Forecasting Renewable Scenarios and Uncertainty Analysis in Microgrids for Self-Sufficiency and Reliability: Estimation of Extreme Scenarios for 2040 in El Hierro (Spain)

Abstract

This study evaluates the feasibility of fully renewable energy systems on El Hierro, the smallest and most isolated Canary Archipelago Island (Spain), contributing to the broader effort to decarbonize the European economy. By 2040, the island’s energy demand is projected to reach 80–110 GWh annually, assuming full economic decarbonization. Currently, El Hierro faces challenges due to its dependence on fossil fuels and inherent variability of renewable sources. To ensure system reliability, the study emphasizes the integration of renewable and storage technologies. Two scenarios are modeled using HOMER Pro 3.18.4 software with probabilistic methods to capture variability in generation and demand. The first scenario, BAU, represents the current system enhanced with electric vehicles. While the second, Efficiency, incorporates energy efficiency improvements and collective mobility policies. Both prioritize electrification and derive an optimal generation mix based on economic and technical constraints, to minimize Levelized Cost Of Energy (LCOE). The approach takes advantage of El Hierro’s abundant solar and wind resources, complemented by reversible pumped hydro storage and megabatteries. Fully renewable systems can meet demand reliably, producing about 30% energy surplus with an LCOE of roughly 10 c€/kWh. The final BAU scenario includes 53 MW of solar PV, 16 MW of wind, and a storage system of 40 MW–800 MWh. The Efficiency scenario has 42 MW of solar PV, 11.5 MW of wind, and 35 MW–550 MWh of storage. Uncertainty analysis indicates that maintaining system reliability requires an approximate 10% increase in both installed capacity and costs. This translates into an additional 7 MW of solar PV and 6 MW–23.5 MWh of batteries in the BAU, and 6 MW and 4 MW–16 MWh in the Efficiency.

1. Introduction

1.1. Background on Renewable Energy Systems: The Path to Sustainability

The IEA report “World Energy Outlook 2023” [1] states that global energy demand has uninterruptedly increased over the last few decades, except for the interruption caused by the COVID-19 pandemic in 2020. However, the upward trend resumed in 2021 despite the ongoing pandemic [2]. Historically and up to the present, most of this produced energy comes from fossil fuels, and similar figures are observed in electricity generation. It is true that in recent years, various countries have been investing in the implementation of different renewable energy generation sources. Nevertheless, nearly two-thirds of energy is still generated through fossil fuels [3], although these figures are gradually decreasing year by year [4].

As many experts emphasize, reliance on fossil fuel-based generation is unsustainable and poses two fundamental problems. Firstly, the inevitable depletion of fossil fuels in the medium term if the current consumption rate is maintained [5,6]. The second problem is the pollutant emissions, especially the substantial release of GreenHouse Gases (GHGs) during energy production when fossil fuels are used as fuels or raw materials [7,8].

Therefore, there are many reasons for renewable energies to be present or even to be the only sources of generation used, with the aim of reducing or eliminating fossil fuels and their consequent emissions [9]. Focusing on electricity generation, the use of renewable energies is a necessity, as otherwise, it is impossible to achieve the ambitious GHG emissions reduction targets [10]. Furthermore, the sharp increase in electricity use, which encompasses the final energy consumption of all countries [11], is expected to exceed 30% of the total amount in a short time in many of them [12].

The described problem is further complicated in remote regions, such as islands [13]. This is because their small size and inaccessible location make connecting to a large grid difficult or even technically or economically impossible. Therefore, in the case of islands, the typical solution is to have an energy system based on fossil fuels (coal, gas, and/or diesel), as these systems have the advantage of high reliability, which is needed for systems that cannot rely on a grid to avoid supply interruptions. However, the disadvantages of these systems include significant emission problems and a strong dependence on a large and complex supply chain. Additionally, the countries producing these fuels are often unstable, posing the risk of shortages that can reduce system reliability, compounded by the inconvenience of fuel prices experiencing large fluctuations due to frequent and unexpected price hikes caused by cartel decisions [13].

Consequently, from both environmental and strategic (energy autonomy) points of view, renewable energies are an option that will be implemented in many places in the near future; indeed, they already exist to varying degrees of implementation in many locations [14,15]. However, when renewable energies take on a significant role in power generation, a series of challenges arises, primarily associated with the intrinsic variability and unpredictability of these sources [16,17]. Specifically, the two renewable sources currently capable of meeting existing energy needs, solar PhotoVoltaic (PV) and wind power, exhibit strong variability, with longer or shorter periods of low or strong wind, solar cycles, cloudy or rainy days, etc. This means that to meet energy needs with these sources, the system must be oversized and include large storage systems to absorb at least part of the excess and make it available when needed [18]. Even so, some degree of energy surplus will inevitably occur, although in this situation such excesses are expected to be more limited [19]. Throughout the system’s operational lifetime, there will be phases characterized by substantial surpluses and others in which the system operates near its capacity limits. These fluctuations will be further affected by the gradual degradation of performance and potential reliability concerns over time [20]. Consequently, the optimal sizing of generation and storage components should be determined primarily from an economic standpoint; that is, by installing generation capacities and storage systems capable of meeting energy demands at the lowest possible cost. On islands, however, storage facilities (such as pumped-hydro stations or large-scale battery systems) and extensive renewable generation installations (wind or solar PV) face additional challenges arising from the scarcity of suitable locations and restrictions linked to environmental protection, tourism activities, and the competing use of land or marine resources by local populations [21].

1.2. Specific Challenges of Islands and the Particular Case of El Hierro

The Canary Islands, located between 100 and 300 km from the Moroccan coast and about 1500 km from mainland Europe, form an archipelago of seven inhabited islands with a population exceeding 2 million [22]. The region’s total electricity demand has remained stable at around 9 to 9.5 TWh per year, aside from the temporary decline during the COVID-19 pandemic. Despite significant renewable potential, the share of clean energy remains modest. In 2023, renewables contributed about 2 TWh—more than 20% of total generation—with wind energy covering roughly 80% of that amount and solar PV accounting for most of the remainder [22]. Within this context, El Hierro—the smallest of Spain’s Canary Islands—accounts for just 0.5% of the archipelago’s total electricity consumption, with an annual demand of around 50 GWh.

