Storage and demand response contribution to firm capacity: Analysis of the Spanish electricity system

Energy Reports 8 (2022) 10546–10560 Contents lists available at ScienceDirect Energy Reports journal homepage: www.elsevier.com/locate/egyr Research paper Storage and demand response contribution to firm capacity: Analysis of the Spanish electricity system Teresa Freire-Barcelóa,∗, Francisco Martín-Martíneza, Álvaro Sánchez-Mirallesa, Michel Riviera, Tomás Gómez San Romána, Sébastien Huclina,b, José Pablo Chaves Ávilaa, Andres Ramosa aInstitute for Research in Technology (IIT), ICAI School of Engineering, Universidad Pontificia Comillas, Santa Cruz de Marcenado 26, 28015, Madrid, Spain bBasque Centre for Climate Change (BC3), Sede Building 1, 1st floor, Scientific Campus, 48940, Leioa, Spain article info Article history: Received 22 April 2022 Received in revised form 1 July 2022 Accepted 3 August 2022 Available online 27 August 2022 Keywords: Firm capacity Demand response Demand growth Batteries Pumped-hydro storage abstract Provision of firm capacity will become a challenge in power systems dominated by renewable generation. This paper analyzes the competitiveness and role of battery storage, six types of pumpedhydro storage, open cycle gas turbine (OCGT), and demand response (DR) technologies in providing the firm capacity required to guarantee the security of supply in a real-size power system such as the Spanish one in horizon 2030. The paper contributes with detailed and realistic modeling of the DR capabilities. Demand is disaggregated by sector and activities and projected towards 2030, applying a growth rate by activity. The load flexibility constraints are considered to ensure the validity of the results. A generation operation planning and expansion model, SPLODER, is conveniently upgraded to properly represent the different storage alternatives addressed in the paper. The results highlight the importance of considering demand response for evaluating long-term firm capacity requirements, showing a non-negligible impact on the investment decisions on the amount of firm capacity required in the system and the optimal shares of wind and solar PV renewable generation. Results also show the dominance of cost-competitiveness of pumped hydro and OCGTs over batteries. Additionally, capacity payments are required to support firm capacity providers’ investments. ©2022 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). 1. Introduction Ensuring the security of supply in the Spanish electricity system is a task that faces multiple challenges in the near future. The Spanish national energy strategy commits to achieving at least 74% of renewable electricity generation by 2030 (PNIEC,2020). Spurred by their increasingly competitive investment costs, there is no doubt the system will mainly rely on wind farms and solar photovoltaic (PV) power plants to meet this target. The production of such renewable generation is fully weather dependent, severely jeopardizing the security of supply, that is, the system’s availability to count on enough available generation to meet the ∗Corresponding author. E-mail addresses: [email protected] (T. Freire-Barceló), [email protected] (F. Martín-Martínez), [email protected] (Á. Sánchez-Miralles), [email protected] (M. Rivier), [email protected] (T.G. San Román), [email protected] (S. Huclin), [email protected] (J.P.C. Ávila), [email protected] (A. Ramos). demand at any time. These renewable sources are substituting thermal generation, which has traditionally provided the system security of supply and flexibility. Therefore, it will be necessary to resort to additional resources to fill the gap left by phased-out thermal generators as firm capacity1providers. Moreover, to be aligned with European goals (Meeus and Nouicer,2020), these new resources should also have low emissions. Storage facilities are one of the most suitable technologies to provide firm capacity. A large-scale battery is one of the options. (Mallapragada et al.,2020) assess its potential as the primary resource of firm capacity, concluding that further cost reduction is necessary for batteries to become a cost-effective alternative. Although available in scarce locations, pumped-hydro storage is another option to be considered due to their maturity, large storage capacity, relatively low capital costs, mainly when they take 1Firm capacity technologies refer to energy sources whose capacity is available at the most critical periods of generation as it is controllable and able to supply energy as needed independently of weather or external conditions (Zachary et al.,2019) safeguarding system adequacy. https://doi.org/10.1016/j.egyr.2022.08.014 2352-4847/©2022 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/bync-nd/4.0/). T. Freire-Barceló, F. Martín-Martínez, Á. Sánchez-Miralles et al. Energy Reports 8 (2022) 10546–10560 advantage of some already installed hydro reservoirs, and fast response capability when needed (Amirante et al.,2017). Many other innovative storage kinds of resources have also been assessed in Korkmaz (2019), such as compressed air energy storage, hydrogen storage, and other developing technologies such as flow batteries and liquid air energy storage. However, none of these more innovative resources is yet close to being cost-competitive. Other options to provide security of supply beyond storage technologies have also been considered in some publications addressing the future of electricity systems. For instance, the use of power plants with carbon capture and storage combined with high interconnectors capacity (Brouwer et al.,2014), or the geographical diversification of wind farms in Germany, show the reduction of firm capacity needed (Bucksteeg,2019). There are other papers similar to this one, where the high penetration of renewables is the issue that result in the pursuit of generation alternatives to guarantee the security of supply of the electricity system. Gaete-Morales et