Measuring the volume of demand-side flexibility in the Belgian day-ahead market

Summary

Elia commissioned N-SIDE to conduct an in-depth analysis of the Belgian wholesale day-ahead market to estimate the amount of demand-side flexibility, or ‘Market Response Volume’ (MRV), already present today. Given Belgium’s portfolio-based approach, this involved a detailed examination of bids submitted by market participants, with particular focus on complex orders. This volume is used as an input in the calculation of the target volume to be procured under the Belgian Capacity Remuneration Mechanism (CRM).

 More concretely, in the scope of this project, we looked for:

• Demand Shedding: A reduction of the load when electricity prices reach a certain level. It can include behind-the-meter generation and/or storage.
• Demand Shifting: Load reduced at a specific period when electricity prices reached a certain level and recovered before or after.

While an existing methodology was in place, the work of N-SIDE evolved around three different axes:

• We reviewed the existing approach and ensured its consistent application,
• We limited the risk of double-counting from merged order books and aggregated curves, and
• We tackled the problem from a different angle based on clustering to validate and improve the calculations.

By refining the methodology and calculation, Elia has gained confidence in the volume of existing demand-side response. This validated value was needed for calibrating the central scenario of the most recent Adequacy & Flexibility Study for Belgium. Furthermore, the methodology used to process these bids also informs the calculation of the strike price for the CRM Auctions for the 2026-2027 and 2029-2030 delivery periods.

Customer Challenge

Belgium’s state introduced a Capacity Mechanism to guarantee sufficient electricity supply. The initial auction for this mechanism was held in October 2021, with the delivery period commencing in November 2025.

To build the scenarios which estimate the demand procured through the CRM, one of the parameters used is the volume of all existing demand-side response active in the Belgian wholesale markets: the Market Response Volume. Although multiple definitions co-exist of what exactly demand response is, we follow the definition of Mathieu et al. (2024): 

Demand response is the actions of customer-sited energy resources downstream of metering points to voluntarily, actively, and temporarily adjust their electricity production and/or consumption in response to signals (e.g., commands, prices, measurements).

Day-ahead electricity markets can fluctuate significantly due to price-inelastic demand, scarcity events, and renewable variability. Demand response helps by smoothing prices, lowering costs, and reducing the risk of involuntary disconnections. 

The main difficulty of the study lay in accurately measuring the volume, given that Belgium, like several other central European nations, utilizes a portfolio-based approach for expressing bids in wholesale markets. This bidding strategy enables participants to combine all their assets into a single, adaptable bid, in contrast to unit-based bidding, which mandates separate bid submissions for every individual physical power plant.

The Solution/Approach

In order to categorize the existing volumes into market response or not, we built a framework that can be summarized as follows:

Figure 1: N-SIDE. (2025). Market Response Volume.

The most significant change in the methodology with respect to the previous approach was the introduction of a clustering-based method, which replaced the existing price-based filter.

Figure 2: N-SIDE. (2025). Market Response Volume.

While this methodology proved effective for complex bids, which allow for the derivation of multiple features, the majority of the volume in wholesale markets involves simple bids. This latter format is less informative than complex bids. For this purpose, we developed a complementary method: a pattern finder algorithm.

The rationale is that some demand-side actors are not affected by fluctuating fundamental conditions (e.g. fuel prices). The assumption is that such actors would place bids with a higher frequency than usual at the same price, and potentially the same volumes as opposed to one-off behaviours. We developed a customized algorithm that analyzes all simple bids and extracts patterns in bidding behaviour. 

The pattern-finding algorithm’s scope was expanded in 2025. Initially, in the first version of the study, it was only applied to simple bids priced below the marginal cost of a high-cost thermal unit (such as Open Cycle Gas Turbines). This adjustment was made to the entire spectrum of simple bids to achieve a more reliable measurement of the demand-side response volume, rather than just an upper bound.

Distribution of MRV volume from fixed-price patterns in winter 2024-25

Figure 3: N-SIDE. (2025). Market Response Volume.

The clustering-based approach brought three improvements to the exercise:

1. Decouple from gas prices. Our approach allowed us to determine a volume which is less influenced by the prices of gas in wholesale markets.
2. Confidence. The analytical method allows to determine a lower and upper bound of demand side response in day-ahead markets.
3. Transparency. We can explain the final volumes of MRV as an addition of volumes coming from individual, smaller clusters rather than a unique volume. As an example, we illustrate below the classification of simple bids in five different clusters, out of which two are categorized as MRV:

Figure 4: N-SIDE. (2025). Market Response Volume.

Benefits

The results delivered by N-SIDE had two immediate impacts for Elia:

1. The volume of existing DSR used to calibrate the central scenario of the latest Adequacy & Flexibility Study for Belgium. Our computed values are a direct input to the base case, which is publicly available in Public consultation on the methodology, the basis data and scenarios used for the study regarding the adequacy and flexibility needs of the Belgian power system for the period 2026-2036

Figure 5: Elia. (2024). Adequacy and Flexibility Study 2026–2036

2. The so-called strike price for the CRM Auctions for the delivery period 2026-2027 and 2029-2030. To harmonize processes, Elia used our consistent approach to the treatment of bids (parsing and collapsing) to determine the price limits as documented in WG Adequacy – 13 October 2025. For more details, the process to calculate the price based on the aggregated curves is detailed in Strike price calibration methodology.

Strike price calibration – methodology reminder Elia. (2024). Working Group Adequacy Meeting Documents

1. Gathering information on submitted DA bids (both demand/supply, simple/complex) based on N-SIDE input, which is aligned with the methodology outlined in this case study.
2. Create a single aggregated curve for each peak hour (8-20) on winter working days
3. Take the average aggregated curve for each winter
4. Create a weighted average curve for the past three winters, weights are the total average volume offered in each curve
5. Normalize the final curve, and define a 75% and 85% of total offered volume

Figure 6: Elia. (2024). Working Group Adequacy Meeting Documents

Conclusion

This collaboration provides Elia with a consistent method for processing input data from the Power Exchanges, which will be valuable for future Flexibility and Adequacy studies.

The methodology can be used to formulate past and existing demand response active in wholesale markets. The framework will continue to be used by Elia for calculating the demand curve of the CRM, the strike price, and additional input for flexibility tracking. 

• A complete version of our study can be found in Ad hoc WG 12th July (2024) and an updated version in  WG Adequacy 28th August (2025)