10 January 2023
DARE – Dynamic daily dimensioning of reserve needs
To keep the frequency at 50 Hz, the Elia Group needs to dispose of balancing services, which is also referred to as reserve capacity or frequency restoration reserve (FRR).
Why this project
Before 2020, reserve capacity needs were statically determined by Elia, i.e. once per year, and approved by the national regulatory authority. In the project DARE “Dynamic Application for Reserves”, a software tool has been developed to calculate on daily basis the reserve capacity needs for the next day. This calculation is based on the expected imbalance risks in the system allows to better fit the reserve capacity to the risks in the system.
The methodology has been developed by Market Development in cooperation with the external consultant N-SIDE. It is implemented via a user-friendly software tool in cooperation with the system operators (National Control Center).
To calculate the FRR needs for the next day, a Machine Learning approach is implemented. In a first step, historical data on system imbalances and system conditions are used to train each month an algorithm to be able to predict the risk of facing system imbalances for the next day. This includes the risk following renewable prediction errors, as well as the risk of facing the forced outage of a power plant or HVDC interconnector. In the second step, this algorithm is used each day before 7 AM in order to determine the reserve capacity needs for the next day. For each block of four hours of the next day, the capacity to be procured is published on the website and tendered by Elia.
OperationsSince February 2020, every morning (before 7 AM), the trained machine learning model is run and the results are validated by the system operator in order to determine the balancing capacity for each block of 4 hours of the next day. These volumes are thereafter published and tendered (if needed).
ValidationThe results are verified and validated before publication by the system operator.
Fall-back OptionsIn case the data are not available to make reliable forecasts, fall-back options have been put in place to ensure reliable operations.