An advanced reliability reserve incentivizes flexibility investments while safeguarding the electricity market Authors: Franziska Klaucke, Karsten Neuhoff, Alexander Roth, Wolf-Peter Schill, Leon Stolle Published: 2026 ABSTRACT SummaryTo ensure security of supply in the power sector, many countries are considering capacity mechanisms. Simultaneously, the expansion of variable renewables increases the need for power sector flexibility, with promising options on the demand side. We analyze how a centralized capacity market and an advanced reliability reserve with a moderately high activation price affect demand-side flexibility investments in a German 2030 case study. Using a capacity expansion model with detailed demand-side flexibility potentials, we show that a centralized capacity market caps wholesale market prices, thereby reducing incentives for flexibility investments. The reliability reserve induces seven times higher flexibility investments in energy-intensive industries and twice as high investments in district heating storage, while achieving security of supply at similar electricity supply costs. The reliability reserve could create a learning environment for flexibility technologies to support the transition to climate neutrality, and policymakers should consider it in forthcoming capacity mechanism decisions.Subject areas: Energy engineering, Energy management, Energy modeling, Business, Economics Introduction The decarbonization of the German energy system largely relies on the use of renewable energy sources.1,2,3 While wind and solar photovoltaics (PV) provide emission-free electricity, their variable generation patterns present a challenge to the electricity system, in which supply and demand must match at any time. With rising shares of variable renewables, flexibility in the electricity system becomes increasingly important to cope with resulting imbalances.4 The necessary flexibility can be provided by both the supply and the demand side. Examples include dispatchable power plants, load shifting by industrial consumers and households, electricity storage, heat storage, or intermediary product storage, as well as cross-border electricity trade.5,6,7,8 Among various flexibility technologies, demand-side options have emerged as promising, yet often under-utilized.9 By shifting load from hours with little renewable generation to hours with high renewable availability, electricity costs of consumers can be reduced. If load shifting lowers the need for expensive peak capacity, it may also incur overall system cost savings.5 Further potential advantages include reduced use of fossil fuels, respective energy imports and carbon emissions,10 and a potentially lower need for grid expansion.11 Germany provides an interesting case study, as it plans to massively increase variable renewable electricity generation and its use across different sectors.12 It currently has an energy only market, which is complemented with different types of reserves to guarantee security of supply in exceptional situations when the wholesale market does not clear.13 This includes a small capacity reserve, which has never been activated so far.14 Yet, the growing importance of renewable energy sources, and therefore the increased variability of electricity supply, has recently reignited a policy debate on security of supply. The previous federal German government (2021–2025) considered different policy options for capacity mechanisms,15 and the current government plans to follow up on this.16 The discussion in Germany focuses on the need for and the design of capacity mechanisms, which remunerate the provision of generation capacity to ensure that sufficient dispatchable capacity is available when needed.17 While traditionally designed to support conventional thermal power plants, capacity mechanisms must now be reassessed in light of their potential interactions with demand-side flexibility, which has often been neglected in the literature. In this article, we analyze how two main types of capacity mechanisms, capacity markets and reserves, interact with demand-side flexibility technologies. To do so, we use an open-source capacity expansion model of the German electricity sector that includes various demand-side flexibility options. Under textbook assumptions, prices in a perfect energy only market set adequate incentives not only for efficient dispatch of all generation and flexibility technologies, but also for sufficient capacity investments to ensure resource adequacy.18 However, in reality there are several barriers that prevent sufficient investment, especially in technologies that serve peak loads, which may only run for a few hours per year and thus rely on high scarcity prices to refinance their investment costs. First, investors in such power plants may be unable to recover their fixed and investment costs due to price caps set by the regulator to protect consumers from scarcity prices. These would decrease the revenues of electricity generators, leading to a “missing money” problem.19 Second, even if there was sufficient revenue, investors might not be able to finance their investment due to incomplete risk trading, or “missing markets” for long-term risk allocation.20,21 Third, politically induced rising shares of variable renewable energy generation lead to increasing wholesale price uncertainty. This may exacerbate concerns about insufficient investment in firm capacities and security of supply.22 As a solution to these challenges, different designs of capacity mechanisms have been discussed and implemented in various electricity markets around the world.23 Their goal is to provide sufficient long-term investment signals by replacing uncertain revenues from scarcity pricing with constant capacity payments.17 Two design options that have been widely implemented so far are centralized capacity markets and capacity reserves. There are also less common capacity mechanisms, such as France’s decentralized capacity market,24 from which we abstract in the following, as it is scheduled to be replaced by a centralized capacity market in November 2026.25 In a centralized capacity market, the regulator determines the necessary capacity to ensure a desired level of security of supply and contracts this capacity via auctions. The winning bidders then receive a constant capacity payment in addition to the revenues they generate from selling electricity on the wholesale market.26 The concept of a capacity reserve, also referred to as strategic reserve, is fundamentally different. Here, the regulator contracts generation capacity outside the wholesale market and exclusively maintains it in a reserve. This reserve capacity is allowed to produce only when a dispatch criterion is reached, for example when the market does not clear or when a certain price level, typically a very high one, is exceeded.27 In the following, we focus on a particular type of such a capacity reserve with a moderately high activation price, which we refer to as an “advanced reliability reserve.”28 The verdict on which type of capacity mechanism is preferable, however, is far from