al. (2019) provides the analysis of the Chilean system in horizon 2050. Ruhnau and Qvist (2022) compare different storage types (hydrogen storage, pumped-hydro storage & batteries) to guarantee the security of supply in the German electricity system. Arion (2020) analyze Moldova’s pumped-hydro storage needs and Lu et al. (2021) assess China’s options to achieve a carbon-neutral electricity system where DR is mentioned qualitatively. However, it is not assessed its impact on the electricity system. As presented, in the literature, firm capacity requirements are primarily addressed, in addition to generation units, with pumped-hydro storage and lithium-ion batteries. The main contribution of this paper is the analysis of an additional competitor in the provision of firmness, neglected in those studies, which is the impact of demand-side management in a high renewables penetration electricity system. Besides, the paper contributes with the upgrades performed in the model to enable it to be used for the first time for this purpose. DR is expected to rapidly increase to comply with European Commission directives (European Parliament,2019), although regulatory, technological, and social barriers (Freire-Barceló et al.,2022) need to be addressed. New automation technologies and increasing customer engagement (Gómez-Barredo et al.,2021) may have a nonnegligible impact in many aspects, also regarding the firmness requisites and generation investment planning. Moreover, the information available in the literature about the origin of electricity demand and the corresponding flexibility, was completely outdated (Red Eléctrica de España,1998; Instituto para la Diversificación y Ahorro de la Energía,2016a). Therefore, a disaggregated representation of the demand differentiates demand growth rates towards 2030 by demand use and to accurately and realistically represent the demand response capabilities of each consumption category. Thus, this valuable and useful information can be used for developing different types of studies. Overall, the work presented in this paper is a natural followup of the one presented in Huclin et al. (2022), where firm coefficients are determined for the different storage technologies. Using those coefficients and the ones that can be found in the literature, the paper analyzes and discusses the need for new firm resources to maintain the security of supply of the Spanish electricity system in horizon 2030 and to consider the generation volatility in a scenario 2030 with a high share of renewables. The SPLODER model, a generation expansion planning model, is the tool used for the analysis. The first version of the model has been presented in Martínez et al. (2017) where the core equations were introduced with the novelty of a disaggregated representation of the demand by usage types such as heating and cooling, domestic hot water or electric vehicles. This fact limits the capability to shift demand freely since each consumption type have specific constraints. The model has already been applied in previous studies such as Martín-Martínez et al. (2017), in which the analysis and scenario definition was focused on comparing centralized vs. distributed generation alternatives considering flexible loads. The model formulation has also been upgraded and used in Gerres et al. (2019), including new remuneration mechanisms required to achieve the renewable penetration targets together with enough firm capacity provision. In addition, the model is already prepared to manage flexible demand and to develop this study, it has been upgraded to properly represent in detail different firm capacity providers, namely different kinds of pumped-hydro storages, large-scale batteries, and OCGT, as well as the consideration of the demand-side response. Thus, the model is used for the first time to analyze the resources required to provide firm capacity and the competitiveness among them, and how the full potential of DR may impact these results as well as the optimal investments in renewable generation. The Spanish electricity system is weakly interconnected, and it can be considered as an energy island (THE LOCAL,2022;Wilson and Muñoz,2022). Therefore, neglecting interconnections allows obtaining insights into the possible evolution of power systems with high penetration of variable renewable energy resources. The main contributions of this paper are threefold. Firstly, the paper contributes by providing a detailed Spanish demand growth disaggregated analysis, obtaining two extreme cases. This allows the consideration of a detailed and realistic representation of DR, previously modeled, in a real-sized electricity system and assessing its role and relevance when planning the future firm capacity provision, as it considerably diminishes the required investments. Secondly, the paper contributes to analyzing the competitiveness among different firm capacity technology providers under different scenarios. Thirdly, the methodology applied resorts to an optimization tool, the SPLODER model. SPLODER mathematical formulation presented in Martínez et al. (2017) and Gerres et al. (2019) has been upgraded with the inclusion of storage technologies to enable the analysis of the contribution of DR and other technologies to firm capacity requirements. These changes are thoroughly explained in Section 3, thus contributing with a more complete model in the literature. The rest of the paper presents the following structure. Section 2presents a demand growth analysis necessary for characterizing the Spanish electricity system until 2030 and identifies demand management capabilities as another firm capacity resource. Section 3describes the new formulation added to SPLODER, differentiating multiple different types of centralized storage resources, and the description of the associated required input data. Section 4presents the scenarios assessed in the paper based on the national policy targets and extended to cover different sensitivities aligned with this study’s main aim. Results are discussed in Section 5. Finally, Section 6assesses the findings and identifies additional future research needs. 