El Hierro stands as a European pioneer in the transition toward renewable and self-sufficient energy systems. Through the integration of wind, solar PV, and hydropower, the island has become a reference model for sustainable energy in isolated territories [23]. Although full energy autonomy has yet to be reached, the commissioning of the Gorona del Viento hydrowind plant over a decade ago marked a milestone in demonstrating that even small, remote communities can effectively lead decarbonization efforts.

Spain, in line with European Union and Paris Agreement commitments, has pledged to achieve climate neutrality by 2050, balancing GHG emissions with removal actions [24]. The National Integrated Energy and Climate Plan (PNIEC 2021–2030) sets an interim target of a 23% emissions reduction by 2030 compared with 1990 levels [25]. Within this framework, the Canary Islands—rich in solar and wind resources—are moving forward under an accelerated timeline. The regional PTECan strategy aims to complete decarbonization by 2040, ten years ahead of the national and EU objectives, reinforcing the archipelago’s role as a testing ground for advanced renewable energy systems [26,27].

As a consequence, El Hierro serves as a pioneering testbed for the large-scale integration of renewable energy in isolated power systems. Since 2014, the island has operated an autonomous, zero-emission electricity system that combines wind generation with a reversible pumped-storage facility. Although the project’s conceptual development began before the island’s designation as a UNESCO Biosphere Reserve in 2000, this status has further reinforced its sustainability objectives. The initiative, known as the “Hydro-Wind Development of the Island of El Hierro”, was undertaken by Gorona del Viento El Hierro S.A. (GDV), a consortium formed by the Island Council of El Hierro, UNELCO S.A., the Canary Islands Government, and the Instituto Tecnológico de Canarias [23].

The system’s design enables the wind farm to supply most of the island’s demand while channeling surplus energy into a pumped-hydro system. Water is transferred from a lower to an upper reservoir when excess electricity is available and released to generate power during low-wind periods. This hybrid configuration effectively converts an intermittent resource into a controllable and continuous supply, representing a significant innovation in the renewable energy sector.

Nevertheless, in periods when renewables cannot fully meet demand, backup electricity is supplied by the diesel-powered Llanos Blancos plant, owned by ENDESA. Although the hydrowind system was initially expected to satisfy up to 75% of the island’s total demand, operational results over nearly a decade indicate an even split between renewable and diesel generation, each providing roughly 50% of El Hierro’s electricity.

1.3. Overview of Research Organization and Major Contributions

Proper energy system sizing is essential for isolated grids, where reliability is essential. The article develops an uncertainty-based modeling framework aimed at assessing the optimal energy generation mix required to meet the projected electricity demand, applying it to the case study of El Hierro Island in 2040. As previously mentioned, the study performs predictive analyses of future demand and evaluates the most likely extreme scenarios expected to occur by that year, comparing their outcomes. In general terms, most previous studies have relied on deterministic analyses. These typically simulate different scenarios under specific assumptions—such as maximizing renewable generation, minimizing system costs, or reducing surpluses—usually using representative or typical input data for resources and demand (e.g., typical data, adverse behaviour, or average values). The main innovation of the present research lies in the implementation of Best Estimate Plus Uncertainty (BEPU) techniques through the application of Wilks’ formula. This approach enables the probabilistic estimation of the required generation and storage capacities, ensuring the desired reliability levels (in this study, coverage and confidence levels of 95%/95%). Ultimately, this probabilistic framework shifts from traditional cost-based deterministic optimization to a reliability-oriented design of energy systems. It accounts for meteorological variability and the evolution of demand, thereby providing a more robust foundation for the sizing of systems in isolated grids such as those of the Canary Islands.

The first, a Business-as-Usual (BAU) scenario, assumes continuity of current practices while electrifying all sectors, including transport. The second scenario emphasizes energy efficiency, Demand-Side Management (DSM), and sustainable mobility, reducing consumption and emissions through collective transport and other low-carbon measures. Together, these represent the two most plausible extreme cases for a fully decarbonized island economy, incorporating the variability and uncertainty inherent in the analyses.

Unlike the Canary archipelago as a whole [28,29,30], El Hierro’s path to full electrification does not require alternative energy carriers, as the island lacks energy-intensive industries and long-distance heavy transport requirements. The analysis draws on historical solar and wind resource data, selecting the most adverse conditions to ensure the proposed energy mix maintains the required reliability levels at 95% coverage and confidence thresholds.

The BAU scenario projects the highest energy demand, while the efficiency-focused scenario significantly shrinks the required generation mix through DSM implementation and sustainable mobility measures. Both cases are evaluated to the 2040 horizon, aligning with the Canary Islands’ decarbonization timeline—ten years ahead of the EU’s target.

Optimization criteria considered to reach the optimal scenario include economic feasibility, as well as restrictions related to land occupation due to the presence of protected areas. In addition, land occupation considerations have been taken into account, as well as the maximum installation capacities of certain technologies. For example, the maximum power of solar PV generation on rooftops has been analyzed to avoid the increase in visual pollution caused by solar farms.

As previously mentioned, all these considerations have been applied to a practical case study involving a small island completely isolated from any larger electrical grid and with no short- or medium-term prospects of being connected to one, such as El Hierro. Similarly, the main conclusions are directly transferable to other “comparable” locations, that is, those with similar renewable resources and demand profiles, and without connection to a central electrical grid. Moreover, the developed methodology can be applied to any other site, considering only its specific local characteristics. Therefore, our methodology is fully scalable and replicable.

To achieve the objectives outlined, Section 2 presents a complete overview of the current electricity supply situation on El Hierro island. Section 3 details the methodology developed for this analysis, offering context for the study and thoroughly describing the characteristics and necessary data of all relevant systems. These include both generation and storage technologies required to conduct precise simulations within the defined timeframe. Section 4 presents the main simulation results, along with a thorough analysis and discussion of the findings. The section focuses on the forecasting analysis of electric demand and renewable resource availability, considering the two specific scenarios proposed. This section also presents the main simulation results, along with a thorough analysis and discussion of the findings. Finally, Section 5 summarizes the key conclusions derived from the study regarding the generation system and outlines potential directions for future research.

2. The Electric System of El Hierro

The Canary Islands face structural limitations in available generation technologies due to their isolation and the small scale of each island’s power system. Only Lanzarote and Fuerteventura are interconnected, making fossil fuels the dominant source to ensure reliability [22].