conclusive. Employing an agent-based model calibrated to the German electricity market, Keles et al.29 find that both a centralized capacity market and a capacity reserve can provide security of supply. Despite the capacity market being prone to overcapacity, it is found to be the cheaper design option in the long term. Höschle et al.,30 modeling capacity expansion as a non-cooperative game for the Belgian market, find that both mechanisms reduce total consumer costs compared to an energy only market. The centralized capacity market is economically more efficient, as capacity in the reserve cannot participate in the balancing market, which is also needed for security of supply. Weiss et al.,31 comparing market designs in a 100% renewable Israeli electricity system, conversely find that both a reserve and capacity market increase security of supply, but that the capacity market does so at higher costs due to overcapacity. Jimenez et al.32 find that capacity reserves, centralized and decentralized capacity markets can improve security of supply compared to an energy-only market, in a decarbonized electricity market of the Netherlands. Capacity markets do so at lower costs while also reducing price volatility relative to the capacity reserve. Previous analyses, however, have mostly not explicitly addressed the interactions between different types of capacity mechanisms and demand-side flexibility. Potential reasons for this may be that the availability of techno-economic data on flexibility options is limited,33 and that the aspect of flexibility only had limited relevance in legacy power systems dominated by dispatchable thermal generators. Keles et al.29 distinguish between exogenous scenarios with high or low availability of sheddable load and apply it to an energy only market, but not to capacity mechanisms. Weiss et al.31 restrict demand-side flexibility to battery storage, which they assume to be the only technology contracted in the strategic reserve. Jimenez et al.32 consider different consumer groups with different values of lost load able to provide demand response, as well as shedding of industrial heat load. They find that a strategic reserve activates more demand response, as the reserve plants are only activated at an activation price of 4,000 euro/MWhel, which lies above the costs of demand response. However, the authors do not consider investment costs for demand-side flexibility. Rious et al.34 calculate potential revenues of demand response operators in France, but do not model the market equilibrium. Demand-side flexibility can also be considered as a substitute for thermal peak generation capacity29,34,35 and can thus contribute to security of supply.36,37 Capacity markets, however, discriminate against demand-side flexibility via multiple channels. Firstly, pre-qualification procedures for capacity markets disadvantage flexibility technologies,38 excluding them from obtaining capacity payments through capacity mechanisms.39 For example, different types of storage can be operated for different durations before they run empty, depending on their storage capacity and the filling level. As a consequence, complex de-rating calculations are required to make their contribution to security of supply comparable to firm generation capacity.39 Secondly, capacity markets distort investment decisions by creating a bias for peak generation technologies with low fixed costs but high operating costs.33 This occurs as technologies with high operating costs refinance a larger share of their costs through scarcity prices. In a capacity market, this revenue stream is replaced by capacity payments and spot market prices are less volatile.40 As, however, exploiting this price volatility is the main source of revenue for flexibility technologies, a capacity market reduces their profitability.41 Thirdly, capacity markets distort bilateral contracting and risk trading.28 In this article, we first gather the potentials and techno-economic parameters for the most relevant demand response potentials in Germany, including various industrial processes and district heating. We then include this portfolio of demand-side flexibility options in a capacity expansion model of the German electricity market. We apply the model to a stylized 2030 scenario with a high share of renewable generation, comparing a centralized capacity market and an advanced reliability reserve which is activated if wholesale electricity prices reach 500 euro/MWhel. Importantly, we assume that demand-side flexibility technologies are neither included in the capacity market nor in the reliability reserve, as they could not compete with firm generators on a level playing field in either of these capacity mechanisms. We optimize both the investments in different types of demand-side flexibility and their hourly utilization. For clarity, we focus on a Germany-only setting to illustrate the workings of the two capacity mechanisms, without complicating the numerical analysis by including other countries and cross-border electricity flows. Both the capacity expansion model and the input data are described in the STAR Methods section. Our contribution to the literature is thus twofold: First, we evaluate the effects of capacity mechanism design on investment and utilization of demand response and storage technologies, based on a comprehensive review of their potential and costs in Germany. Second, we aim to advance the literature on the design of capacity reserves and add to the comparative analysis of different capacity mechanisms. Results Overall power plant portfolio The optimal capacity portfolio, resulting from the first step of the optimization, differs between the two capacity mechanisms (Figure 1). In the advanced reliability reserve, the installed capacity that directly participates in the wholesale market, namely open-cycle gas turbines (OCGTs), is significantly lower than in the centralized capacity market. Instead, extensive generation capacities (35 GWel) are held in the reserve and do not participate in the wholesale market. Installed firm capacity by technology for the capacity market and the advanced reliability reserve While the capacity portfolio of the reliability reserve is not part of the optimization, we assume that it consists of newly built OCGTs, which have low specific investment costs. In reality, however, the reserve could also absorb existing plants that leave the wholesale market. This has already been the case for the existing capacity reserve in Germany. Impacts on wholesale electricity prices The main difference between a centralized capacity market and an advanced reliability reserve is the ability of the capacities covered by these mechanisms to participate in the wholesale electricity market. While power plants that take part in a centralized capacity market can fully operate on the wholesale electricity market, power plants contained in the reserve cannot, but are only activated by the regulator in case the wholesale market price reaches the activation price. This difference has profound consequences for the price formation in the wholesale market. We illustrate this with