ACER Opinion 12-2025 on the French National Resource Adequacy Assessment

No 12/2025

OPINION

on the differences between the national resource adequacy assessment of France and the 2024 European resource adequacy assessment

15 December 2025

A C E R O P I N I O N N O 1 2 / 2 0 2 5

Executive summary

The European Network of Transmission System Operators for Electricity (ENTSO-E) carry out the European resource adequacy assessments (ERAAs) on an annual basis. To complement ERAA, Member States may also carry out a national resource adequacy assessment (NRAA). If a NRAA identifies adequacy concerns not reflected in the European assessment, the Member State must explain the differences and submit the national assessment to ACER for opinion.

Key findings of the French NRAA

The French NRAA and ERAA 2024 are aligned in their identification of adequacy concerns for all target years except 2030. For the 2030/2031 target year, the national assessment identifies an adequacy concern, with the loss of load expectation reaching 19.8 hours. This exceeds the French reliability standard of 2 hours and reveals an adequacy concern that was not identified in ERAA 2024.

ACER’s assessment of differences

ACER identified several methodological best practices in the French NRAA. • Consideration of the contribution of capacity mechanisms in other Member States. This enables a more realistic view of the state and development of the EU electricity system. • Inclusion of ancillary service revenues. This improves the representation of investors’ business cases by accounting for an important additional revenue stream for flexible resources like batteries. • Enhanced consistency between the investment and adequacy modules. This ensures the overall coherence and robustness of the assessment. ACER finds three unjustified methodological differences between the French NRAA and ERAA 2024. • Risk aversion approach: A share of the highest revenue years is discarded and the possibility for investors to hedge part of their risk through forward markets is not considered. • Price limit increase: The repeated occurrence of price spikes on spot markets results in the increase of their price cap . The French NRAA does not replicate this dynamic leading to an underestimation of potential market revenues. • Flow-based modelling: The French NRAA applies a flow-based allocation for target year 2026/27 but uses a NTC allocation for target years 2028/29, 2030/31 and 2035/36. This does not accurately represent the French cross-border exchanges that do not rely on NTC allocation. Additional analysis conducted: To demonstrate that the adequacy concern identified for 2030/2031 would persist even without the above differences, RTE carried out additional sensitivity analyses. These analyses indicate that the adequacy concern remains, although its magnitude decreases, when the non-justified differences are removed.

Conclusion

Based on the French NRAA and the supplementary analysis provided by RTE, ACER considers the assessment reliable in identifying an adequacy concern for France for the 2030/2031 target year. ACER nonetheless suggests that RTE considers the recommendations made in this Opinion to ensure that the magnitude of the adequacy concern is more accurately captured.

A C E R O P I N I O N N O 1 2 / 2 0 2 5

1. Background

1 Article 20 of Regulation (EU) 2019/943 (Electricity Regulation) requires Member States to monitor resource adequacy within their territory based on the European Resource Adequacy Assessment (ERAA). It also allows them to complement the ERAA by conducting a National Resource Adequacy Assessment (NRAA) according to Article 24 of the Electricity Regulation. Either of the two assessments can identify resource adequacy concerns, but both assessments must be based on the ERAA methodology. The adequacy concerns identified through these assessments should be in the first place addressed by eliminating regulatory distortions or market failures, and, where it does not suffice, by introducing capacity mechanisms.

2 According to Article 24(3) of the Electricity Regulation, where the NRAA identifies an adequacy concern that was not identified in the ERAA, the NRAA must include reasons for the divergence between the two assessments, including details of the sensitivities used and the underlying assumptions. The Member State must publish the NRAA and submit it to ACER for opinion. ACER’s opinion assesses, on a case-by-case basis, whether the differences between the two assessments are justified. ACER primarily considers the differences identified and reasoned by the Member State in the submission, but may also identify further differences, where they have a material impact on the results.

3 The national body that is responsible for the NRAA, which can be a Transmission System Operator (TSO) or a different body designated by the Member State, must take due account the ACER's opinion, and, where necessary, amend its assessment. Where it decides not to take ACER's opinion fully into account, it must publish a report with detailed reasons.

4 The latest ERAA (ERAA 2024) was approved by the ACER Decision on 11 August 2025. To complement it, the Directorate General for Energy and Climate of France (the French Ministry) requested the French TSO, Reseau de Transport d’Electricite (RTE), to conduct a National Resource Adequacy Assessment (French NRAA or NRAA).

5 On 27 October 2025, the French Ministry submitted its NRAA to ACER. The submission includes reasons for its divergence from the ERAA 2024, as well as the details of the sensitivities used and the underlying assumptions.

6 This Opinion follows the submission of the French NRAA and evaluates whether the differences between the French NRAA and the ERAA 2024 are justified.

7 This Opinion is addressed to the French Ministry and issued in accordance with Article 9(2) of Regulation (EU) 2019/942 and Article 24(3) of the Electricity Regulation.

2. Summary of the French NRAA

8 The aim of the French NRAA is to complement the ERAA 2024 by adopting different input assumptions and modelling approaches that more adequately reflect the French electricity system based on historical observations and recent power system evolutions.

9 The French NRAA and ERAA both models 4 target years (2026, 2028, 2030 and 2035). While the French NRAA shows an adequacy concern in all target years, ERAA 2024 shows adequacy concerns for target years 2026, 2028 and 2035 but not for 2030. The results of the central A C E R O P I N I O N N O 1 2 / 2 0 2 5 scenario of the French NRAA hence diverge from ERAA 2024 for the target year 2030, where the estimated LOLE is much higher than in ERAA 2024. This is summarised in Table 1.