El Hierro exemplifies these challenges. Historically dependent on diesel generators, the island sought energy independence with the 2014 commissioning of the Gorona del Viento hydro-wind plant [23]. However, over ten years, fully renewable operation has averaged fewer than 100 days annually, though around 50% of total yearly demand has been covered by renewables.

These systems also require high margins for the peak demand-to-installed power ratio, approximately 25% in recent years. Moreover, considering the peak demand-to-actual available power ratio, this becomes significantly higher at many times, around 40% to 70% [31]. Their design and operational criteria therefore differ substantially from mainland grids, emphasizing reliability over cost-efficiency. Seasonal tourism-driven demand and strong environmental constraints further raise generation costs, with roughly 75% of expenses being variable due to costly imported fuel and transportation, while investment and Operational and Maintenance (O&M) costs play a lesser role.

2.1. The Electricity Demand

Hourly electricity demand data for El Hierro were obtained from the Spanish market operator Red Eléctrica de España (REE) [32], available from 1 January 2014, onward. Historical records indicate remarkable stability in the island’s electricity consumption over the past twenty years, averaging around 40–50 GWh annually [22].

A modest long-term growth trend is observed: demand rose from below 10 GWh in 1985 to approximately 40 GWh by 2005, reaching around 50 GWh in recent years. This stable demand pattern is consistent with the trends in energy production displayed in Figure 1, which shows that between 2010 and 2022, total electricity generation on the island has remained largely constant, with no major fluctuations in output.

El Hierro has the warmest climate in the Canary Islands, with mild temperature variations year-round. Winters average 18–19 °C and summers 24–25 °C, with only a few cooler days between December and March. Because of this thermal stability, electricity demand is only weakly influenced by climate.

Analysis of hourly demand patterns indicates that consumption is lowest in winter (December–February), highest in summer (July–September), and moderate in spring and autumn (Figure 2). These variations are mainly driven by tourism, which peaks during summer. The daily load curve follows the typical profile of residential and service sectors, showing two demand peaks—one in the morning and another in the early evening—and a drop during late-night hours.

Data from the Spanish Electricity Grid (REE) [32] for 2014–2024 reveal three daily demand categories, ranked by consumption level: holidays and Sundays (lowest), Saturdays and Mondays (intermediate), and midweek days (highest). Seasonal analyses also identify three main demand patterns, illustrating both daily and monthly variability linked to tourism dynamics and mild climatic conditions. Another aspect is the three seasonal patterns shown in Figure 3, which display the hourly demand profiles for typical season days of 2023 (Figure 3a) and the monthly variability over the whole hourly database between 2014 and 2024 (Figure 3b).

2.2. The Electricity Generation System

El Hierro, the most remote of the Canary Islands from the African coast, has maintained a population of fewer than 11,000 residents for decades. Recognized by UNESCO as a World Biosphere Reserve in 2000 [33], the island has long pursued a model of sustainable development. This commitment was reinforced by the Sustainability Plan adopted on 22 November 1997, which set a clear goal of transforming the island’s energy system toward full sustainability [23].

Strong local environmental awareness, together with the coordination of governmental and institutional actors, enabled the creation of the Gorona del Viento wind-hydro project—an initiative designed to achieve 100% renewable electricity supply. The project is managed by a consortium comprising the Island Council of El Hierro, the regional electric utility, and the Government of the Canary Islands [23].

2.2.1. The Diesel Generation Plant

Historically, electricity on El Hierro has been generated using fossil fuel-based systems, primarily small-scale diesel units housed at the Llanos Blancos Thermal Power Plant in the municipality of Valverde. The facility currently operates ten generators with net power outputs ranging from 670 to 1900 kW and gross outputs between 780 and 2000 kW, providing a total net capacity of 13.04 MW and a gross capacity of 14.91 MW [34]. The plant’s maximum net generation capacity, entirely diesel-fueled, therefore stands at 13.04 MW, with a net utilization factor of about 40% in recent years.

Following the commissioning of the hydro-wind power plant, annual electricity generation from the diesel units has declined notably—from around 40 GWh to approximately 20–25 GWh per year (Figure 4a). This transition has reduced the island’s dependence on fossil fuels to roughly 50%, marking significant progress toward its renewable energy objectives [22].

El Hierro’s aging diesel generators, averaging over 25 years old [34], supply the island’s electricity from the Llanos Blancos plant. Power is stepped up to 20 kV for distribution across the island. The facility receives fuel via a pipeline from DISA, stored in two main 250 m³ tanks and three smaller daily use tanks [34].

2.2.2. The Hydrowind Power Plant

The Gorona del Viento hydrowind complex is situated in the municipality of Valverde, between Pico de los Espárragos and Llanos Blancos [35]. It comprises an integrated system that combines wind and hydroelectric generation: an 11.5 MW wind farm with five Enercon E-70 E4 turbines (Enercon GmbH, Aurich, Germany), an 11.3 MW hydroelectric plant with four Pelton turbines (Pelton Water Wheel Company, San Francisco, CA, USA) of 2.83 MW each, and a 6 MW pumping station. Together, they provide a total energy storage capacity of approximately 225 MWh, enabling flexible operation and improved integration of renewable generation on the island.

The operation of the Gorona del Viento system adapts dynamically to available generation capacity. When wind conditions are favorable, the wind farm supplies electricity to the island and uses any surplus to pump water from the lower to the upper reservoir [23]. During calm or low-wind periods, water is released from the upper reservoir, descending nearly 655 m to drive the Pelton turbines and generate electricity, compensating for reduced wind output.

Recent performance data show that the wind farm achieves around 3000 equivalent operating hours per year, producing between 30 and 40 GWh, while the pumping process consumes about 20 GWh. As a result, the net annual energy contribution from the hydrowind system ranges from approximately 20 to 24 GWh. Following testing and commissioning in 2014–2015, the plant has been fully operational and integrated into El Hierro’s grid since 2016. The data depicted in Figure 4b illustrate this evolution, with a marked rise in production during the commissioning phase and stabilization thereafter.