price-duration curves for several weather years, which sort all hourly prices of a year in descending order, starting with the highest price on the left-hand side (Figure 2). Wholesale market price-duration curves for both capacity mechanism scenarios and different weather years (2008–2014) Depending on the weather year, peak prices differ greatly between a capacity market and a reliability reserve (left, Figure 2) for a maximum of around 400 h. In a centralized capacity market, there is always sufficient generation capacity from gas-fired power plants during these hours, and their marginal costs determine the price in all peak hours. The resulting maximum price is around 130 euro/MWhel, which corresponds to the marginal production costs of a gas turbine in the model, considering fuel and carbon prices as well as other variable costs. In contrast, wholesale prices gradually rise to 500 euro/MWhel in some hours in the scenario with an advanced reliability reserve. The electricity price never exceeds 500 euro/MWhel, as this price triggers the activation of the reserve, which has always sufficient capacity to serve demand. During most other hours of the year, the results for the centralized capacity market and the reliability reserve hardly differ (right, Figure 2). On the right-hand side of the price-duration curve, the differences between weather years turn out to be substantially greater than those between the two policy scenarios. Investments in flexibility technologies In line with the different wholesale price patterns emerging in the two scenarios, potential revenues for flexibility technologies greatly differ. Hence, there are also different incentives for respective investments in the second step of the optimization. With its flat peak prices, the centralized capacity market leads to substantially lower investments in different demand-side technologies in the electricity market compared to the advanced reliability reserve (Figure 3). The centralized capacity market leads to such a high capacity of firm power plants that prices only rise to the variable costs of the most expensive plant. This reduces potential revenues of flexibility technologies, which are thus crowded out on the wholesale market. Storage capacities for enabling demand-side flexibility for both capacity mechanism scenarios In contrast, investments into flexibility technologies are much higher if an advanced reliability reserve complements the electricity market. Model results show that investments into storage for flexibilizing processes in the energy-intensive industry are nearly seven times higher than under a capacity market. Investments into process thermal storage increase by around a quarter. In the case of district heating energy storage, the reliability reserve leads to more than twice as much investment. As discussed below, the precise model outcomes hinge on the weather year. However, the qualitative effects remain robust also using different weather years (Figure S2). Activation of the advanced reliability reserve In each year, the reliability reserve is only activated for very few hours. Depending on the weather year, it is used between 0 and 64 h, as shown in Figure 4. And even when the reserve is activated, it is mostly used at less than half of its capacity. In the seven weather years modeled here, the reserve is activated for only 175 h (0.29% of all hours). Operating hours of the advanced reliability reserve in different weather years Considering the extremely rare activation of the reserve in only a few hours with very high output, it seems plausible that the reserve could in reality be smaller than modeled here. A similar reasoning applies for the centralized capacity market. Ultimately, a weather situation with cold temperatures and low wind and solar power generation, as observed on a December day in 2009, leads to such a high remaining demand for electricity that the reserve must be dimensioned almost twice as large as would be necessary for all other 2,554 days covered in the analysis. Average electricity supply costs Figure 5 shows average annual wholesale prices for all modeled weather years, as well as the required payments for the capacity mechanism, allocated to the total electricity demand as a levy per megawatt hour. We define the sum of the two as average supply costs of electricity. Note that we do not consider other price components of end user tariffs here, such as grid fees or taxes. Average supply costs depend primarily on the assumed fuel prices, which we hold constant in our analysis, and should thus not be interpreted as policy-relevant forecasts for 2030. Average wholesale electricity prices and levies for capacity mechanisms in different weather years (2008–2014) We first compare capacity payment differences between the two mechanisms. The respective cost allocation for the capacity market is 5.9 euro/MWhel, and varies between 1.8 and 2.0 euro/MWhel for the reserve. The main reason for the higher levy in the capacity market scenario is that all power plants receive a payment that reflects the annualized investment costs plus annual fixed costs of OCGTs, which are the marginal power plants that receive the capacity payment. As the capacity portfolio does not differ between weather years, also this capacity payment is constant across weather years. In the case of the advanced reliability reserve, the capacity payments only finance the power plants in the reserve. In contrast to the capacity market, the reserve plants also get compensation for their variable costs through the capacity payments, and not only for their capacity costs, when they are activated. The levy is the sum of capacity and activation payments for all plants in the reserve, minus the revenues from selling the respective amount of electricity at the activation price. Accordingly, the levy slightly varies across years and is somewhat lower in weather years in which the reserve is activated more often (e.g., 2012). Average wholesale electricity prices are slightly lower in the capacity market scenario than in the scenario with a reliability reserve, as prices in the former are effectively capped at the marginal costs of open-cycle gas plants. Yet, differences between weather years are larger than differences between the two capacity mechanisms. Combining average wholesale prices and levies, overall electricity supply costs hardly differ between the two capacity mechanisms, averaged across the weather years 2008 until 2014. In the current model calibration, the capacity market leads to slightly lower average supply costs of 76.4 euro/MWhel than the reliability reserve at 77.6 euro/MWhel. Considering a range of uncertainties in the model parameterization, this cost difference of less than 2% can be interpreted such that neither of the two options has a clear social welfare advantage in the static setting modeled here. That is, the economic inefficiencies introduced by both capacity mechanisms largely balance each other in our model parameterization: the capacity market discriminates against flexibility technologies, and the advanced reliability reserve leads by design to