10 This divergence in the LOLE results is due to differences in the modelling choices and assumptions between the French NRAA and the ERAA 2024. These differences are discussed and assessed by ACER in the following section.

3. ACER’s assessment

11 ACER welcomes the French NRAA and considers it a well-developed assessment of the French electricity system, enriched with relevant methodological details. The report provides a clear description of the modelling assumptions, the differences to ERAA 2024, and the adequacy results.

12 The differences between the French NRAA and the ERAA 2024 falls into three categories: (i) scope, (ii) inputs, and (iii) method. Each difference is listed in Table 2, along with its impact on LOLE. A detailed description of each difference is provided later in this section. A C E R O P I N I O N N O 1 2 / 2 0 2 5

13 The French NRAA is based on a single modelling tool that differs from the one used in the ERAA. While the ERAA relies on the Plexos model, the French NRAA is developed using the Antares tool. As a result, modelling choices could differ between the two assessments. ACER considers that Antares is an appropriate tool to run a resource adequacy assessment.

3.1. Scope

Difference #1 Regional scope (see Section 2.2.2 in the NRAA)

14 Description: An NRAA is an assessment with a regional scope compared to the pan-European ERAA. In the French NRAA, coverage is highlighted in Figure 1: France in dark green and the other modelled countries in green, some of them being divided into several bidding zones (Denmark, Italy, Norway, Sweden, United Kingdom).

15 Relevance: Reducing the geographical scope may affect the adequacy results because it determines the extent to which cross-border exchanges are taken into account. If the geographical scope is too narrow, the assessment may fail to capture the contribution of neighbouring systems during scarcity hours and therefore underestimate the level of support available to the French electricity system.

16 Assessment: The reduced spatial coverage applied in the French NRAA includes the relevant set of countries, namely France’s direct neighbours as well as other bidding zones that may significantly influence France’s cross-border exchanges. ACER considers that this regional scope is consistent with Article 24 of Regulation (EU) 2019/943 and therefore that the difference compared with ERAA is justified. Difference #2 Target years (see Section 2.1.2 in the NRAA)

17 Description: The French NRAA considers overlapping years (2026/27, 2028/29, 2030/31, and 2035/36) -modelled from July of year N to June of year N+1. By contrast, ERAA uses calendar years (2026, 2028, 2030, and 2035), modelled from January to December.

18 Relevance: Using overlapping years centres the simulation on the winter period, which corresponds to the period during which scarcity risks are most likely to materialise in France.

19 Assessment: ACER considers that centring the assessment horizons around the winter period better reflects the thermo-sensitivity of the French power system and enables a more accurate representation of seasonal constraints. This approach leads to a more realistic assessment of scarcity periods in France and is therefore considered justified. A C E R O P I N I O N N O 1 2 / 2 0 2 5

3.2. Input

Difference #3 Demand and supply hypothesis (see Section 3.2 in the NRAA)

20 Description: The French NRAA applies lower demand trajectories than ERAA 2024, leading to differences in both demand levels and demand profiles. These assumptions reflect downward revisions in expected electrification, notably in the industrial and hydrogen sectors.

21 The generation capacity assumptions used in the French NRAA are broadly consistent with those of ERAA 2024, with slight upward adjustments for solar capacity, downward adjustments for offshore wind capacity, and a partial phase-out of oil and gas units. These adjustments were introduced either to account for the different temporal structure of the target years modelled (see Difference 2) or to reflect the most recent information collected during the public consultation carried out for the French NRAA .

22 Relevance: Demand projections in the French NRAA are lower than those in ERAA 2024 for all target years, with the greatest difference being 34 TWh in 2035 (see Table 3). As a result, demand-supply equilibrium is expected to be met more easily. TWh French NRAA ERAA 2024 Target years 26/27 28/29 30/31 35/36 2026 2028 2030 2035 Domestic 455 479 508 583 483 513 535 617 consumption

23 At the same time, the French NRAA adopts higher supply capacity assumptions than ERAA 2024. It projects higher level of solar capacities, although the gap tends to close for later target years (from +32.5% in 2026/2027 to +4.5% in 2035/2036). The same occur for wind capacities, with assumptions nearly converging in later target years (from +7.8% in 2026/2027 to -0.4%% in 2035/2036). In that case, the slightly lower wind offshore projections are offset by greater wind onshore capacities. Regarding fossil thermal capacities, CCGTs show a slight phase-out from 6.7 GW to 5.9 GW contrary to ERAA 2024 which assumes a constant 6.6 GW capacity. The French NRAA also adopts less conservative assumption for other fossil generation. There is no difference concerning nuclear and hydro inputs capacities. Overall, the French NRAA assumes higher input capacities, as illustrated in Figure 2. A C E R O P I N I O N N O 1 2 / 2 0 2 5

24 Assessment: The French NRAA combines lower demand and higher supply than ERAA 2024. These differences are not expected to lead to an overestimation of adequacy concerns. ACER considers the differences in demand and supply assumptions are justified as they reflect the latest updates of the evolution of the French power system including input from recent public consultation conducted for the purpose of the French NRAA . Difference #4 Monte-Carlo scenarios (see Section 2.2.3 in the NRAA)

25 Description: The French NRAA uses 200 climate years representative of projected climate conditions as of 2025. The climate database is provided by Météo-France, the French meteorological public body. These are combined with 60 thermal power plant and 60 HVDC power line availability patterns, resulting in 1,000 Monte Carlo scenarios. All 1,000 scenarios are used in the Economic Dispatch (ED) module, while a subset of 20 MC scenarios is selected for the Economic Viability Assessment (EVA) module.