As illustrated in Figure 4a,b, renewable energy production on El Hierro increased dramatically from nearly zero to about 50% of total generation between 2014 and 2016, following the commissioning of the Gorona del Viento hydrowind power plant. Since 2016–2017, the renewable share has stabilized at this level, remaining below the project’s target of 73.4%.

However, significant energy losses have been observed due to system losses, Figure 4b. The wind farm’s 11.5 MW capacity exceeds the 6 MW pumping capacity, preventing full utilization of wind generation peaks. This limitation, particularly evident during low-demand nighttime hours and the winter season, often results in storage saturation when the upper reservoir reaches full capacity. As a result, excess renewable energy must be curtailed, with total losses estimated at 30–40% of generation over nearly a decade of operation [23].

3. Methodology

The methodology section includes not only the approach of the analysis using the software, but also the major points, which are those aspects related to the estimation of the future demands in the year of analysis, as well as the ones related to the availability of existing resources. As mentioned throughout the introduction, the development and practical application of these methodologies for demand forecasting and energy resource behavior are central to this work, representing a significant innovation with the application of a BEPU approach. In contrast to many studies that typically model using resource data from a “typical year”, just replacing some or all fossil fuel-based generation with renewables or meeting current demands with such technologies, this research extends far beyond exploring the areas discussed above.

3.1. The Possible Simulation Tools

The literature shows that several simulation tools, such as HOMER, EnergyPLAN, H2RES, MARKAL, and LEAP, are effective for energy planning at various scales, from small systems to national levels [36,37]. These tools are particularly useful for modeling scenarios with high levels of Variable Renewable Energy (VRE), typically using hourly time steps for one-year simulations. Consequently, several tools were considered suitable for implementing this work, with HOMER chosen for its detailed analysis capabilities. The software, called Hybrid Optimization Model for Multiple Energy Resources (HOMER), has been developed by the National Renewable Energy Laboratory (NREL) [38]. It is a widely recognized tool for simulating the operation of national or regional energy systems on an hourly basis [11,39]. It is valued for assessing the energy, environmental, and economic impacts of various energy strategies, enabling the comparison of multiple options. Additionally, HOMER addresses factors like reducing installed power, minimizing energy waste, and managing land use or visual pollution.

In this study, a statistical analysis of resource behavior was conducted using the available historical data, estimating the Probability Density Functions (PDFs) for solar and wind resources. A similar analysis was performed for the demand history, estimating future behavior assuming no significant changes in consumption patterns. This allowed for estimating the behavior of a BAU scenario without demand reduction measures and involving huge EV penetration. Different charging profiles for this technology were analyzed to estimate the likely shape of the demand curve. On the other side, a scenario with demand reduction measures and also involving huge EV penetration, including efficiency measures, collective mobility measures, among other policies. Consequently, the main characteristics of the resources available for electricity generation and the expected demand for the analyzed year were determined. These estimations were made on an hourly basis, allowing for the hourly balancing of generation and demand in each of the presented scenarios, as will be described in more detail later.

Finally, the subsection on the presentation of the HOMER not only describes the general characteristics of the HOMER, but also provides the information that must be entered into the code in order for the HOMER to perform its calculations. Information regarding technical characteristics and costs of the different technologies, in the current case information on wind turbines and the existing pumping station in Gorona del Viento, as well as other technologies to be used to achieve energy independence with zero emissions, which in the current case correspond to the use of solar PV generation and additional use of storage through the installation of mega-battery systems. It must also be provided with the information related to the demand, in the case of the software used, breaking down individually the consumption of Electric Vehicles (EVs) of the remaining electrical consumption, so it should be provided both demand curves. In the current case, the balances have been made hourly, so the information on resources and demand must be provided hourly.

Therefore, this section of methodologies has been divided into three subsections, the first one analyzes the estimations of wind and solar resource conditions, the second one shows the demand estimates, the third one displays the technical information of the different subsystems and the fourth one describing the key aspects related to the HOMER code.

3.2. Monte Carlo Analysis Through the Wilks’ Formula

No modeling technique is free from uncertainty, so it is important to consider it when making decisions. Approaches to managing uncertainty range from alternative modeling methods to stochastic programming, where Monte Carlo Analysis (MCA) stands out as a key evaluation tool. This approach involves simultaneously propagating the model or scenario’s input parameters through to the output variables. Inputs are typically represented by probability distributions, bounded by realistic minimum and maximum values. After assigning distributions to each input, the model is run multiple times, each time using a different set of inputs. The outcomes yield varying outputs or variables of interest.

Typically, pure Monte Carlo methods require hundreds or even thousands of model runs. However, computational time can become prohibitive, so a smaller number of runs is often needed. Wilks’ formula, developed in the 1940s, offers an alternative with reduced computational demand [40,41]. Wilks devised a method to determine the sample size required to establish one-sided or two-sided tolerance limits at a specified confidence level. Initially applied in the nuclear safety field, it has also been adopted for uncertainty analysis in modeling. For example, Wilks’ formula has been used to assess uncertainties in various energy systems, such as the future Lithuanian energy system [42], and at an island level, to evaluate energy generation and future electricity and/or hydrogen demand in the Canary Islands [17,30].

Wilks’ sample size estimation for two-sided tolerance intervals, with a probability (γ) of falling within the two-sided confidence interval (β), is given by Equation (1):

$$n=\min \left\{n\in \mathbb{N}|\left(n-1\right){\gamma }^n-n{\gamma }^{n-1}+1\ge \beta \right\}$$

(1)

Consequently, the desired confidence and coverage levels must be set. The first one expresses the degree of certainty or probability with which the true value of a population parameter is expected to fall within the interval calculated from a sample. While the second one, the coverage level, refers to the empirical or theoretical proportion of times that, when repeating the sampling and constructing confidence intervals, those intervals actually contain the true value of the parameter. In summary, the system’s different performances would be captured, taking into account the variability in the behavior of the uncertain variables considered, thereby allowing the system to be sized to operate with the desired levels of confidence and coverage. Thus, for the typical levels of coverage and confidence that prevail in many industrial fields, i.e., 95% for both confidence and coverage levels (β = 0.95 and γ = 0.95), based on Equation (1), the number of simulations required to establish unilateral and bilateral tolerance intervals is calculated as n = 59 and n = 93, respectively. When the desired coverage and confidence levels reach 99% for both, the number of simulations required increases to n = 459 and n = 662 for unilateral and bilateral tolerance intervals, respectively. It is essential to note that these formulas are only valid when the tolerance interval is applied to a single output variable or to several independent variables. This balance between computational effort and statistical rigor makes the approach well suited for energy system modeling under uncertainty.