2. Demand growth analysis In the next future, electricity systems will experience a deep change in the demand side towards more efficient energy consumption. Besides, decarbonization targets will boost higher levels of electrification so that significant growth of demand flexibility is expected to be available. The more electric loads, the easier it becomes to profit from their adaptability to consume at different times without compromising the electricity end usage and the consumers’ comfort. To take advantage of controllable loads, it is necessary to compute the load flexibility potential and the limits that consumption can be managed. To this end, four categories of electricity consumption within the residential 10547 T. Freire-Barceló, F. Martín-Martínez, Á. Sánchez-Miralles et al. Energy Reports 8 (2022) 10546–10560 Table 1 2015 electricity demand in Spain. Electricity demand 2015 Sector TWh % Residential 72.73 29% Services 76.24 31% Transport 6.40 3% Industry 91.86 37% Total 247.22 100% Table 2 2015 electricity demand of the residential sector in Spain. Electricity demand 2015 Residential sector TWh % Heating 4.42 6% Cooling 3.40 5% DHW 4.48 6% Lighting and others 60.43 83% Total residential 72.73 100% and services sectors have been considered to be controllable in different ways: heating & cooling (climatization), domestic hot water (DHW), refrigeration (cold chain, freezers, and fridges) and electric vehicles (EV). It is relevant to come up with an estimation of the amount of demand associated with each of these categories and which proportion of each one will be ready to be controllable. To estimate the amount of demand in the Spanish sector, classified according to the above-mentioned categories, the following methodology has been applied: – First, load hourly data of the Spanish electricity system for the year 2015 are considered for each sector (residential, services, transport, industry) as the base profiles. This year has been selected because the information available is very detailed by different usage types. In addition, the pandemic does not influence consumption and clearly serves as a reference year for estimating growth rates (Instituto para la Diversificación y Ahorro de la Energía,2011). – Second, residential and commercial sectors’ data are further disaggregated using a set of complementarity reports to reach the granularity required to identify the potential controllable, that is, climatization, DHW, refrigeration, and estimation of EVs loads. This disaggregation was deeply studied during the years 2014–2018. – Third, a literature review has been performed to set a range of annual demand growth for each category. – Fourth, the two extreme values found in the literature for the 2030 demand growth will define the range in which it is located and the potentially controllable load that the model will consider. Furthermore, to validate the estimated growth, it has been checked that there are no inconsistencies with the information available up to 2022. Table 1 shows the 2015 electricity demand breakdown in the four main consumption sectors, obtained as the average value from reports (PNIEC,2020;International Energy Agency, 2015;Linares and Declercq,2018;Deloitte,2018b;Government of Spain,2018). As shown in Tables 2 &3, the ‘‘Residential’’ and ‘‘Services’’ consumption sectors have then been further broken down into different categories based on Instituto para la Diversificación y Ahorro de la Energía (2016a), Persson and Werner (2015) and Consultores (2005), using the ‘‘Lighting and others’’ category to gather the rest of demand, considered as inflexible. The transport sector electricity consumption for 2015 has been calculated as the sum of EV and electric trains consumption. Table 3 2015 electricity demand of the services sector in Spain. Electricity demand 2015 Services sector TWh % Heating 13.89 18% Cooling 11.11 15% DHW 0.80 1% Lighting and others 47.91 63% Refrigeration 2.52 3% Total services 76.24 100% Table 4 2015 EV fleet in Spain. EV 2015 EV fleet [N◦of vehicles] 5848 Table 5 2015 total transport sector electricity consumption in Spain. Transport 2015 EV consumption [TWh] 0.017 Trains [TWh] 6.39 Total [TWh] 6.40 Table 6 2030 min. and max. demand for the residential sector [TWh]. Residential demand 2030 min 2030 max Heating 36.87 48.29 Cooling 5.65 13.48 DHW 14.79 22.77 Lighting and others 39.68 45.46 Total residential 97 130 The following calculation has been applied to estimate the part of it corresponding to EV. The EV fleet published in European Commission (2015) and presented in Table 4 has been used as the starting point. Then, an average EV consumption of 0.2 kWh/km has been assumed based on the average electric car consumption in Oak Ridge (2022), and considering the inefficiency due to the charging and discharging cycle as assessed in Iclodean et al. (2017). Finally, Spain’s average daily car uses is assumed to be 40 km/day, as shown in European Environment Agency (2019). These assumptions led to an estimation of 17 GWh for the total EV electricity demand in 2015, as presented in Table 5. The remaining transport sector electricity consumption in Table 1 has been associated with the railway sector. Once the 2015 demand breakdown has been set, minimum and maximum demand growth values for 2030 have been estimated for the different sectors and consumption categories. For residential and service sectors, forecasts published in Linares and Declercq (2018), Instituto para la Diversificación y Ahorro de la Energía (2011), Jakubcionis and Carlsson (2018) and Instituto para la Diversificación y Ahorro de la Energía (2016b) have been contrasted to determine the demand growth rate for the different consumption categories. The most optimistic and pessimistic predictions published across the reviewed publications have been selected to define a range with the minimum and the maximum demand growth for 2030. The residential and services sector’s minimum and maximum consumption are presented in Tables 6 and 7, respectively. Transport sector demand growth assumed in the study and presented in Table 8 is based on the most extreme values concerning the expected EV fleet growth stated by PNIEC (2020) and International Energy Agency (2015) . 