prices above the costs of the marginal power plant when it is activated. Discussion Long-term outlook beyond 2030 The analysis is conducted in a static setting for the year 2030, based on policy-relevant assumptions on electricity demand and renewable generation shares. However, this 2030 perspective neglects the fact that the use of variable wind and solar power is planned to increase significantly in Germany in the longer run, which is in line with the country’s 2045 climate neutrality target. Accordingly, the system benefits of demand-side flexibility are likely to increase further over time, and the benefits of early investments in such flexibility are thus underestimated in our analysis that implicitly assumes a 2030 long-run equilibrium. As the reliability reserve unlocks much more demand-side flexibility, it is likely to become also more advantageous over time compared to a centralized capacity market, which crowds out demand-side flexibility via an oversupply of dispatchable power plants. This crowding out of flexible assets through firm power plants, which would be locked into the system for a long time, is likely to incur increasing efficiency losses over time. Furthermore, the modeled flexibility potentials can be considered a lower bound estimate for a long-run decarbonized energy system beyond the year 2030: for example, the model does not yet consider flexibility potentials in households and commerce, as well as flexibility related to electric vehicles, which are all likely to become more relevant beyond 2030. Significant demand-side flexibility potential can still be tapped here in the future, particularly against the backdrop of advancing electrification and sector coupling. The role of firm generation assets could thus decrease if different types of storage and other flexibility technologies become more widespread after 2030. In turn, this could impact the optimal dimensioning of all types of capacity mechanisms. Advanced reliability reserve design considerations and policy implications The practical implementation of a reliability reserve in Germany would require more detailed work on the dimensioning, procurement, and activation routines. However, we argue that the practical implementation of the reserve appears to have further advantages compared to a centralized capacity market, beyond the flexibility benefits quantified in our model analysis. In Germany, an advanced reliability reserve could be implemented quickly. The regulator already has substantial experience with the tendering and operation of different types of capacity reserves.13 In an initial phase, the reliability reserve could even consist of power plants that are already part of existing reserves or that will be decommissioned in the coming years for economic reasons. A central design characteristic of the reliability reserve is determining its size. While centralized capacity markets are prone to oversizing,31,32 for example because of high levels of political or regulatory risk aversion, this is likely also true for reliability reserves. Yet, the reserve appears to be significantly more adaptable and less irreversible than a centralized capacity market. This is particularly true as decisions on the size of the reserve do not have a direct impact on supply and demand and thus on wholesale market price trends, and ultimately on the incentives for investments in flexibility and generation capacity in the electricity market. The adaptable size of the reserve could further emerge as a great advantage when the optimal reserve dimensioning might vary over time because of ongoing technological developments. Three energy market characteristics likely decrease the reserve requirements: first, an increasing price elasticity of electricity demand in addition to the specifically modeled investments in demand-side flexibility could help to reduce the size of the reserve. Second, cross-border European electricity trade could help to reduce the reserve capacity, similarly to its mitigating effect on storage requirements.7 Third, there may be further potential to negotiate additional contracts with large consumers to reduce their peak loads for the very few peak hours in which the reserve would be activated. Another important design parameter of the reliability reserve is its activation price. We propose 500 euro/MWhel to consider the trade-off between regulatory credibility of maintaining this price, and sufficient investment incentives via wholesale market prices. In other words, the activation price has to be low enough to be credible, and high enough to provide investment incentives. So far, real-world reserves have often been only activated when the electricity market does not clear, which implies extremely high activation prices. In Germany, the upper reference price limit on the day-ahead market is currently at 4,000 euro/MWhel,42 which is also used as an activation price for the reserve by Jimenez et al.32 However, we argue that such a high activation price might not be credible because the regulator would be pressured to mitigate such high prices, especially when they persist over a longer time, by activating the reserve already at lower prices. For example, the 2025 German coalition agreement includes a statement that existing reserve plants could be dispatched with the aim of lowering wholesale market prices.16 We argue that such interventions are unlikely to happen in case of a moderately high activation price of 500 euro/MWhel and thus consider this price to be “low enough.” At the same time, such an activation price appears to be “high enough” to provide sufficient incentives for investments into new generation capacity and demand-side flexibility. We consider it unlikely that investors of power plants or storage options would rely on much higher peak prices than 500 euro/MWhel in their investment decisions. In contrast, activating the reserve at substantially lower prices, e.g., at 150 euro/MWhel,43 may deteriorate investment incentives in the wholesale market. In a sensitivity analysis, we explore the effects of a 20% lower or higher activation price (see section S3.2). While all major results are qualitatively robust, there are a few, and mostly minor, quantitative changes. For example, a lower activation price leads to a slightly larger reserve (Figure S3), minor changes in price-duration curves (Figure S4), somewhat lower investments in flexibility technologies (Figure S5), increased operation hours of the reserve (Figure S6), and slightly higher average electricity supply costs (Figure S7). For a higher activation price, the opposite is true. Importantly, regulators would have to ensure that the advanced reliability reserve is filled via a non-discriminatory competitive auction that is open to both existing plants and new builds. This way, a challenge of some previous capacity reserves could be avoided, where existing plants sometimes could bargain to be included. Finally, maintaining a large reserve capacity which is only used very rarely to the full extent might raise inefficiency concerns in the public debate. Policy makers or regulators would thus have