26 On the other hand, ERAA 2024 relies on 36 weather scenarios representing future climate conditions and issued from the Pan-European Climate Database (PECD) 4.1. These weather scenarios are combined with 15 thermal and interconnections availability patterns, yielding 540 Monte Carlo scenarios. For the EVA, only a subset of 3 weather scenarios and an average of all outage patterns are used.

27 Relevance: Climate years/weather scenarios capture the inherent uncertainty related to probabilistic events of weather and impact the hourly demand and generation profiles as well as their seasonality. Considering the French power system’s thermo-sensibility, the underlying assumptions of climate years/weather scenarios have a substantial impact on adequacy results.

28 Assessment: The large number of climate years considered in the French NRAA both in the EVA and the ED allow for a broader probabilistic representation of adequacy situations of the French power system. It is expected to enhance the robustness of the assessment and is therefore justified. A C E R O P I N I O N N O 1 2 / 2 0 2 5 Difference #5 FB domains for target year 2026/27 (see Section 3.2.6 in the NRAA)

29 Description: RTE set the flow-based (FB) constraints for the 2026/27 target year using the Porygon methodology. This results in FB domains that differ from those applied in ERAA 2024 but are more closely aligned with the inputs of ERAA 2025. In practice, this leads to lower theoretical maximum import and export net positions for France. RTE provided comparison for the winter domains of the 2026/27 target year (3 in ERAA vs. 2 the French NRAA) as shown in Figure 3. Figure 3: Differences in theoretical maximum import and export net position for France

30 In addition, RTE compared the maximum import observed during scarcity hours for each French border in the French NRAA with those of ERAA 2024, as shown in Figure 4. The comparison shows that, despite the smaller FB domains used in the French NRAA, the maximum energy imported occurring in the French NRAA remain broadly comparable with the one in the ERAA 2024 (maximum imports being 5.2% smaller across all borders).

31 Relevance: The size of FB domains directly determines the cross-border exchange capability available to France. Smaller domains can underestimate import capability during periods of system stress, hence leading to higher adequacy risk outcome. Likewise, they may underestimate export capability, reducing the economic viability of domestic capacity and ultimately contributing to increased adequacy concerns.

32 Assessment: Given that the maximum import capacity during scarcity hours is consistent with the values simulated in ERAA 2024, ACER considers that the French NRAA approach provides a Page 9 of 23 A C E R O P I N I O N N O 1 2 / 2 0 2 5 realistic representation of France’s cross-border exchanges in scarcity hours. The resulting difference in the FB domain is therefore considered justified. Difference #6 Nuclear availability patterns (see Section 2.3.1.4 in the NRAA)

33 Description: Although both assessments consider the same 63 GW French nuclear fleet capacity (existing reactors and the Flamanville EPR) for all target years, the French NRAA assumes broader availability patterns compared to the ERAA. Notably, the French NRAA considers 60 nuclear availability patterns, while the ERAA only considers one. Such assumption aims to reflect uncertainties linked to planned and forced outages, such as maintenance outage durations and delays. This results in higher dispersion of annual nuclear potential generation (ranging from 290 TWh to 450 TWh potential generation pre-simulation).

34 Additionally, and following Electricite de France’s (EDF) revision of nuclear availability assumption, the average nuclear winter availability, fixed at 50 GW in the ERAA, has been revised up to 51 GW for target years 2026/2027 and 2028/2029, and up until 52 GW for later target years. Overall, the updated nuclear availability profiles are consistent with historical trends as RTE illustrated in Figure 5, as well as in France’s Multi Annual Energy Plan.

35 Relevance: Nuclear generation represents a substantial share of power supply in France. Small changes in assumed availability significantly affect adequacy margins, import dependency, and economic viability of resources. In that sense, considering a greater dispersion of nuclear availability patterns is expected to increase the average estimated LOLE, as lower availability patterns are expected to lead to more scarcity situations. The impact on the economic viability of non-nuclear resources less predictable, considering that some availability patterns used in the French NRAA result in greater nuclear generation than in the ERAA.

36 Assessment: The French NRAA improves the representation of nuclear operational constraints such as uncertainties regarding the duration of outages and updates the availability projection using the latest official information. These differences compared to ERAA 2024 enhance the assessment’s robustness; ACER, therefore, considers the difference justified. A C E R O P I N I O N N O 1 2 / 2 0 2 5

3.3. Method

Difference #7 EVA approach (see Section 2.2.6. in the NRAA)

37 Description: The French NRAA EVA evaluates the economic viability of resources using a revenue-based iterative modelling approach applied to capacities located in France. By contrast, ERAA applies a system-wide cost-minimisation approach, in which the optimal capacity mix is determined by minimising total investment and operational costs. To perform the EVA, RTE first runs an Economic Dispatch (ED) with the initial capacity mix over 1 000 Monte Carlo (MC) scenarios. On this basis, a scoring algorithm is applied to select a subset of 20 MC scenarios, which is broader than the subset of 3 weather scenarios used in ERAA . This subset is then used to assess whether market entry candidates, existing, and refurbished thermal, DSR, and batteries units would be economically viable without a capacity mechanism.