10548 T. Freire-Barceló, F. Martín-Martínez, Á. Sánchez-Miralles et al. Energy Reports 8 (2022) 10546–10560 Table 7 2030 min. and max. demand for the services sector [TWh]. Services demand 2030 min 2030 max Heating 14.71 22.47 Cooling 12.03 17.08 DHW 0.89 1.78 Lighting and others 56.92 43.49 Refrigeration 2.80 2.81 Trains 6.60 8.77 Total including trains 93.94 96.39 Table 8 2030 min. and max. demand for the transport sector. Transport demand 2030 min 2030 max EV fleet [N◦of vehicles] 300,000 5,000,000 EV consumption [TWh] 0.876 14.6 Table 9 2030 min. and max. demand for the industrial sector [TWh]. Industrial demand 2030 min 2030 max Industry 103.6 124.9 Table 10 2030 total min. max. and average demand [TWh]. Total demand 2030 min 2030 max Average Residential 97 130 114 Services 94 96 95 Industry 104 125 114 Transport 1 15 8 Total 295 366 331 Industrial demand growth for 2030 is expected to be in the range described in Linares and Declercq (2018) and presented in Table 9. In the scenarios presented below, intermediate growths inbetween the range presented (2030Min and 2030Max) are considered individually for each sector and consumption category. Table 10 presents the minimum, the maximum and the average of the total electricity demand by sector. As a final step in the demand characterization, different degrees of penetration of demand response are considered in the study. It is assumed that only a fraction of all the potentially controllable loads will be ready to participate in demand response programs by 2030. Finally, it is important to properly model the actual capabilities of demand response of such manageable loads. Demand management corresponds to a demand shift among hours. But demand cannot be shifted in any way. Each of these loads obeys to a process that limits its response capabilities. For instance, the heating and cooling demand cannot be freely shifted over time as it should ensure that the temperature of the building does not go out of a preset comfort band. The SPLODER model enables a very detailed representation of the demand. The way different consumption categories that can provide DR are modeled is detailed in Martín-Martínez et al. (2017), but overall is as follows: •EVs: a given portion of the EV demand if considered fully flexible and manageable during the 24 h. The rest of EV demand is considered a non-flexible one and follows a predetermined charging profile, split up between base and peak hours, as presented in Gerres et al. (2019). •Heating and cooling: a reference and comfort band temperature is set according to predefined temperature bands. The building thermal inertia is considered to simulate the temperature evolution. Buildings are clustered according to their geographical area with different external temperatures according to the month and their level of thermal isolation. •DHW: the portion assumed to be flexible can be managed freely during a whole day as it is associated with the stored hot water inertia. •Refrigeration: is flexible if the average temperature follows a reference temperature. The temperature could be two degrees upper or lower this reference, whereas the average is respected at the end of the day. The energy to keep the temperature in reference during a day can be considered constant in an adiabatic system, and it can be freely allocated throughout the day. 3. Characterizing storage resources in SPLODER The SPLODER model performs an optimal generation expansion plan minimizing investment, production, and O&M costs for a given time horizon. Years are represented by a set of clustered representative weeks with hourly time granularity. These four weeks represent the seasons’ winter, spring/fall, summer, and vacations, each of them with different weights along the year. The four-week demand and renewables generation profiles are obtained by applying the k-means clustering algorithm (Hartigan and Wong,1979). Several renewable production profiles are considered in the optimization framework (three renewable scenarios are used in this study); although there is stochasticity, another model with the 8760 h of the year could validate SPLODERs operation decisions; however, this proof is out of the scope of the paper. The main inputs and outputs of the model are summarized in Fig. 1. As further explained in Gerres et al. (2019), the resulting optimal generation and storage mix should comply with two main constraints. First, generation and demand should meet hourly. Hourly energy prices (= C/MWh) are obtained as the dual variable of such generation–demand balance constraint. Second, the model guarantees that enough firm capacity is provided to the system (a 10% reserve margin over the peak demand is used in this study). Each generation technology contributes differently to the system firm capacity. A specific firm capacity coefficient is assigned to each generation technology. An annual based capacity price (= C/MW) is obtained as the dual variable of firm capacity requirement constraint and thus used to model a new capacity market for the Spanish electricity system. Hence, generation units are remunerated both for providing energy and firm capacity to the system, and both incomes are considered to ensure the full cost recovery of all newly installed units. Balancing services provision is out of the scope of this paper. The full description of the SPLODER optimization model, including a detailed explanation and formulation of the objective function and constraints, can be found in Martínez et al. (2017) and Gerres et al. (2019). For this paper, SPLODER has been