to clearly communicate the workings of the reliability reserve and its potential benefits over alternative options. This includes not only providing supply security but also fostering flexibility technologies and ultimately strengthening the wholesale market. Alternative capacity mechanisms Beyond the centralized capacity market and the reliability reserve analyzed here, further capacity mechanisms have been discussed in Germany and beyond. In particular, our analysis abstracts from decentralized capacity market concepts. In theory, decentralized mechanisms such as mandatory hedging on futures markets or decentralized capacity market designs could also provide incentives for the activation of demand-side flexibility options similar to the reliability reserve. However, the practical feasibility and functionality of these decentralized mechanisms with regard to effectively ensuring a high level of security of supply is yet to be demonstrated.13 We consider the reliability reserve to be a more viable alternative for practical implementation. Nonetheless, future research may further investigate the relative benefits of the reliability reserve and decentralized mechanisms. Policy takeaways and further research Both a centralized capacity market and an advanced reliability reserve are generally suitable to ensure security of supply in electricity markets. To reflect the increasing need for flexibility in power markets based on variable wind and solar energy, capacity mechanisms should be designed to provide strong incentives for the development of various flexibility technologies. Our model results illustrate that a centralized capacity market does not deliver these incentives in a 2030 scenario for Germany. Wholesale electricity prices are effectively capped at the variable costs of the marginal power plant, in our case an OCGT. This severely limits the revenue opportunities for demand-side flexibility options that rely on higher peak prices. In comparison, the advanced reliability reserve allows higher wholesale price spreads up to the activation price and thus unlocks significantly more investment into demand-side flexibility. Compared to a centralized capacity market, the reliability reserve induces seven times higher flexibility investments in energy-intensive industries and twice as high investments in district heating storage. Investments into process thermal storage increase by around a quarter. An advanced reliability reserve, properly designed to guarantee the security of supply while maintaining investment incentives, could thus be a preferable alternative to a centralized capacity market. It avoids very high wholesale prices above the activation price of 500 euro/MWhel, but still allows for sufficient price variation to enable investments in demand-side flexibility. Overall, the reserve could thus strengthen the wholesale market environment. Furthermore, it creates a learning environment for flexibility technologies, which become ever more relevant with increasing shares of variable renewable energy sources. This appears to be beneficial during the transition to climate neutral energy systems. An advanced reliability reserve could also be introduced faster and is more flexible than a centralized capacity market. We thus conclude that policymakers should consider the reliability reserve concept in upcoming decisions on capacity mechanisms in Germany and beyond. Considering the knowledge gaps that still exist, we see promising avenues for further research. For instance, the interplay of capacity mechanisms with support mechanisms for renewable electricity generation—which are also about to be reformed in Germany—are of interest. Further, analyzing the interactions of capacity mechanisms with decentralized options for flexibility and sector coupling appears promising, including prosumers who operate behind-the-meter batteries, electric vehicles or heat pumps. Such distributed assets are unlikely to be easily included in capacity mechanisms and their operations would rather have to be governed by strong wholesale price signals. Last, but not least, further research into the European coordination of the design, procurement, and operation of capacity mechanisms would be highly desirable. Limitations of the study The primary objective of this model analysis is not to produce precise forecasts of the composition of the German power plant fleet, future energy prices, or the dimensioning of capacity mechanisms. Rather, our analysis seeks to generate insights into how different capacity mechanism designs influence the investments and utilization of demand-side flexibility. For tractability, the electricity sector model comes with several simplifications. First, the analysis is based on a deterministic model with perfect foresight. This assumption may lead to an overly optimistic use of demand-side flexibility technologies, as they can ideally adjust to variable electricity generation from wind and solar power and utilize respective wholesale market price fluctuations. Likewise, investments into flexibility technologies could be lower under imperfect foresight, as intra- and inter-annual wholesale market price variations are uncertain, and investors would have to discount their expected revenues. However, such model distortions are generally affecting investments of all generation and flexibility assets, and are likely to be more relevant for the operation of long-duration storage technologies44 as compared to shorter-duration flexibility options which are in the focus here. Next, we assume that all types of generators are generally available when needed, and that market actors do not withhold capacity strategically. This is true for both cases: the centralized capacity market and the reliability reserve. In reality this might have to be ensured by additional regulation that creates respective incentives. For example, it could be made mandatory for all generators included in the capacity market to sign a reliability option, which forces them to pay back the difference between the market price and an agreed strike price in case the former exceeds the latter. Likewise, a non-availability penalty could be set for the generators contained in the reliability reserve. For simplicity, we abstract from such additional market imperfections and respective instruments in our stylized model setup. Further, we do not explicitly model electricity grids, but implicitly assume that sufficient grid capacity is available to avoid any bottlenecks in the transmission and distribution grids within Germany. In case of binding regional transmission grid constraints, it could be required to add spatial components to either of the two capacity mechanisms. Regulators might have to ensure that sufficient capacity is provided at particular locations in the constrained grid. While this applies both to a centralized capacity market and to a reliability reserve, such provisions could arguably be implemented more easily for a reserve. For example, the regulator could simply specify minimum procurement restrictions for reserve capacity at particular locations. In addition, respective