38 The scoring algorithm of the French NRAA is based on 3 criteria capturing scarcity events and revenues: the average LOLE, the average yearly revenues of an OCGT, and the average yearly revenues of an OCGT considering risk aversion (see Difference 7). The subset of 20 MC scenarios that performs best against these criteria across all target years is retained for the EVA. At the start of each iteration, each unit’s Net Present Value (NPV) is computed based on simulated revenues across all target years. In case of positive NPV, the asset is evaluated as economically viable, hence enters or remains in the market. Inversely, assets with negative NPV decommission or do not enter the market. Then, starting from the first target year, the least profitable existing capacity is removed from the capacity mix, while the most profitable entry candidate is added. This process is iterated until no further decision arises. Then, the same evaluation is done for the next target year, until all target years are evaluated and a new iteration is launched. The iterative process continues until a stable, viable, capacity mix composition is reached.

39 In ERAA, the EVA is carried out in a single optimisation step using a cost-minimisation objective. Similarly to the French NRAA, the ERAA does perform the EVA with a reduced number of weather scenarios, selecting a subset of 3 weather scenarios out of the 36 of the full set and an average outage pattern. The selection algorithm is based on the Wasserstein distance and the criterion used is CCGT and OCGT net revenues. Based on this selected subset, the overall cost of the system is minimized, starting from an initial capacity mix and adjusting technologies’ capacities and operational dispatch considering their commissioning/decommissioning/mothballing/life-extension potential. The EVA stops when the cost-minimizing capacity mix is reached. To calculate the adequacy metrics, the ED is run with this resulting optimal, cost-minimizing capacity mix on the full set of weather scenarios.

40 Because the French NRAA uses the same dispatch model in both the ED and the EVA and relies on a larger subset of 20 Monte Carlo scenarios explicitly selected to reflect scarcity events and revenues, the EVA provides a more accurate representation of the potential revenues available to investors. This, in turn, results in a high degree of consistency between EVA and ED outcomes in the French NRAA, as illustrated in Figure 6 and Figure 7. Consequently, the French NRAA EVA approach achieves closer alignment between the EVA and ED modules than ERAA 2024 both in terms of adequacy metrics and revenue outcomes. A C E R O P I N I O N N O 1 2 / 2 0 2 5

41 Relevance: Consistency between the EVA and ED is essential to ensure that the investment decisions modelled in the EVA are based on a realistic representation of the revenues ultimately observed in the ED. If the EVA is run on an insufficient or non-representative subset of MC scenarios, it may misrepresent scarcity patterns and revenues, thereby biasing investment signals and, ultimately, adequacy outcomes.

42 Assessment: ACER considers that the iterative, revenue-based EVA approach used in the French NRAA complies with Article 6(2) of the ERAA methodology and constitutes good practice in line with the European Commission’s report on the assessment of possibilities of streamlining and simplifying the process of applying a capacity mechanism. The computational efficiency of the iterative revenue-based design makes it possible to use a larger, representative subset of Monte Carlo scenarios and to avoid additional simplifications of the dispatch module in the EVA, thereby enhancing the internal consistency between the two modules. This is confirmed by the comparison provided by RTE, which shows a strong alignment of LOLE and revenue outcomes between the EVA subset and the full Monte Carlo set (see Figure 6 and Figure 7). ACER therefore considers that the difference in the EVA approach is a methodological improvement compared to ERAA 2024 and, therefore, justified. Difference #8 Risk-aversion approach (see Section 2.3.4. in the NRAA)

43 Description: In the French NRAA, risk aversion is modelled endogenously through an adjustment applied within the net present value calculation of assets when assessing their economic viability (See Difference 6). Instead of determining the net present value on the basis of the full expected revenue across all Monte Carlo scenarios, the French NRAA applies risk-adjusted revenue A C E R O P I N I O N N O 1 2 / 2 0 2 5 directly. For each target year, this measure is derived by excluding the 10 percent of Monte Carlo scenarios with the highest revenues and calculating the expected revenue over the remaining MC scenarios.

44 Description: When performing the EVA, the French NRAA models investor risk aversion using a yearly Conditional Value at Risk with 10 % parameter (CvaR 10%) approach. The method consists of computing annual revenues for each MC years performed in the EVA, and excluding the 10% with the highest average revenues when computing the Net Present Value (NPV) of assessed capacities.

45 This approach does not take into account the possibility for investors to hedge part of their revenue risk in forward markets. It results from the systematic exclusion of the highest revenue MC scenarios in all target years, irrespective of the availability of forward market instruments that could partially mitigate revenue volatility. RTE justifies this approach on the grounds that forward products lack liquidity beyond Y+3, and across all horizons for peak products. Hence, the liquidity of forward products limits their ability to provide meaningful hedging opportunities for existing peaking and all new technologies and therefore limits investors’ reliance on such instruments when evaluating investments.