upgraded to improve the modeling of firm capacity resources, enabling the characterization of different pumped-hydro storage types and adding centralized battery storage as candidate technologies to be considered in the future generation and storage mix. This section focuses specifically on describing the storage technologies considered in this study and the detailed formulation of equations added to the existing SPLODER model to represent their behavior properly. According to the Spanish context, six different categories of pumped-hydro storage have been modeled as candidates for expanding the system. They have different storage capacity sizes and investment costs representing basically two options: build an artificial upper reservoir associated with an existing storage hydro power plant or build a new penstock with a reversible turbine in an already existing pumped-hydro storage power plant (CEEPR, 2020). Table 11 summarizes the set of new pumped hydro storage 10549 T. Freire-Barceló, F. Martín-Martínez, Á. Sánchez-Miralles et al. Energy Reports 8 (2022) 10546–10560 Fig. 1. Main inputs and outputs of SPLODER (Gerres et al.,2019). Table 11 Storage data. Agent Max. install. [MW] Min. install. [MW] Annualized install. cost [= C/MW] Fix annual O&M [= C/MW] Var annual O&M [= C/MWh] Firm Coeff. Round-trip eff. Charge hours Discharge hours Batt_cent 0 0 133,799 5550 0.00025 0.69 0.9 4 4 Sto_8h_1 1000 400 39,255 9000 3 0.96 0.75 8 8 Sto_20h_1 2000 400 52,340 12,000 3 0.96 0.75 20 20 Sto_20h_2 5800 400 65,424 15,000 3 0.96 0.75 20 20 Sto_40h_1 800 400 35,983 8250 3 0.96 0.75 40 40 Sto_40h_2 600 400 62,153 14,250 3 0.96 0.75 40 40 Sto_60h_1 1500 400 55,611 12,750 3 0.96 0.75 60 60 types included in SPLODER. Table 11 also shows the centralized large-scale Li-Ion batteries that have been modeled and considered in the study. The economy of scale of centralized storage makes distributed batteries unprofitable, thus discouraging their deployment from a centralized point of view. The same thing happens with distributed solar generation, although in this case, a fixed amount has been set as input in the model (Red Eléctrica de España,2019), considering their natural deployment due to individual motivations or local tariffs incentives. However, profits from solving local distributed system congestion problems are not considered. Installation, fix, and variable O&M costs for batteries are based on Mongird et al. (2019). Pumped-storage hydro data, including maximum available installation capacities for the different options and the associated firm capacity coefficients, are estimated based on CEEPR (2020) and PNIEC (2020). Additionally, the annualized installation costs for the different storage options have been estimated with public pumped-storage hydro projects with different storage capacities (Repsol,2021;European Commission, 2019;Roca,2019,2020). These storage facilities are modeled within an upgraded version of SPLODER. The detailed formulation of equations is provided, preceded by the used nomenclature (Table 12): All new storage types, both pumping hydro and batteries, have been modeled similarly. Constraints (1),(2),(3),(4) &(5), control the state of charge (SOC) of all storage types, forcing it not to exceed the maximum storage capacity according to investment decisions. (INSTALLEDist +newInstallist)×DISCHARHOURSist ≥socist,p,w,m,h (1) (INSTALLEDist +newInstallist)×DISCHARHOURSist ≥socist,p,w,m,h−1+chargei,p,w,m,h×YIELDist ∀ist,p, w, m,h>1 (2) (INSTALLEDist +newInstallist)×DISCHARHOURSist ≥socist,p,w,m−1,24 +chargei,p,w,m,h×YIELDist ∀ist,p, w, m,h=1 (3) (INSTALLEDist +newInstallist)×DISCHARHOURSist ≥socist,p,w−1,7,24 +chargei,p,w,m,h×YIELDist ∀ist,p, w, m=1,h1 (4) (INSTALLEDist +newInstallist)×DISCHARHOURSist ≥socist,p,4,7,24 +chargei,p,w,m,h ×YIELDist∀ist,p, w =1,m=1,h1 (5) Constraints (6),(7),(8) &(9), are the boundary conditions for the maximum energy that can be discharged at each time of the year. The YIELDist considered in the model gathers the round-trip efficiency. Hence, it is not necessary to multiply the dischargei,p,w,m,hvariable again. socist,p,w,m,h−1≥dischargei,p,w,m,h∀ist,p, w, m,h>1 (6) socist,p,w,m−1,24 ≥dischargei,p,w,m,h∀ist,p, w, m,h=1 (7) socist,p,w−1,7,24 ≥dischargei,p,w,m,h∀ist,p, w −1,m=1,h=1 (8) socist,p,4,7,24 ≥dischargei,p,w,m,h∀ist,p, w =1,m=1,h=1 (9) Constraints (10),(11),(12) &(13) calculate the SOC at every hour for each storage type. The SOC at each hour equals to the SOC at the previous hour, plus the charged energy, minus the discharged energy. The four equations differ only in reference to the previous hour SOC, in which, due to the temporal granularity of the model, the sets to be referred to slightly change with time (depending on the week, the day and the hour). socist,p,w,m,h=socist,p,w,m,h−1+(chargei,p,w,m,h×YIELDist ) −dischargei,p,w,m,h∀ist,p, w, m,h>1 (10) 10550 T. Freire-Barceló, F. Martín-Martínez, Á. Sánchez-Miralles et al. Energy Reports 8 (2022) 10546–10560 Table 12 Sets, parameters, and variables added. Sets i Technology type {1–27} ist ϵi New storage types {1–7} p Renewable scenario {1–3} w Week {1–4} m Day of the week {1–7} h Hour {1–24} Parameters INSTALLEDiExisting power previously installed for each technology i [MW] CHARHOURSiCharging hours for each type of storage [h] DISCHARHOURSiDischarging hours for each type of storage [h] YIELDiStorage round-trip efficiency by technology i [%] MAXINSTALLiMaximum capacity to be installed of each technology i [MW] Variables finalinstallediTotal capacity in place for each technology i [MW] newInstalliNew installed capacity for each technology i [MW] soci,p,w,m,hState of charge of hydro plant at each hour [MWh] chargei,p,w,m,hPumped hydro storage charge at each hour [MW] dischargei,p,w,m,hPumped hydro storage discharge at each hour [MW] energySelli,p,w,m,hHourly energy sold by each technology i [MWh] energyBoughti,p,w,m,hHourly energy bought by each technology i [MWh] socist,p,w,m,h=socist,p,w,m−1,24 +(chargei,p,w,m,h×YIELDist ) −dischargei,p,w,m,h∀ist,p, w, m>1,h=1 (11) socist,p,w,m,h=socist,p,w−1,7,24 +(chargei,p,w,m,h×YIELDist ) −dischargei,p,w,m,h∀ist,p, w > 1,m=1,h=1 (12) socist,p,w,m,h=socist,p,4,7,24 +(chargei,p,w,m,h×YIELDist ) −dischargei,p,w,m,h∀ist,p, w =1,m=1,h=1 (13) Constraint (14) sets the SOC to be the same at the beginning and end of the year (first hour of the first representative week and last hour of the last representative week) for all storage types in order to better represent the storage potential throughout the year. socist,p,1,1,0=socist,p,4,7,24 (14) Constraints (15) &(16) set the charging and discharging speed rate depending on the storage hours’ capacity. (INSTALLEDist +newInstallist)×DISCHARHOURSist /CHARHOURSist ≥chargeist,p,w,m,h(15) (INSTALLEDist +newInstallist)×DISCHARHOURSist /CHARHOURSist ≥dischargeist,p,w,m,h(16) For each pumping storage type, the new installed capacity cannot exceed the total available capacity to be built, as stated in (17). MAXINSTALList ≥newInstallist (17) In (18), the three different renewable scenarios considered in the study are forced to start at the same state of charge for each storage type, which is set to be 60% of their total installed capacity. socist,p,1,1,0=0.6×(INSTALLEDist +newInstallist) ×DISCHARHOURSist ∀p(18) Additionally, (19) guarantees that each technology considers its given yield when buying electricity. energySelli,p,w,m,h=energyBoughti,p,w,m,h×YIELDist ∀ist,p, w, m,h (19) 4. Case study and scenario definition This section describes the scenarios considered in the study. Since the study focuses on the cost competitiveness of the different firm capacity providers and how DR may impact the overall firm capacity needs and that competitiveness, four sets of scenarios have been built. These four blocks are characterized by the parameter which sensitivity is analyzed. These sets of scenarios and sensitivities addressed are presented in Table 13: A base case scenario is used as a reference for each scenario set. All base cases evolve from the baseline scenario. The baseline scenario assumes there are no DR capabilities in the system and looks for an optimal, minimum cost, mix of generation and storage technologies for the 2030 Spanish electricity system, assuming the following set of technical and cost parameters. As explained in Section 2, demand growth from 2015 data is set using an intermediate value between the minimum and maximum growth rates for each disaggregated category of demand. Table 14 summarizes the firm capacity coefficients assumed for each technology. These values are obtained from Red Eléctrica (2020), except for batteries with four hours of discharging rate (NationalgridESO,2020) and pumped hydro storage (National Grid,2017). The values corresponding to the candidate pumped storage hydro facilities are presented in Table 11. Table 15 summarizes the 2019 existing generation capacity expected to be still available by 2030. Values are extracted from PNIEC (2020). Table 16 presents the values assumed for the investment costs and the fix and variable O&M costs for both conventional and renewable technologies, updated from previous studies (Gerres et al.,2019) and based on additional Refs. International Renewable Energy Agency (2017), European Commision (2018), Allen (2017), Larsen and Rønnov (2018), The National Renewable Energy Laboratory (NREL) (2018) and PNIEC (2020). The values corresponding to storage facilities (including hydropower) are presented in Table 11. Table 17 summarizes the assumptions adopted for fuel prices, CO2emission costs, and taxes for pollutant technologies. Prices for CO2and gas are based on International Energy Agency (2019): 4.1. Baseline_NewFC scenario: Firm coefficient sensitivities The SPLODER model results may be pretty sensitive to the firm capacity coefficient (FC) parameter adopted for each technology. 10551 T. Freire-Barceló, F. Martín-Martínez, Á. Sánchez-Miralles et al. Energy Reports 8 (2022) 10546–10560 Table 13 Scenarios sets definition. Set Main feature analyzed Base case Sensitivities built upon base case scenario A Firm coefficient Baseline Baseline_NewFC B Percentage of DR deployment Baseline_NewFC(0%DR) 25%DR 50%DR C Reduction of battery price 25%DR PriceBatt_L PriceBatt_LL D CO2and gas prices Baseline EVC_XHigh EVC_High EVC_Low Table 14 Firmness coefficients. Technology Firm capacity coefficient Nuclear 0.97 OCGT 0.96 CCGT 0.96 Cogeneration 0.55 Biomass/Biogas 0.55 Solar thermal 0.14 Hydro (reservoir) 0.44 Hydro (run-of-river) 0.25 Existing pumped hydro storage 0.77 Solar photovoltaics 0 Wind power 0.07 Li-Ion batteries 0.69 Table 15 2019 existing generation capacity expected to be still available by 2030. Technology Installed capacity (MW) Nuclear power plants 3050 OCGT 0 CCGT 24,560 Cogeneration 3745 Biomass/Biogas 2146 Solar thermal 2299 Hydro (reservoir) 15,614 Hydro (run-of-river) 636 Existing pumped hydro storage 3329 Solar photovoltaics 8372 Wind Power 25,553 Li-Ion batteries 0 Table 16 2030 generation technologies’ costs. Investment costs (= C/kW) Annual fixed O&M cost (= C/kW-year) Variable O&M cost (= C/MWh) Nuclear – 108.3 – OCGT 544.1 18.4 11.0 CCGT 845.1 19.3 2.0 Hydropower (All) – 68.8 3.0 Solar PV (utility) 500 10 – Solar thermal 4396.6 49.6 0.46 Wind 950 29 – Table 17 2030 fuel costs, CO2 emissions costs and individual taxes. Fuel cost (= C/MWh) CO2cost (84 = C/tonCO2) Taxes (= C/MWh) Nuclear 8.72 – 15.02 OCGT 48.88 42.42 4.68 CCGT 32.58 28 4.68 Cogeneration – 48.78 – In a system dominated by renewable generation, it is quite often that scarcity periods last longer than 4 h (which is the charging and discharging cycle of batteries) and sometimes last even longer than 8 h (which is the charging and discharging cycle of 8 h pumped hydro storage) (Huclin et al.,2022). To be conservative, the security of supply of an electricity system should not fall upon storage with less than 10 h of storage capacity. For this reason, authors evaluated in Huclin et al. (2022) the firm coefficient of the different storages, coming up with lower FCs for batteries and Table 18 Firmness coefficients sensitivities. Scenario Batt_cent Sto_8h_1 Baseline 0.69 0.96 Baseline_newFC 0.294 0.567 Table 19 Scenarios with distributed PV panels under different DR percentages. Scenario DR Climate DR DHW DR EV DR REF 0DR 0 0 0 0 25DR 25% 25% 25% 25% 50DR 50% 50% 50% 50% 8 h cycle pumped hydro storage, and with these new FCs has been built another scenario, referred to as Baseline_newFC, being these values more accurate for future scenarios. Considering (Huclin et al.,2022;NationalGrid,2018), the new values adopted for these two technologies are shown in Table 18 and are used for all the rest of the scenarios. 