reserve activation procedures would have to be defined such that grid constraints are reflected. Here, regulators could potentially leverage experiences from the existing grid reserve present in Germany today.13 In a centralized capacity market, however, it would not be straightforward to reconcile spatial differentiation with a single capacity product and a uniform capacity market price. For the sake of clarity and simplicity, we further exclude other flexibility options such as cross-border electricity exchange or flexible sector coupling technologies like electric vehicles or decentralized heat pumps in households.45 As a result, we underestimate the flexibility potential in the German power sector compared to real-world conditions, and overestimate the reliance on dispatchable capacities in Germany. The outcomes of this study should therefore be interpreted as an upper bound of the thermal power plant portfolio size for the weather years analyzed here. In reality, the European interconnection is an important source of flexibility,7 and coordinating capacity mechanisms across European countries remains important.28 In addition, we base our analysis on a sample of seven weather years. While this can be considered an improvement compared to many analyses that use a single weather year only, we may not capture rare but very prolonged renewable energy drought events, which are expected to become more relevant in future renewable energy systems.46,47,48,49 Expanding the dataset to include a broader range of weather years that covers more extreme events could lead to higher optimal capacities of firm generation, long-duration storage or demand-side flexibility technologies. The capacity results of this analysis are based on the challenging weather year 2009, yet results differ when other weather years are used (Figure S2). However, the qualitative result of higher investments into flexibility with a reliability reserve remains stable across years. Finally, our results hinge on the assumption that demand-side flexibility technologies cannot effectively participate in centralized capacity markets. This is in line with literature that concluded that capacity markets typically discriminate against demand-side flexibilities.38 It should be noted that some existing capacity markets do partially allow for the inclusion of demand-side flexibilities. However, the results are mixed: In the UK, demand-side options play a lesser role than initially expected, among others due to cumbersome pre-qualification processes and auction design favoring larger market participants, influenced by lobbying of large incumbent generators.50 In November 2018, the UK capacity market was even suspended after the EU General Court annulled its State-aid approval, citing concerns about the treatment of demand response. Following a formal investigation, the Commission re-approved it and adjustments were made to the rules, including a decrease in the minimum bidding threshold from 2 MW to 1 MW.51,52 In the American PJM market, demand-side flexibility can more easily participate due to a wider range of capacity products, e.g., delivery restricted only to seasonal periods or specific reaction times.52 The capacity products must thus be defined such that they consider not only the operational needs of the system, but also the technological characteristics of resources in the market.53 Generally, the participation of demand-side flexibilities in capacity markets requires applying non-discriminatory de-rating factors,33 the design of which appears to be a major challenge.54 Resource availability Lead contact Requests for further information and resources should be directed to and will be fulfilled by the lead contact, Wolf-Peter Schill (wschill@diw.de). Materials availability This study did not generate new unique reagents. Data and code availability •Part of the input data, all output data and all figures are available in a public repository: https://doi.org/10.5281/zenodo.15662854. Part of the input data, all output data and all figures are available in a public repository: https://doi.org/10.5281/zenodo.15662854. •The model code and all input data are available in a public repository: https://gitlab.com/diw-evu/projects/advanced-reliability-reserve. The model code and all input data are available in a public repository: https://gitlab.com/diw-evu/projects/advanced-reliability-reserve. •Any additional information required to reanalyze the data reported in this study is available from the lead contact upon request. Any additional information required to reanalyze the data reported in this study is available from the lead contact upon request. Acknowledgments We gratefully acknowledge internal funding by the German Institute for Economic Research (DIW Berlin) via the Bridge Project “SichER”. We thank participants of internal DIW Berlin seminars for constructive discussions on previous drafts. This article builds on a longer DIW Berlin project report that includes additional qualitative discussion of different capacity mechanisms and preliminary model results.13 Author contributions F.K.: investigation, data curation, formal analysis, methodology, and writing – original draft; K.N.: conceptualization, methodology, investigation, writing – review and editing, and funding acquisition; A.R.: methodology, software, formal analysis, investigation, data curation, writing – original draft, and visualization; W.-P.S.: conceptualization, methodology, formal analysis, investigation, writing – original draft, and funding acquisition; L.S.: methodology, investigation, and writing – original draft. Declaration of interests The authors declare no competing interests. STAR★Methods Key resources table REAGENT or RESOURCESOURCEIDENTIFIERDataModel and input dataGitlabhttps://gitlab.com/diw-evu/projects/advanced-reliability-reserveOutput data and figuresZenodohttps://doi.org/10.5281/zenodo.15662854CodePython 3.11.14Python Software Foundationhttps://www.python.org/downloads/GAMS 49.4.0GAMS Development Corphttps://www.gams.com/DIETERpy 2.0.0a3Gaete-Morales (2021)55https://gitlab.com/diw-evu/dieter_public/dieterpy/-/tree/9661ab1ce2dcdd05ce93b689ac315917200ea3fc/ Experimental model and study participant details Omitted as our study does not involve biological models. Method details Capacity expansion model To model the German electricity sector, we use the open-source capacity expansion model DIETER.55,56 Different versions of this model have been used in previous research to investigate the energy system implications of electricity storage,49,57 electric heating,58,59,60 electric mobility,61,62 green hydrogen63 and the European interconnection7 in the renewable energy transition. It is a linear program that minimizes overall system costs, i.e., investments and hourly use of various generation and storage technologies, for given, price-inelastic demand and weather data. Its results can be interpreted as the outcome of an ideal, frictionless market with perfect foresight. The model takes all consecutive hours of a full year into account to properly represent renewable energy variability and the use of flexibility