46 ERAA 2024 represents investor risk aversion through an increase in the Weighted Average Cost of Capital (WACC) via an additional technology-specific hurdle premium. The model first calculates expected revenues under risk-neutral assumptions and then applies the WACC together with the hurdle premium. This approach reflects standard industry practice and is explicitly recognized in the European Commission’s report on the assessment of possibilities of streamlining and simplifying the process of applying a capacity mechanism. According to the ENTSO-E survey conducted in 2025, 20 out of 21 respondents reported using a hurdle-premiumbased approach to account for risk aversion, and half indicated that they also use additional riskaversion tools alongside hurdles rates.

47 The technology-specific hurdle premiums also allow ERAA to implicitly take forward markets into account. These premiums are calibrated to reflect typical investor expectations, including the extent to which forward markets mitigate revenue volatility. For example, investment in new thermal units is evaluated using a higher premium than decommissioning decisions, as a smaller share of risks can be hedged at the time of investment due to the longer time horizon. This approach is consistent with the methodology described in Boudt et al. , which forms the basis for the calculation of hurdle premiums in ERAA. Unlike the approach used in the French NRAA, the hurdle-premium methodology in ERAA does not discard the highest-revenue Monte Carlo years but instead discounts their contribution through the application of a WACC and technologyspecific hurdle premiums.

48 To justify its risk-aversion assumptions, the French TSO provided the responses to question 16 of its public consultation (“Does the hypothesis of risk aversion in anticipating energy market revenues appear grounded to you? If so, do you support its modelling through the use of Conditional Value-at-Risk with a 90% threshold?”). The four stakeholders that provided relevant answers are physical asset owners that may invest in new capacities. They indicate that investors and financiers tend to exclude uncertain, exceptional revenue episodes such as price spikes from their business cases, citing limited forward market liquidity, regulatory interventions and hedging practices. Respondents generally support using a CVaR approach to operationalize this risk aversion, with the 10 percent parameter serving as a reasonable reference, while also noting that sensitivity analyses on the level of this parameter would be advisable.

49 RTE further justifies its treatment of forward markets by referring to one respondent’s feedback that underlines the limited risk-mitigation opportunities provided by forward products. According to this stakeholder, low liquidity, especially beyond Y+4, does not offer sufficient hedging opportunities for future investment revenues. In addition, the respondent argues that forward A C E R O P I N I O N N O 1 2 / 2 0 2 5 market hedging implies a significant revenue discount, limiting the ability to benefit from scarcityrelated revenues. RTE considers that this view is consistent with the one of the French NRA, for example in their quarterly market monitoring report that highlights the lack of liquidity of forward products beyond Y+3.

50 Relevance: The assumed level of risk aversion directly influences investment incentives. Higher risk aversion reduces estimated revenues and may discourage investment, which can result in higher LOLE outcomes. At the same time, forward markets reduce investor exposure to revenue volatility. Where hedging is realistically available and forward markets are sufficiently liquid for some of the assets evaluated, not reflecting this possibility leads to an overestimation of risk aversion, a reduction in the level of economically viable capacity and, ultimately, less realistic adequacy outcomes. In this context, incorporating technology-specific parameters, as done through technology-specific hurdles premiums, is recommended, as revenues patterns and exposure to market uncertainty differ significantly across technologies and may shape how investors assess risk.

51 Assessment (forward markets): ACER considers that the French NRAA provides evidence that not all risks faced by investors can be hedged through forward markets. However, ACER also considers that the portion of risk that can be hedged should nevertheless be taken into account. For this reason, ACER finds that the French NRAA’s assumption that investors do not hedge any risk through forward markets is not justified. ACER recommends that future editions of the French NRAA explicitly consider the possibility for investors to hedge part of their risk through forward markets.

52 Assessment (treatment of high-revenue years): ACER finds the reasoning in the French NRAA (that investors are unlikely to base their investment decisions on revenues arising mainly from Monte Carlo scenarios that are very unlikely to occur) to be reasonable. However, ACER considers the assumption that investors completely disregard such potential revenues to be too restrictive. Exceptional revenues should, at least, be partially considered when modelling investment decisions, as is the case when hurdle premiums are used in ERAA, where highrevenue years are discounted rather than discarded. For this reason, ACER considers that discounting rather than fully excluding the highest revenue MC scenarios is more realistic. ACER therefore recommends that future editions of the French NRAA apply an appropriate discount to the highest-revenue MC scenarios rather than removing them entirely from the calculation of expected revenues.

53 To quantify the impact of the CvaR and reflect the hedging possibility of forward markets, RTE provided a sensitivity analysis using a lower risk aversion parameter. Although RTE argues that the CVaR 5% represent an extreme case of hedging possibilities since it is applied to all capacities and all horizons whereas liquidity of forward products is limited, this sensitivity with less risk averse actors could represent some forward hedging possibilities. Results for 2030/2031 are shown in Table 4: Adequacy outcomes comparison – Reference case vs. CvaR 5% sensitivity. LOLE (h) EENS (GWh) A C E R O P I N I O N N O 1 2 / 2 0 2 5 Reference case – CVaR 19.8 73.8 10% Sensitivity – CvaR 5% 12.5 41.1

54 The sensitivity confirms that lower risk aversion improves adequacy (LOLE falls from 19.8h to 12.5h and EENS from 73.8 GWh to 41.1 GWh. This highlights the importance of the calibration of risk aversion parameters and the proper consideration of forward markets.

55 Nevertheless, as the adequacy concerns remain even under the sensitivity, ACER concludes that this unjustified risk aversion choice, when considered in isolation, does not undermine the overall validity of the French NRAA results.