4.2. DR scenarios: Percentage of DR sensitivities Three additional scenarios built upon the Baseline_NewFC scenario are considered to analyze the impact of DR on the firm capacity requirements of the system. The three scenarios include a fixed amount of solar distributed generation as explained below. Baseline_NewFC scenario neglected any DR capability in the system as the 0DR scenario does. Two additional scenarios (25DR and 50DR scenarios) are built for which, respectively, 25% and 50% of the total load identified as controllable (see Section 2) is considered ready to actively participate in demand response programs. These segments of load correspond to heating and cooling (climate), domestic hot water (DHW), electric vehicles (EV) and refrigeration (REF) as shown in Table 19. An amount of distributed small-size self-consumption solar PV generation is assumed to be already installed by 2030 for these three scenarios (0DR, 25DR and 50DR). A conservative fixed preset amount, shown in Table 20, has been considered. This amount has been estimated assuming that there will be a 25% of new installation capacity between the two extreme values found in the literature, a minimum growth by 2030 of 0.6 GW (Deloitte Advisory,2017) and a maximum one of 6.5 GW (Deloitte,2018a), upon the currently 1 GW installed capacity (Red Eléctrica de España,2019). This amount is geographically allocated by climate zones around Spain according to Red Eléctrica de España (2019) and associated with the three demand sectors (residential, services and industry) according to NationalgridESO (2020). Nevertheless, these distributed solar PV panels have a very marginal impact on the results of this study. They substitute utility-scale solar PV installation needs (actually at a slightly larger ratio than 1:1 as some network losses are avoided) but do not contribute to firm capacity. 4.3. Battery low price scenarios: Battery price sensitivities As results will show later, batteries are far from being competitive, provided the installation costs assumed in scenarios so 10552 T. Freire-Barceló, F. Martín-Martínez, Á. Sánchez-Miralles et al. Energy Reports 8 (2022) 10546–10560 Table 20 Assumed installed PV distributed capacity in 2030. Technology Installed capacity (MW) Solar distributed 2467 Table 21 Batteries installation costs. Scenario Batt_cent install cost [= C/kW] 25DR 133.8 PriceBatt_L 30 PriceBatt_LL 20 Table 22 CO2and gas prices scenarios. Scenario CO2[= C/tonCO2] Natural gas [= C/MMBTU] EVC_XHigh 83 27 EVC_High 62 18 Baseline 84 6 EVC_Low 90 4 far. Two additional scenarios (PriceBatt_L and PriceBat_LL) have been considered lowering the installation cost of batteries, as presented in Table 21. Both are built upon the 25DR scenario. The main purpose of the first scenario, PriceBatt_L, is to identify the battery installation cost threshold below which batteries begin to be competitive enough to be competitive. For that purpose, batteries installation costs have been reduced in steps of 1= C/kW until results show some battery installation, substituting OCGT. This happens for an installation cost decrease of 77%, as shown in Table 21. The second scenario, PriceBatt_LL, allows a larger penetration of batteries to understand the impact of the technological mix of renewables and the competitiveness of other firm capacity providers. This is achieved with a further reduction in batteries installation costs (85% of reduction), as shown in Table 21. 4.4. Equivalent variable cost scenarios: CO2and gas prices sensitivities CO2and gas prices have a substantial impact on electricity wholesale market prices. Therefore, a set of scenarios has been built to specifically assess their incidence on generation investment priorities. The Baseline of this set of scenarios assumes the same prices for gas and CO2in 2030 than all previously described sets of scenarios. They follow the values stated in the WEO2019 (International Energy Agency,2019) for 2030. Then, three sensitivities to these prices are performed. The EVC_XHigh scenario is built upon the first semester of 2022 average CO2 (SENDECO,2022) and natural gas (MIBGAS,2022) prices, which have historically influenced the market in the Iberian Peninsula. EVC_High scenario assumes extremely high prices for gas and moderately high prices for CO2, similar to those the world faced in the summer, autumn, and winter of 2021, taken respectively from MIBGAS (2021) and SENDECO (2021). It assumes those prices will remain similar in 2030. Finally, the EVC_Low scenario considers the more recent forecasts up to a day for CO2(Simon, 2021) and natural gas (Sönnichsen,2022) price evolution. These three scenarios provide a sensitive sensibility analysis of the gas and CO2emission prices. These values are presented in Table 22. Unit conversion types considered in these cases are: 1$ →0.84= C and 1 MWh→3.41MMBTU The price tendency of each of the three scenarios has been clarified by calculating the equivalent variable cost (EVC) in = C/MWh for the two-generation technologies affected by CO2and gas prices: OCGT and CCGT. EVC integrates into a single production cost per technology the actual impact of both the gas and Table 23 Equivalent variable cost for OCGT and CCGT. Scenario OCGT [= C/MWh] CCGT [= C/MWh] EVC_XHigh 258 167 EVC_High 180 115 Baseline 102 63 EVC_Low 90 54 CO2emission allowance prices. Table 23 summarizes the resulting EVC for both technologies. It provides sensible information on the assumption made in each scenario. The production cost of both technologies respectively increases and decreases in the EVC_High and EVC_Low sensitivity scenarios compared to the Baseline one. 5. Results This section presents and discusses the results provided by the SPLODER model. Results are organized following the sequence of the blocks of scenarios described previously. 