units and to minimise any possible bottlenecks.
Activation of Ancillary Services by the VPP Algorithm
If the Virtual Power Plant receives the signal from the transmission system operator to deliver balancing power, the algorithm of the VPP selects the appropriate plants from the pool. Technical and economic criteria play a role in this selection. Both the energy costs of a plant and the technical requirements are included in the activation process. Additionally, the VPP algorithm includes a randomizer, which has to guarantee a fair distribution of technical units within the pool. The calls are distributed statistically across the individual installations so that the same installations are not always called up.
If an activation is communicated by the transmission system operator, the required quantity of flexible reserves is distributed among several units, even if a single unit could also provide the required quantity on its own. If a large unit were to fail to deliver, the effort to compensate for this failure would be greater than if a small unit failed. For this reason, it is also more desirable for several units to run at partial load than for one unit to be switched off completely (when negative balancing power is called up). The reasons are easy to understand: On one hand, the full start of a unit takes longer than if one unit had to ramp up from partial load to full load again. On the other hand, when a unit is switched off completely, there is a greater risk of complications that could be occurring during the re-start. During the activation, the quality of the service is constantly monitored. If a unit does not meet the set target value, further units are activated to provide the desired amount of balancing energy.
In short: Since there is a high number and diversity of units in the virtual power plant, the underlying algorithms try to form this heterogeneity into a coordinated swarm with precise control commands and detailed planning data.