56 Recommendation: ACER recommends that future French NRAA editions refine the current risk aversion approach using the discounting method as explained above. Difference #9 With-CM approach (see Section 2.3.3. in the NRAA)

57 Description: In the French NRAA, the EVA is performed only for France. For all other Member States, capacity levels are adjusted exogenously so that, where a reliability standard is defined, LOLE levels are equal to that standard. This leads to capacity addition in Member States where, pre-adjustment, the reliability standard is not met and capacity reduction in those where it is exceeded. For the target year 2030/2031, the French NRAA thus identifies around 22 GW additional capacity in Germany, while 15 GW are considered in excess in Italy and 16 GW in Spain.

58 In contrast, ERAA models investment and decommissioning decisions endogenously across all Member States. Investment decisions are based on a pan-European economic viability assessment, allowing capacity to be added or removed wherever it is economically justified.

59 Relevance: Policy choices in other Member States directly influence France’s adequacy outcomes through cross-border exchanges. If capacity is underestimated in a neighboring Member State, its contribution to France’s adequacy will also be underestimated. Conversely, if capacity is overestimated, its contribution will also be overestimated.

60 Assessment: Article 24 of Regulation (EU) 2019/943 requires the NRAA to have a regional scope. This means that the NRAA shall consider the contribution of other Member States through crossborder exchanges. ACER considers that the approach of assuming that other Member States are at their reliability standard is an appropriate approach for a national assessment, as it provides a transparent and realistic representation of their contribution. By restricting endogenous investment decisions to France while externally constraining capacities in neighbouring Member States to their reliability standard, the French NRAA is, in ACER’s view, consistent with the regional scope requirement.

61 In terms of impact, ACER observes that in ERAA 2024, several neighbouring Member States, particularly in the CORE region, exhibit LOLE levels above their reliability standards, indicating undercapacity. As a results, the risk that the French NRAA underestimates the contribution of cross-border exchanges to France’s adequacy is limited.

62 Given the above, ACER considers the French NRAA’s approach where neighbouring Member States’s capacities are adjusted to exactly meet their respective reliability standards to be appropriate. This approach provides a balanced and plausible estimation of cross-border contributions to adequacy and is hence considered justified. A C E R O P I N I O N N O 1 2 / 2 0 2 5

63 Furthermore, ACER notes that it is good practice to dimension capacity mechanisms under the assumption that other Member States will, at a minimum, meet their respective reliability standards. This approach helps to mitigate the risk of over-procurement, where multiple Member States might contract capacity to address the same (regional) adequacy concerns in an uncoordinated manner.

64 Recommendation: For dimensioning a potential capacity mechanism, ACER recommends maintaining the assumption that all other Member States meet their respective reliability standards. Difference #10 Price cap modelling (see Section 3.3.4.4. in the NRAA)

65 Description: The French NRAA assumes a constant electricity market price cap of 4,000 EUR/MWh across all target years. In contrast, ERAA includes the possibility of increases in the price cap, based on the expected evolution of price limits derived from the previous ERAA cycle. ERAA 2024 therefore applies rising caps: 4,500 EUR/MWh in 2026, 5,000 EUR/MWh in 2028, 6,000 EUR/MWh in 2030, and 6,500 EUR/MWh in 2035.

66 Relevance: A lower price cap reduces the potential scarcity revenues, which may discourage investment, thus increasing adequacy concerns. For example, in 2035, the French NRAA assumes a cap of 4,000 EUR/MWh, while ERAA assumes 6,500 EUR/MWh. Under identical scarcity conditions, this implies that scarcity revenues are limited by a cap that is 38.5% lower in the French NRAA.

67 Assessment: The French NRAA justifies its static price limit assumption by stating that market participants may not trust future increases. To support this, RTE refers to the public consultation conducted for the purpose of the French NRAA, in which one respondent argued that there is a risk of regulatory intervention via a possible revision of the methodology on price cap increase. RTE also points out to the ENTSO-E survey in which only one respondent indicated assuming future developments in the price cap over time.

68 ACER notes that the non-consideration of price cap increases is inconsistent with past development, as the market price cap has already risen from 3,000 to 4,000 EUR/MWh. Thus, an assumption that excludes any future increase is not in line with former observations.

69 In addition, Article 10(2) of Regulation (EU) 2019/943 requires a transparent mechanism to adjust bidding limits when the existing limits are expected to be reached. This provision makes clear that price limits are intended to evolve so as not to restrict free price formation. Assuming a fixed cap over the entire horizon effectively ignores this legally foreseen adjustment mechanism.

70 As the French NRAA’s assumption of a fixed price cap does not align with the regulatory framework or with past observed market developments, ACER considers that applying it is not justified.

71 To quantify the impact of the difference, RTE provided a sensitivity analysis using the ERAA 2024 price cap trajectory. Results for 2030/2031 are shown in Table 5. LOLE (h) EENS (GWh) Reference case – Price cap 19.8 73.8 = 4000 EUR/MWh A C E R O P I N I O N N O 1 2 / 2 0 2 5 Sensitivity – ERAA 2024 16.9 60.5 price caps

72 The sensitivity confirms that higher, evolving price caps improve adequacy (LOLE falls from 19.8h to 16.9h and EENS from 73.8 GWh to 60.5 GWh). This highlights the importance of correctly reflecting the regulatory framework and observed market developments.