3.3. Forecasting the Weather Conditions

Referring to the methodology described earlier, it is necessary to estimate the performance for the uncertain input parameters related to the two renewable resources used in this work, specifically wind and solar resources. This requires understanding the behavior of wind and solar irradiation. Historical data on solar irradiance and wind speed from PVGIS [43] for the period from 1 January 2005 to 31 December 2024 are utilized. Hourly values of these variables are taken to estimate their respective hourly curves.

HOMER requires a text file, typically containing hourly values of different variables which show marked variability: 8760 values for common years and 8784 for leap years (as in the case studied here, year 2040). The input text file consists of the hourly mean values of the variable in question. Specifically, a complete synthetic year is constructed for each of the 93 synthetic time series, with a 95/95% confidence and coverage level in the outputs, according to Wilks’ formula and the BEPU methodology applied in this study (Equation (1)).

To simulate hourly variability in wind and solar generation, a block resampling method based on a conditional bootstrapping framework was used [44]. The previously mentioned historical dataset was employed. Generating synthetic time series or cases is necessary to evaluate the variability or randomness in power generation, particularly for renewable energy sources.

For this method, the historical dataset consisting of 20 years in the current analysis is divided into daily blocks. Each block represents the same calendar day from every year in the dataset, containing 24 hourly values of the resource, such as irradiance or wind speed. For each day of the synthetic year, its corresponding month is identified. Then, a day is randomly selected from the same month across any of the historical years. All hourly values of the selected day are placed in the corresponding position of the synthetic year. This produces a sequence of capacity factors recorded but randomly ordered. This approach preserves daily and seasonal patterns in the sampled data while generating a broad set of possible annual generation profiles that reflect the empirical distribution of historical capacity factors.

3.4. Forecasting of Electricity Demand Curves

In order to estimate the electricity demand forecast for El Hierro island by 2040, it was considered that the mean hourly demand will have a similar shape as the actual one. To perform the analysis of the current demand shape calculations, the demand curve has been considered but split by sectors (residential, retail, industrial, public administrative, housing and other uses) [45,46,47]. Next, the current hourly demand of each sector has been multiplied by a factor that takes into account its estimated increase or decrease up to 2040. These curves are multiplied by the factor associated with economic growth from now to 2040 (considering the increase in consumption due to economic growth).

A similar procedure for estimating the demand curves to the one mentioned for solar and wind generation has been carried out, except that in this case the REE data for the island is available from 2014 to the present, so that the 11 years available have been used to characterize the demand. The only exception is the use of the coefficient that takes into account the increase or decrease in demand according to the scenario considered (basically BAU or application of consumption reduction measures). In other words, the curves will be similar to the current ones, but scaled according to the considerations of demand evolution up to 2040, except for the contribution of the EV fleet.

For the trend estimation of electricity demand, the Random Forest technique is used, taking Gross Domestic Product (GDP) and population as factors (having reached high correlations with both variables, close to 90%, although with greater weight of GDP) considered as a growing series as a result of GDP and, to a lesser extent, population [46]. Reaching an increase from 30–70% over current demand figures, i.e., consumption in El Hierro will increase to about 80–110 GWh by 2040.

But in addition to the trend demand, it is important to know the demand that would be obtained by including energy efficiency policies. The mandatory policies established in the framework of the PNIEC 2023-30 [25] and PTECan [27] are taken as a basis. In this case, with the adoption of the maximum optimization measures, approximately slightly reducing the current demand, specifically a demand of 52 GWh, while a moderate increase up to around 65 GWh if efficiency policies and the rest of environmental policies are not implemented in such an effective way.

Regarding the increase in electricity demand, considering the EVs’ effect, the projected growth of the vehicle fleet was estimated using advanced statistical regression techniques, utilizing the historical evolution of the vehicle fleet, population, and the gross domestic product of the islands as explanatory signals [47]. Additionally, these estimates consider both conservative and more disruptive projections of the evolution of collective transport, achieving a reduction rate of the number of vehicles per inhabitant of 25% (from 0.862 to 0.626 vehicles per inhabitant) by 2040 compared to 2020, considering reductions due to the strong implementation of collective transport measures, sustainable mobility, etc. Then, considering the full penetration of EVs in road transport and a conservative ratio of vehicles per inhabitant (similar to the current one), estimates suggest around an 80% increase compared to current electricity demand values. Specifically, an EV demand of 43.34 GWh per year, i.e., reaching a total electric demand of approximately 108.35 GWh in 2040 [47]. Conversely, using the low ratio of vehicles per inhabitant and considering that the electrification of road transport would be associated with light vehicles other than buses and trucks, where these solutions present certain problems, the increase will reduce to around 31.50 GWh/year [46], i.e., a total electric demand of about 83.5 GWh per year.

These consumption figures are based on the estimates presented in

Table 1. Summary of the EV Fleet Characteristics assuming 100% electrification and the conservative tendency (based on [47,48]).

Vehicle Type	Number of Vehicles	Consumption (kWh/km)	Average Distance (km/day)	Charge/Discharge Power		Capacity		Yearly Consumption (GWh)
				Unitary (kW)	Total (MW)	Unitary (kWh)	Total (MWh)
Car	5618	0.15	50	3.7	20.79	80	449.4	18.49
Motorcycle	802	0.06	20	3.7	2.967	20	16.04	0.373
Van	1627	0.18	60	3.7	6.02	100	162.7	6.425
Bus	47	0.95	300	50	2.35	250	11.75	4.913
Truck	1885	0.64	30	50	94.25	300	565.5	13.14

. The main characteristics of the different types of vehicles for the year 2040 are shown, considering the specific conditions existing on the island of El Hierro. Because of the island’s small size, EVs there typically cover only short distances.