technologies. In this analysis, we use a reduced version of the model that focuses on Germany only, but includes a wide range of demand-side flexibility technology options. Details regarding the modeling, technical specifications, and costs of these demand-side flexibility technologies are provided in the next section. Other key input data for the model include time series of the demand for electricity, heat and green hydrogen, time series on the availability of variable renewables, as well as cost assumptions and bounds for investments in various technologies. While the capacities of variable wind and solar, hydropower and bio energy are fixed exogenously, the remaining part of the power plant fleet is determined endogenously for a challenging weather year with a high peak residual load (2009). This includes in particular the capacities of gas-fired power plants, batteries, hydrogen storage as well as investments in various flexibility options on the demand side. These capacities are then fixed for the model runs of other weather years. Key input parameters Variable renewable energy capacities are fixed at the values targeted by the German government for 2030: Photovoltaics (215GWel), onshore wind power (115GWel), and offshore wind power (30GWel). The capacities of oil-fired, bio energy, hydropower and pumped hydro storage power plants are also fixed (Table S1). We assume that the coal phase-out will be completed by 2030. This means that neither lignite nor hard coal power plants are available to the model. In contrast, the model endogenously determines installed capacity of two types of gas-fired power plants (gas turbine, combined-cycle gas), as well as battery and hydrogen storage, including electrolyzers and hydrogen power plants. In addition, the model can choose to invest in various demand-side flexibility options, as discussed in the next subsection. Techno-economic assumptions are summarized in Table S1. We further assume a CO2 price of 130euro/ton. The annual electric load, excluding the electricity demand for district heating, is assumed to be around 670TWhel for all years. The annual difference between this number and the current demand, which accounts for additional sector coupling, is distributed equally over all hours. This approach neither results in load peaks that are overly inflated, nor in underestimating future demand peaks by implicitly assuming too much flexibility, when compared to more extreme alternative methods such as linear scaling or off-peak distribution of additional demand. We add the amount of electricity required to cover district heating, which ranges between 87 and 103TWhth (average 95TWhth) across all scenarios and years, which corresponds to around 30% of the space and water heating demand of multifamily and commercial buildings. This translates to roughly 19 to 25TWhel (average 22TWhel) of electric energy. Therefore, total annual electricity demand is around 700TWhel, which ranges between current policy goals and recent system studies.2,3,64 The time series for electricity demand65 and renewable availability66 come from ENTSO-E’s European Resource Adequacy Assessment (ERAA) 2023. The time series for heat demand and efficiency of heat pumps are based on the When2Heat dataset.67 Hence, depending on the weather year, not only the electricity generation of wind and solar PV varies, but also electricity demand related to space heating. Considered demand-side flexibility options We consider demand-side flexibility options (also referred to as demand response) for energy-intensive industries, process heat, and district heating. Technical potentials, parameters, and costs are projected for the year 2030 based on literature data. This includes the potentials for load change, their maximum duration of load change in hours, and associated costs.68 We assume that the various options only require investments in energy storage or intermediate product storage to unlock demand-side flexibility, while the necessary production capacity already exists (see S2.1). Note that both the product storage and thermal storage options considered here have similar effects on the power sector as batteries or other electricity storage technologies. Accordingly, their representation in the model is also similar to electricity storage. The details of the calculations, assumptions, and literature data used to determine the demand response potentials are explained in S2. We analyze the flexibility potentials of electric steel, aluminum, paper, cement, chlor-alkali process, and air separation separately, relying on the existing literature (Tables S2 and S3). These processes are of particular interest due to their high electricity intensity.69 Energy-intensive industry The flexibility potentials of installed electrical load of other energy-intensive industries are estimated as an aggregate, based on their reported electricity demand.70 Due to the increasing future electrification, the electricity demand has been scaled up by 39% in 2030.71 We assume that 10% of the installed electrical load could be run flexibly (see assumptions in S2). To reduce complexity, we further assume that electricity-intensive parts of a production process can be shifted in time by storing the products in a physical storage (load shifting). Load shedding, i.e., the temporary reduction of electricity-based production without compensation at a later point, also referred to as “demand destruction”, is not considered. Figure S1 shows the resulting potentials for flexible loads in energy-intensive industries for 2030. The costs of demand response account for both the costs of retrieving the demand response potential and the investment costs of the storage system. The applied costs and assumptions are described in S2.1. The model decides endogenously, based on storage and activation costs, which amount of energy can be shifted at maximum. For example, all technologies with a load change duration of 3 h (see S2.1, assumption 6) and an overall shiftable load of 1558MWel (Figure S1) have the potential to add a maximum storage size up to 4674MWhel. Technologies with correspondingly higher durations can also be equipped with larger storage. Process heat Process heat constitutes the largest share of industrial energy consumption in Europe. Currently, natural gas is the dominant energy carrier, supplying 35% of total energy demand of process heating in 2019 in the EU-27.72 The electrification of process heat is therefore crucial to achieve future climate targets.71 In contrast to the seasonal demand profile of space heating, the demand for process heat does not vary throughout the year, as we assume no seasonality of industrial production.5 We consider process heat up to a temperature level of 500°C, at which heat pumps and resistance heaters (electrode boiler) can be used on an industrial scale according to the current state of the technology.72,73 Thereby, the process heat is provided by high-temperature heat pumps for temperatures up to 160°C and by resistance heaters for temperatures between 200°C and 500°C.74 We base the calculation of future electrical energy demand for the