73 Nevertheless, as adequacy concerns remain even under the sensitivity, ACER concludes that this unjustified modelling choice, when considered in isolation, does not undermine the overall validity of the French NRAA results.

74 Recommendation: ACER recommends that future French NRAA editions consider the price cap mechanism in the model. Difference #11 Cross-border exchanges (see Section 2.3.2. in the NRAA)

75 Description: The French NRAA applies a flow-based approach for the target year 2026/2027 but reverts to a net transfer capacity (NTC) approach for the 2028/2029, 2030/2031 and 2035/2036. RTE justified this by the limited visibility on future Critical Network Elements and Contingencies (CNECs) for these later horizons and by stating that reusing 2026 flow-based domains would be overly conservative and not reflective of expected grid developments.

76 In contrast, ERAA 2024 uses a flow-based methodology for all target years. This represents an evolution from ERAA 2023, which relied on a hybrid approach combining 2025 flow-based domains with ATC-based extensions for later horizons.

77 Relevance: Flow-based market coupling generally allows more efficient cross-border capacity allocation than an NTC approach. Using NTCs may therefore underestimate the contribution of neighbouring Member States during scarcity periods. This may not lead to a realistic estimation of adequacy concerns and inflate the volume required to meet the reliability standard, thereby increasing the risk of over-procurement under a capacity mechanism.

78 Assessment: The Electricity Regulation requires the use of flow-based capacity calculation. The French NRAA is therefore not compliant with this requirement for the target years modelled using NTCs, and ACER considers this deviation not justified.

79 To quantify the impact of the difference, RTE provided additional elements for 2030/2031, the only target year with diverging LOLE outcomes compared to ERAA 2024: • RTE compared maximum imports from the CORE region: 7.6 GW in the French NRAA versus 10.9 GW in ERAA 2024. The French NRAA maximum import value corresponds to the historical NTC value established for BE-FR and DE-FR borders before the before the introduction of flow-based market coupling, augmented by capacity increases from new projects. This comparison thus indicates a difference of around 3 GW in import capacity. • To complement the comparison, RTE performed a sensitivity analysis adding 3 GW of additional import capacity from CORE (1.5 GW on BE-FR, 1.5 GW on DE-FR), bringing import capacity levels closer to the one of ERAA 2024. As shown in Table 6, the impact on adequacy outcomes is limited, with LOLE decreasing by 0.2h for target year 2030/2031. French NRAA Reference case Sensitivity – NTC values A C E R O P I N I O N N O 1 2 / 2 0 2 5 LOLE (h) 19.8 19.6

80 These results indicate that, although the NTC assumption restricts cross-border exchanges compared to ERAA 2024, its quantified impact on French adequacy outcomes in 2030/2031 remain limited. ACER, therefore, concludes that this unjustified modelling choice, when considered in isolation, does not undermine the overall validity of the French NRAA results.

81 Recommendation: Future French NRAA editions adopt a flow-based approach for all target years to ensure compliance with the Electricity Regulation and consistency with the ERAA methodology. This recommendation also applies to the dimensioning and the Maximum Entry Capacity (MEC) calculation for any potential capacity mechanism. ACER considers it essential that capacity-mechanism procurement relies on flow-based modelling. As a mitigation, until flowbased domains can be implemented for all target years, ACER considers that, at least for the closest-to-delivery auction, both the MEC and the volume to be procured should be computed using flow-based modelling. Difference #12 Hydro (see Section 2.3.1.3. in the NRAA)

82 Description: The French NRAA, like ERAA 2024, models hydropower facilities based on their type, installed capacity, and technical and economic parameters. However, it introduces a threestep heuristic reservoir management approach to better estimate the weekly available hydropower generation within the Antares optimisation window.

83 This heuristic is based on input inflow data and the shape of the residual load as follows: • First, a monthly pre-allocation is performed in proportion to residual load to reflect the seasonality of the French power system. • Second, daily generation targets are derived from the monthly pre-allocation, accounting for reservoir constraints, and then aggregated into weekly generation targets. • Third, hydropower dispatch is optimized at the hourly level in the ED based on costminimization, subject to the weekly generation constraints determined in the second step. The application of this heuristic constrains hydropower generation profiles so that they more closely follow expected residual load patterns, thereby ensuring consistency with historically observed reservoir management behaviour.

84 Relevance: Hydropower plays a central role in adequacy assessments because of its ability to shift energy across hours, days, and seasons. The modelling approach used affects the operational flexibility of hydropower units, their ability to provide energy during scarcity, and the consistency of dispatch with historical practice and environmental constraints. Differences in hydropower modelling can affect LOLE outcome, particularly in systems with significant reservoirs and pumped-storage capacity such as France.

85 Assessment: By introducing a residual-load-based pre-allocation heuristic, the French NRAA ensures that hydropower generation patterns remain consistent with historical reservoir management practices. This provides a conservative yet credible estimation of hydropower’s contribution to adequacy. ACER therefore considers the difference justified. Difference #13 Alternative revenues (see Section 2.2.6. in the NRAA)

86 Description: The French NRAA EVA explicitly accounts for additional revenues from ancillary services such as balancing capacities provision. These revenues are included when estimating A C E R O P I N I O N N O 1 2 / 2 0 2 5 capacities’ NPV, thereby influencing investment and retirement decisions. The calibration of ancillary revenues is performed through two complementary approaches carried out ex-ante.

