The needs for balancing resources i.e. Frequency Restoration Reserves (FRR) in the Belgium Load Frequency Control (LFC) block can be covered by different means, such as (a) contracted balancing capacity, (b) non-contracted balancing energy bids (known as free bids), and (c) sharing of reserves. ELIA is investigating the possibility to reduce, for a given balancing capacity product and a given 4-hour time block, the procurement of balancing capacity, and the corresponding procurement cost by predicting the available volume of free bids and reserve sharing.
Thus, the main goal of the project was to develop, test, and validate a method that forecasts in D-1, i.e. before the balancing capacity procurement, the volume of non-contracted balancing energy bids that will be available in ELIA’s LFC block. This forecast was developed, separately, for automated (aFRR) and manual (mFRR) reserves and for both upward and downward directions. The developed model is also used for computing available mFRR sharing (via Available Transfer Capacity (ATC) after intraday).
Figure: Fit of the forecasting tool within the procurement time-line
The methodology followed during the project is aligned with industry standards, namely, a first step where data is collected, cleaned, and transformed, continued by a feature importance analysis, then designing proper performance indicators for the models, as well as preparing an adequate learning environment. Finally, the models are evaluated and assessed using the KPIs agreed between ELIA and N-SIDE.
Our models are able to forecast, on average, a volume of non-contracted balancing means (including reserve sharing) of almost 1940 MW for mFRR downward and 500 MW for mFRR upward, with a 99% reliability. Although several regression models were tested, the most performant one was a quantile regression based on random forests.
Read all the details of the project in the public consultation published by ELIA: "Study on the daily prediction of non-contracted balancing energy bids", 22 December 2021. Available at: https://www.elia.be/en/public-consultation/20211001_public-consultation-on-the-daily-prediction-of-non-contracted-balancing-energy-bids
Would you like to know more about this project with TSO Elia?
Alberte holds a double master's degree in Energy Engineering from KTH (Sweden) and KU Leuven (Belgium) with a specialization in Smart Grids. He works on applications of mathematical programming and machine learning within electric power system operations, such as grid expansion planning or sizing of balancing capacity.Alberte Bouso