However, it is not only important to quantify the demand associated with the use of EVs but also to determine when this recharging occurs. Figure 5 illustrates the projected normalized curves of the hourly EV recharging profile proposed for 2040, split into light private vehicles and the rest of EVs (heavy goods and passenger transport, and light commercial and passenger vehicles, as well as agricultural machinery) [47,49]. The 95% confidence interval for the probability functions of both contributions is also presented. This charging profile has been characterized according to the estimated consumer habits, assuming average behaviors based on the charging point to which they are connected. Thus, the provided typical profile has been obtained by weighting the recharging habits of light private vehicles joined to light commercial, public service vehicles, together with heavy transport and public service vehicles. The major contributions of light private vehicles are related to the usual habits in residential parking lots, workplaces, hotels, shopping centers, regulated parking areas, and service stations. While the rest of the vehicle categories basically have a nighttime charging pattern, outside normal working hours, and in almost all cases at the company site or in some cases at the worker’s home. Thus, as can be seen in Table 1, approximately 43.5% of the consumption is for light private vehicles and the remaining 56.5% for the rest of the vehicle types. These profiles exhibit a considerable degree of generality applicable to various locations, primarily showing variations attributable to the specific customs of the population in the analyzed area. Therefore, although generally analogous in most countries, minor adjustments may be necessary to accommodate the idiosyncrasies of local inhabitants’ behavior.

Charging behavior is strongly influenced by seasonal patterns, with demand reaching its highest levels during the summer months as a result of increased travel activity. According to data from Spain’s General Directorate of Traffic (DGT) [50], there is a marked surge in traffic during the summer. Tendency which is driven by individuals heading to coastal destinations, rural retreats, and second homes, a tendency which also appears in El Hierro Island, since tourism has its peak in the summer months. These pronounced seasonal fluctuations play a central role in ensuring the accuracy of annual demand forecasts and the effectiveness of energy planning strategies. Figure 6 illustrates the average monthly variations observed in 2021, highlighting the impact of these seasonal trends on charging demand. The monthly variability of the peninsular traffic is quite similar to El Hierro’s demand variations shown in Figure 3b, which are mainly due to the seasonality of tourism, which has a significant effect on vehicle use. Therefore, this monthly variability has been considered in the prediction of demand in 2040.

The demand behavior profiles assuming continuation of current trends with total EV penetration are shown in Figure 7. It presents the hourly profile of a typical winter and summer day for the expected demand in 2040 on El Hierro. These profiles have several components: the first is the evolution of the current demand profile scaled by expected increases mainly due to population and GDP growth (orange line). Added to this is the expected evolution assuming full adoption of EVs for land transport, with the vehicle fleet continuing current trends. This contribution, as shown in Figure 5, consists of two subcomponents: private light vehicles (gray line) and other vehicles (yellow line). This scenario and an additional environmental measures scenario, including efficiency and promotion of collective transportation, will be described in detail ahead.

When analyzing the predicted total demand curves against current ones, the late afternoon peak is still present and even more pronounced. The nighttime minimum remains but is now less marked, meaning the curve has slightly flattened. The typical midday dip has also partially smoothed out, now staying near morning values. Overall, there has been a strong increase in demand mainly caused by the emergence of electric vehicles and also a certain flattening of the hourly demand curve.

To provide an initial overview of the general hourly demand behavior, Figure 7 displays the final shape of the averaged hourly demand curves. However, these hourly curves are not used directly in the analyses as the are averaged values, but rather the hourly values of the hourly curves that are sampled from the solar and wind resource PDFs and the demand curves are used to capture the variability and uncertainty of the hourly demand curves.

Specifically, seasonal effects are incorporated by applying the appropriate monthly coefficients. Additionally, uncertainties related to both inter-day and intra-day variability within each week are considered. This is achieved by considering the behaviour for all 52 weeks, using historical data downloaded from the Spanish electricity system operator, REE [32]. This procedure captures the uncertainties for both the EV contribution and the current baseline demand, and are applied to both scenarios under study.

3.5. The Generation and Storage Systems

As is widely known, at present, renewable energy generation on a significant scale is focused on the exploitation of wind and solar resources through the use of wind turbines and solar panels. In the particular case of the island of El Hierro, as mentioned in previous sections, the Gorona del Viento hydrowind power plant has been in operation since 2014. This system consists of five Enercon E-70 E4 wind turbines of 2.3 MW of unit power, totaling a power of approximately 11.5 MW. The characteristics of the wind turbine are detailed in

Table 2. Datasheet of the wind turbine [51,52].

Wind generator	Enercon E-70 E4
Rated power (MW)	2.30
Rotor diameter (m)	71
Height to the hub (m)	98 m
Total height (m)	133.5 m
Cut in wind speed	2.5 m/s
Cur out wind speed	34 m/s
Lifetime (years)	25
Cost of the system (M€/turbine)	2.30
M€/MW	1.00
O&M cost (M€/year)	0.19

. Associated with it is a reversible pumping station, with a pumping power of 6 MW, a turbine power of 11.3 MW and a storage capacity in the upper reservoir of 225 MWh. So, this is the basis for the initial calculations of the generation and storage system proposed in this work.

From these initial installations, for those scenarios that require greater generation or storage capacities, the necessary systems will be added. Given that, at the present time, practically all of the island has wind power generation, diversification is considered adequate. In fact, there is a document from the Canarian government itself that analyzes the generation capacities by solar PV energy in each of the islands [53], but given the large areas that are part of the spatial protection areas of the archipelago and particularly the island of El Hierro, the possibility of exploiting solar generation has been considered, but through self-consumption facilities in buildings. Thus, according to the aforementioned document, there would be 1.2 km² of roof surface. The analyses developed within the framework of this study show that the surface suitable for the installation of PV plants on roofs on the island would amount to 0.8 km². Estimates of a maximum installable power of about 83.5 MW, so that this would be the maximum considered as installable on the island in this document.

Table 3. Inputs used for the PV system [54].

Solar panel	Vertex 550+
Lifetime (years)	25
Derating factor (%)	80
Tracking system	No tracking
Temperature coefficient of power (%/°C)	−0.38
Peak Power (W)	550
Nominal operating cell temperature (°C)	45
Efficiency of the panel at standard conditions (%)	21.1
Cost (€/kW)	900
O&M cost (€/kW·year)	13.5

provides a summary of the information necessary to characterize the selected solar panel.