provision of process heat on a study by the German Environment Agency.75 The metal and mineral industry is excluded because their production processes mainly require heat directly in the processes and not indirectly, e.g., via process steam.74 Additionally, the required temperatures are usually at the upper limit of the temperature levels considered here.75 We assume a standard packed-bed thermal energy storage system with a storage efficiency of 90%76 and with a maximum installed storage capacity of 30% of the considered installed electrical capacity for process heat by the year 2030. The maximum thermal storage period is 72 h. Due to low investment costs, the installed excess capacity for process heat is realized by resistance heaters with a thermal efficiency of 99%.74 Storage losses as a percentage of capacity per day are 3%.8 The following table summarizes the data for process heat. The detailed calculations, data basis and assumptions are described in S2.2. The power rating of process heat installations is fixed in the model. However, the amount of energy, i.e., the size of a thermal storage, is endogenous. Projections of future energy demand for the provision of industrial process heatTemperature range process heat [°C]Thermal energy demand for process heat75 [TWhth/a]Electrical energy demand for high temperature heat pumpsa [TWhel]Electrical energy demand for resistance heatersb [TWhel/a]<1006718.1–100–16070.219–160–50033.1–33.4Sum170.270.5aAssumed coefficient of performance (COP) high temperature heat-pump: 3.7.74bAssumed efficiency factor resistance heater 99%.74 Projections of future energy demand for the provision of industrial process heat Assumed coefficient of performance (COP) high temperature heat-pump: 3.7.74 Assumed efficiency factor resistance heater 99%.74 District heating To consider the flexibility potential of thermal storage in heating networks, we include a simplified district heating market in the model. We assume that an average of 95TWhth of heat per year is provided by district heating, which corresponds to 30% of the German heat demand in residential and commercial buildings. In the model, the entire district heating demand is met by large-scale ground-sourced heat pumps. Load shifting is enabled by sensible thermal storage systems using water as a storage medium. The storage efficiency is 90% with standing losses of 2% per day.8 The size of the thermal storage tank is endogenously determined by the model, while the heat and electricity outputs of heat pumps are exogenously given. Investment costs for the thermal storage amount to 50euro/kWhth.8 Decentralized heat pumps and their buffer storage capacities are not modeled. Despite typically small storage sizes, they may also contribute some short-duration flexibility to the electricity system.59 Two policy scenarios We investigate two policy scenarios with different capacity mechanisms: a centralized capacity market and an advanced reliability reserve. The centralized capacity market supplements the electricity market by specifying a desired amount of firm capacity in the model. We require that 101.3GWel of capacity are provided, which can be covered by natural gas (open or combined-cycle gas turbines), oil, bio energy, hydro reservoirs, and long-duration hydrogen storage. The capacity target is set so that in all modeled weather years the residual load (electric load excluding electricity demand for heat minus generation from variable renewables) can be covered by the generation capacity contracted in the capacity market, including a safety margin of five GWel. However, as the capacities of the other technologies are fixed, only gas and hydrogen power plants are available to the model as free variables. In the scenario with an advanced reliability reserve, the electricity market is complemented by reserve power plants located outside the electricity market. These run only when the electricity price reaches the defined activation price of 500euro/MWhel. We choose this price level to satisfy and balance two objectives. First, the price must be high enough to provide sufficient investment incentives into new generation and flexibility capacity in the wholesale market. Secondly, it must be low enough to be credible: if the activation price is too high, the regulator may face political or societal pressure to activate the reserve already at lower prices to shield consumers from high energy costs (more on this in the discussion section). The size of the reserve is determined in such a way that the sum of firm capacity in the wholesale market and in the reserve equals the size of the capacity market (101.3GWel), resulting in a similar level of resource adequacy. To properly account for temporal correlations between renewable energy scarcity and high demand peaks in winter periods, the model is run in a “summer-to-summer” mode (i.e., from July to June). When, for instance, the year “2009” is mentioned, we refer to the period “July 2009 – June 2010”. For each scenario, we apply a two-step procedure: In the first step, we obtain a least-cost solution and optimal capacities of power plants and demand flexibility technologies for a single weather year. We use the year 2009, which is the weather year with the highest residual load in our sample, and has already served to determine the size of the capacity market. In a second step, the model is run in a dispatch-only mode, using the optimal capacities of the first step, and is iterated over the seven weather years 2008–2014 to analyze the effects on prices and the rest of the electricity sector. Importantly, we assume that demand-side flexibility technologies can neither participate in the capacity market nor in the reserve. In both scenarios, demand-side flexibility technologies thus do not receive any payments from capacity mechanisms, but have to cover their investment and operating costs exclusively by revenues from the wholesale market. Quantification and statistical analysis There are no quantification or statistical analyses to include in this study. Footnotes Supplemental information can be found online at https://doi.org/10.1016/j.isci.2026.116176. Supplemental information Document S1. Figures S1–S7, Tables S1–S3 References Associated Data Supplementary Materials Document S1. Figures S1–S7, Tables S1–S3 Data Availability Statement •Part of the input data, all output data and all figures are available in a public repository: https://doi.org/10.5281/zenodo.15662854. Part of the input data, all output data and all figures are available in a public repository: https://doi.org/10.5281/zenodo.15662854. •The model code and all input data are available in a public repository: https://gitlab.com/diw-evu/projects/advanced-reliability-reserve. The model code and all input data are available in a public repository: https://gitlab.com/diw-evu/projects/advanced-reliability-reserve. •Any additional information required to reanalyze the data reported in this study is available from the lead contact upon request. Any additional information required to reanalyze the data reported in this study is available from the lead contact upon request.