87 First, a historical method relies on historical remuneration patterns observed for different technologies, notably CCGTs, OCGTs, and demand response. Second, a forward-looking, prospective approach models both future ancillary services prices and the volumes of capacities expected to be engaged by 2030. In this prospective model, both demand and supply of ancillary services are explicitly modelled, allowing the optimisation of the participation of FCR/aFRR certified capacities in face of the ancillary services requirements, accounting for their associated costs.

88 By contrast, ERAA does not include ancillary services revenues in its EVA. To ensure that the operational availability of assets reflects their obligations to provide frequency containment and restoration reserves, two modelling approaches are available in ERAA. It either reserves a share of thermal, renewables, or hydro capacity for the provision of system services, or specifies the reserve requirement which could be covered by thermal units. The choice between these two approaches depends on the way individual TSOs provide national data for the ERAA assessment.

89 Relevance: Ancillary services revenues can materially impact the economic viability of certain technologies, especially flexible assets with fast activation capability such as OCGTs, batteries, and hydropower units. The inclusion of those revenues increases the expected income of those capacities, reducing their likelihood of decommissioning, and thereby mitigating adequacy concerns.

90 Assessment: The inclusion of expected revenues from other electricity-related services is explicitly required by Article 6(9)(b) of the ERAA methodology. ACER considers that in this respect, the RTE approach constitutes good practice as it provides a more complete representation of actual market incentives faced by investors in France. The difference with ERAA 2024 is therefore considered justified.

3.4. Additional analysis on difference 8. and 10.

91 As outlined in Section 3.3, some of the differences linked to the modelling of revenues are considered not justified, namely the static price cap, the lack of consideration of forward markets, and the discarding of the revenues from the highest-revenue MC scenarios rather than their discounting. For the unjustified differences, the French NRAA provided sensitivity analyses measuring their impact on adequacy outcomes. These sensitivities highlight that, when considered in isolation, LOLE levels are reduced but remain higher than the French reliability standard. It is, however, necessary to assess whether this would still hold if all unjustified differences were corrected simultaneously.

92 In this context, during the opinion process, ACER, the French Ministry and RTE agreed that RTE would run an additional scenario to demonstrate whether the adequacy concern would still materialize. The parameters of this additional scenario run by RTE are as follows: • Lower the risk aversion parameter from 10% to 5% in order to reflect the hedging possibilities through forward markets; • Price caps aligned with the increasing limits derived in ERAA 2024; • Discounting rather than discarding exceptional revenues A C E R O P I N I O N N O 1 2 / 2 0 2 5

93 Table 7 summarizes the results of the sensitivity analyses done for difference 8. and 10. in isolation and the combined sensitivity with the total impact. Revenues CVaR LOLE EENS Price caps above CVaR parameter (h) (GWh) parameter 4000 Reference case 10% Discarded 19.8 73.8 EUR/MWh CVaR 5% sensitivity 4000 5% Discarded 12.5 41.1 EUR/MWh (see Difference 8) Price cap sensitivity ERAA 2024 10% Discarded 16.9 60.5 (see Difference 10) Combined scenario ERAA 2024 5% Discounted 6.5 16.7

94 The results of this scenario confirm the significant impact of parameters related to revenues on adequacy outcomes, with LOLE levels decreasing from 19.8 hours in the reference case to 6.5 hours. Nevertheless, LOLE remains above the French reliability standard of 2 hours. Accordingly, this additional scenario, while more favorable to investment in capacity, does not alter the conclusion of the French NRAA that indicates insufficient capacity levels in the 2030/2031 target year. ACER considers that this combined scenario would be an appropriate Central Reference Scenario to identify adequacy concerns.

95 ACER further notes that this additional combined scenario does not account for the increased NTC capacities included in the cross-border exchanges sensitivity (See Difference 11). While the impact of these additional NTC capacities between France and the CORE region is expected to be limited (given that the LOLE decreased by only 0.2h in the sensitivity), ACER considers their inclusion represents best practice to assess the combined impact of all unjustified differences on adequacy outcomes. A C E R O P I N I O N N O 1 2 / 2 0 2 5

4. Conclusions

96 ACER has assessed the differences between the French NRAA and ERAA 2024 as listed in Table 8. The combination of these differences leads to a divergence in the adequacy results between the two assessments.

97 ACER considers most of the differences between the French NRAA and the ERAA 2024 edition to be justified. However, ACER finds three differences, #8, #10 and #11, not justified. The previous section explains ACER’s position and offers recommendations for improvements.

98 Nonetheless, as confirmed by the sensitivities provided by the French TSO, the overall impact of these differences on adequacy results (LOLE) does not alter the conclusion of the French NRAA. ACER suggests that the French authorities consider the recommendations made in this Opinion to ensure that the magnitude of the adequacy concern is more accurately captured.

99 Considering the above, ACER is of the view that overall, the French NRAA identifies adequacy concern in a robust manner. ACER also recognises that the revenues-based modelling approach, the assumption that other Member States will, at a minimum, meet their respective reliability standards, and the consideration of alternative revenues are good practices. A C E R O P I N I O N N O 1 2 / 2 0 2 5 This Opinion is addressed to the French Ministry of Energy and Climate - Direction Generale de l’Energie et du Climat (DGEC). Done at Ljubljana, on 15 December 2025. — SIGNED — V. ZULEGER, ACER Director ad interim