As for additional storage needs, the aforementioned Canary Islands government study [55] shows that any future expansion of storage capacity on the island by pumping would require increasing the size of the reservoirs used in the Gorona del Viento hydro pumping. However, there are numerous protected areas on the island, so prior to any intervention, the possible impact should be carefully analyzed. The estimated cost of this technology is about 3.0 and 0.6 M€ per MW of installed power for the hydro turbination and pumping subsystems, respectively, other additional 7.5 × 10⁻³ M€ per MWh [56]. Therefore, in this document, the use of another alternative storage system has been considered, i.e., the use of electrochemical batteries. The operation would be the same as that of the pumping stations, absorbing excess energy and returning it when needed.

Table 4. Standard system specifications of the selected battery system Tesla Megapack 2XL [57].

Maximum AC power (kW)	979
Energy available (MWh)	3.916
Round-trip system Efficiency (%)	93.7
Cost of the module (€)	460,000
O&M cost (€/year)	5300
Lifetime (years)	25

summarizes the specifications of the mega-battery system used, specifically Tesla Megapack 2XL modules with the 4 h configuration.

In fact, there are currently projects that aim to achieve total decarbonization of the island in the not-too-distant future [58]. In the first phase of implementation, it is intended to install 5 MW of solar PV generation and 5 MW more of storage through battery groups, with the intention of going from approximately 50% of current renewable generation to about 80%. In the second phase, the intention is to increase 7 MW more of solar generation and add a second group of batteries of another 5 MW, in order to achieve 100% decarbonization.

3.6. Definition of HOMER Inputs

In recent years, several calculation models have been defined and used in various research articles to estimate electricity demand and generation in different types of scenarios with varying levels of GHG emissions, achieving high reliability. For example, as particular applications of different code simulation tools, Berna et al. [21] presented a forecast for Gran Canaria for the year 2040 in a scenario of total decarbonization of the economy using HOMER software. Prina et al. [59] performed several forecast simulations for the South Tyrol region using EnergyPLAN software. Segurado et al. [60] simulated different scenarios varying the renewable penetration on the island of S. Vicente in Cape Verde using the H2RES v2.8 tool, and Mirjat et al. [61] conducted a long-term analysis of Pakistan using the LEAP tool. Other research works compile reviews of various energy simulation tools. Hall and Buckley [62] reviewed many energy simulation tools for the United Kingdom. Similarly, Ringkjob et al. [36] reviewed the most used tools for energy and electricity systems employing highly renewable resources, and Prina et al. presented a review of existing simulation tools for energy system scenarios applied at the island level [37].

Based on the aforementioned documents and numerous other works in the extensive existing literature, it is evident that a variety of codes can be used to simulate scenarios for energy planning at different levels, demonstrating over the years their capability to carry out these analyses. The analyses can range from a small isolated system to a country or even a continental level, with various tools used for each level. Among the tools used for energy planning at different levels, notable examples include the use of HOMER, EnergyPLAN, H2RES, MARKAL, and LEAP, as shown in the research works of Hall and Buckley [62] or Prina et al. [59]. As mentioned earlier, all of these are suitable for simulating current scenarios or estimating future scenarios with high levels of Variable Renewable Energy (VRE) sources. Generally, they simulate one year with time steps, usually hourly.

Based on the previous comments, it has been considered that several tools are useful for the objective pursued in the current work. In this context, the widely known tool HOMER has been used [38,63]. The software estimates the best system size, the required investment, Levelized Cost Of Electricity (LCOE) and other economical variables based on different simulated energy sources. The scientific community widely uses this software for various applications, such as predicting energy production and consequently choosing the best option for both stand-alone and grid-connected systems, planning the installation of hybrid energy systems, and estimating their feasibility, among others. Thus, the first solution shown is the optimal one (the global minimum optimum), but other options can also be explored. For instance, additional criteria can be considered, such as the need for less installed power, minimizing energy wastage, the weighted combination of several factors, penalties for land occupation, or visual pollution, etc.

As mentioned above, the HOMER simulation code has been used to conduct detailed analyses of the system’s performance, demonstrating its capability for the required simulations over recent years. To implement the code estimations, a rigorous methodology has been followed, which includes a detailed introduction of the necessary input information and an outline of the steps used, as shown in Figure 8. The required input data includes: annual information on energy demand or, for future estimates, their forecasts; technical and cost information of the generation system to be considered (in the current study, wind and solar PV power plants); technical and cost information of the storage system (mega-batteries and reversible pumping); the energy resources available for each generation system (wind and solar resources available at the selected sites); and additional economic data (such as the annual interest rate and the project’s useful life).

Based on all this data input into the software, the most efficient combination of these generation systems can be estimated to supply all the energy demanded by the system. Specifically, the required nominal power, energy generation from each system, necessary storage capacity, and other factors are determined. HOMER also provides economic information such as the LCOE, initial capital, replacement, O&M costs, etc. The selection criteria in the methodology are the economic ones previously mentioned, while maintaining zero CO₂ gas emissions.

The economic criteria inherently entail a balance between the sizing of generation and storage facilities to meet demand requirements. Given that the attained solution is one in which cost is minimized, this primarily implies that the system size is kept as small as possible while consistently meeting energy needs. In other words, an optimal point is reached between oversizing generation and storage systems. To conduct these estimations, economic data from the year 2024 has been utilized as a reference, assuming that cost variations in the technologies used will remain constant, which should be valid in principle, given the high level of maturity of said technologies. The methodology has been tested in two different scenarios: the first replicates the current existing scenario, while the second resizes the system to be fully renewable, always maintaining the non-negotiable 100% reliability required in such systems isolated from a central grid. It has been deemed fundamental to replicate the results from the year 2024, as the software provides much more detailed information, while simultaneously verifying its ability to reproduce existing real-world results with a high degree of reliability.

The HOMER software conducts simulations of the operation of each analyzed system through an energy balance at each defined time step in the code, with the most common being the hourly balance, as in the current study. The program compares the energy demanded at each time step with the one supplied by the generation system under analysis, as well as the optimal operation of the generators and storage systems. Following the simulation of all system configurations, HOMER provides the system with the lowest NPC.