5.1. Firm coefficients analysis Figs. 2 and 3show the generation and storage optimal investment decisions for the period 2019–2030 as provided by SPLODER for these scenarios. Namely, Fig. 2 displays the investments in renewable technologies (wind and solar PV) for both scenarios, while Fig. 3 shows the investments in the rest of the technologies, mainly oriented to provide firm capacity to the system (storage and thermal backup capacity). Fig. 2 shows that the new installed renewable capacity is very large (almost 73 GW) compared to the peak demand value for 2030, which is assumed to be 51.39 GW and given the initially existent installed capacity (see Table 15). This is because storage technologies are helping to decrease renewables’ spillage. Although installation costs are significantly lower for solar PV than for wind, investments in both technologies are balanced. This is partly because solar PV production is concentrated in fewer hours than wind production. Therefore, solar contribution to the system firm capacity is much lower (indeed, it is zero in this study, as the more stressful periods for the system happen during hours without sun). Besides, the concentration of the solar PV in the hours of sunshine ends up cannibalizing the energy income of solar PV. This effect is more significant than for wind since the latter has more variability at different hours. On the other hand, as shown in Fig. 3, batteries are fully discarded, not being competitive compared to other options. On the contrary, 5 out of 7 available pumped-hydro storage options are selected. The mix is completed with thermal OCGT backup generation to meet the system firm capacity requirements.2 By comparing in Figs. 2 and 3, the impact of considering stricter (lower) firm capacity coefficients for shorter-term storage technologies, that is, batteries and 8 h pumping storage, no very relevant changes are shown (Baseline_NewFC results compared to Baseline ones). Renewable investments almost do not change, although wind generation comes out slightly favored at the expense of solar PV. This is because 8 h pumping storage loses some competitiveness due to its reduced firm capacity coefficient, 2If investments in gas generation technologies are to be avoided to meet decarbonization commitments, the OCGTs investment would be replaced by the rest of available hydro pumping options and batteries if also needed. Nevertheless, OCGTs would almost not produce, being their role mainly to provide firm capacity. 10553 T. Freire-Barceló, F. Martín-Martínez, Á. Sánchez-Miralles et al. Energy Reports 8 (2022) 10546–10560 Fig. 2. New renewable investments (2019–2030) in baseline scenarios. Fig. 3. New firm capacity investments (2019–2030) in baseline scenarios. being thus replaced by OCGT. This reduces the system’s total storage capacity, which disfavors solar PV more than wind. Also, the lower the storage capacity, the higher the total firm capacity required in the system since storage supply peak demand. In this case, around 200 MW of additional firm capacity is needed. Wind power provides some firm capacity while PV solar does not. To complete the analysis of the competitiveness of firm capacity providers, a cost recovery analysis for new investments for the baseline_newFC is presented in Fig. 4. This figure shows, on the left-hand bar, the total annualized costs faced by the technology, disaggregated into investment, fuel, O&M and CO2 emissions costs, and on the right-hand bar the total incomes received by the technology, disaggregated into the incomes from the hourly energy production valued at the hourly energy price and the Fig. 4. Cost recovery for new firm capacity investments (2019–2030) in Baseline_NewFC scenario. required income from the firm capacity provision to balance costs and incomes. It could be observed first that all selected technologies do recover their total costs considering an income due to firm capacity provision equal to the missing money for each technology. Indeed, the capacity payment mechanism would follow a marginal price approach for the provision of firm capacity. The incomes from capacity payments would equal or exceed these values for the selected technologies, as OCGTs and the 8 h pumped-hydro storage are the marginal technologies providing firm capacity. OCGT incomes fully come from the provision of firm capacity showing that its role in producing energy will be very marginal.3 On the contrary, although they are also providing firm capacity, pumped-hydro storage technologies do have some income by operating in the energy market.4The energy market-related incomes may recover up to 60% of their total costs for some of these technologies (for instance, the Sto_40h_1), but only 22% for others. The more hours the pumped-hydro storage can store, the more it would operate and more income from the energy market. The results show the relevance of capacity payments to ensure investments cover the system firm capacity requirements. 5.2. DR analysis Fig. 5 shows the investment decisions in renewable technologies, and Fig. 6 those in firm capacity providers’ technologies 3The production of such peaking units is however somehow underestimated in models such as SPLODER since a fully stochastic approach will reveal more situations where these back-up technologies will produce some energy. 4These figures are slightly underestimated due to the same reason as for OCGTs. See previous footnote. 10554