4. Approaches to Scenario Forecasting on El Hierro for 2040

Several documents published by the Canary Islands government focus on the various aspects required to achieve the goal of decarbonizing the economy by 2040. Each one focuses on one of the key actions required to achieve full economic decarbonization. For instance, one analyzes the self-consumption of solar PV generation [53], other than the energy storage [55], the EV strategies [47], the manageable generation [45], renewable marine generation [64], hydrogen production [65] and DSM-Smart Grids [46]. Finally, a summary of the major points is shown in the Canary Islands Energy Transition Plan (Plan de Transición Energética de Canarias—PTECan) [27].

According to the Energy Storage Strategy for the Canary Islands [55], under a Business-as-Usual (BAU) scenario, assuming population growth to exceed 2.5 million inhabitants by 2040 and a GDP growth rate of 2% annually, electricity demand across the archipelago would increase by approximately 3 GWh per year from the current 9 TWh. If energy efficiency and collective mobility measures are implemented, total demand could instead be reduced to around 8.4 TWh. Conversely, based on the EV Strategy [47]. The full deployment of electric vehicles would increase electricity demand by approximately 6–8 TWh per year.

Translating these projections to the island of El Hierro, the expected annual electricity demand under current trends is projected to reach approximately 65 GWh by 2040. Incorporating energy efficiency, DSM, and sustainable mobility policies could reduce this value to slightly below the current demand, around 52 GWh. However, the full electrification of road transport would notably raise consumption. Considering the existing vehicle fleet, this would represent an additional 43.34 GWh (an 80% increase over current demand). If collective mobility measures reduce the vehicle-to-inhabitant ratio from 0.862 to 0.626 by 2040, the increase would moderate to about 31.50 GWh per year. These conditions thus delimit the range of total demand forecasted for 2040, between approximately 81.5 and 108.35 GWh. Given the large storage capacity required to manage a fully renewable system, additional storage capacity has been necessary (there are limitations on the island for expanding the existing pumped storage infrastructure or locating new ones). To this end, coverage through storage with megabatteries has been used (probably this should also favor distributed storage in self-consumption, although for the simulation, Tesla Megapack 2XL battery modules have been used, basically at a macro level, the only difference would be an increase in cost).

Two primary scenarios are therefore analyzed:

• Scenario 1 (BAU scenario): assumes the continuation of current demand trends combined with total EV penetration.
• Scenario 2 (Efficiency scenario): considers the implementation of strong energy-efficiency measures, sustainable mobility, and DSM policies.

The analysis of the two scenarios employs both deterministic and stochastic approaches. The deterministic analysis (Section 4.1) identifies optimal system configurations under average conditions, while the stochastic analysis (Section 4.2) incorporates uncertainty in renewable resources and demand to assess system reliability and necessary oversizing as described in Section 3.

4.1. The Deterministic Approach

This subsection presents the results obtained under a deterministic framework, assuming average conditions for solar and wind resources and forecasted electricity demand profiles for 2040. The generation–demand balance is evaluated using fixed hourly values, without incorporating stochastic variability. The resulting configuration represents the baseline optimal mix that minimizes costs while achieving full demand coverage under standard conditions.

4.1.1. Business-as-Usual (BAU) Scenario

The following paragraphs summarize the main results of the BAU scenario, which considers and analyzes both economic and technical aspects. It analyzes the different resources considered, specifically wind and solar, together with the conditioning factors mentioned earlier (maximum available capacities of reversible storage and solar PV energy under a self-consumption regime).

Table 5. Summary of the installed generation power, storage power and capacity, and energy production of the BAU scenario.

Generation Systems
Technology	Power (MW)		Energy		Surpluses
			GWh	%	GWh	%
Solar PV	53.0		90.96	58.45	-	-
Wind	16.1		64.66	31.55	-	-
Total	69.1		155.62	100	42.47	27.29
Storage Systems
Technology	Power In/Out (MW) 1		Stored Energy (MWh)	Energy In		Efficiency (%)
				MWh	% 2
Hydro-Pump	32	16	750	27.20	17.48	81
Batteries	9.79	9.79	39.16	3.471	2.23	93.7
Total	41.79	25.79	789.16	30.20	19.71	-

¹ Power In/Out refers to the Pumping/Turbination and Charge/Discharge processes for the reverse pumping and batteries, respectively; ² percentage of energy recovered by the storage systems as a percentage of the total generated energy.

summarizes the main characteristics of the generation and storage systems. Solar PV generation is by far the largest installed power, accounting for more than 75% of the total (53 MW of nominal capacity), while wind energy accounts for the remaining almost 25% (slightly above 16 MW). The Capacity Factors (CF) of solar PV and wind energy sources are quite high, especially wind power, with values well above those typical of existing installations on the Iberian Peninsula, reaching almost 20% and exceeding 45%, respectively. It is worth noting the low percentage of system surpluses, which is achieved thanks to the important contribution of the storage systems, which have a combined power of more than 40 MW. This means that they have the capacity to absorb practically all of the unused energy, as in reality, almost all of the excess comes from systems reaching their storage capacity. Despite having almost 800 MWh, which is about 40 times the island’s peak hourly demand, this capacity is not fully utilized. Thus, almost 20% of the energy generated is fed into the grid when necessary, although at the cost of losing just over 2% in the storage processes. The high capacity of the storage systems leads to a reasonable energy excess of about 27% of the total generated electricity.

Table 6. Capital, replacement, O&M, and total discounted costs for the sub-systems of the scenario.

Sub-Systems	Capital (M€)	Replacement (M€)	O&M (M€) 1	Total (M€) 1
PV	48.7	38.19	15.58	101.5
Wind	16.1	12.88	8.41	37.4
Reverse Pumping	73.95	0	17.95	90.9
Mega-Batteries	4.6	7.35	1.154	13.1
Total	141.9	58.4	43.1	242.9

¹ Total costs over the 50-year life of the project.

presents the key variables related to the financial analysis of the proposed system. Total costs are directly influenced by the installed capacity of each subsystem, establishing a direct link to the data in Table 5, which details electrical generation and power for each system. The total capital expenditure required for the entire electrical installation amounts to just under 150 M€. Within this amount, initial investment accounts for nearly 60%, replacement costs of various subsystems represent 24%, and O&M costs make up approximately the remaining 17%. These combined costs lead to nearly 250 M€ overall, resulting in a final LCOE of 103.0 €/MWh, a value that is currently competitive and likely to remain so through 2040.

Analyzing the costs associated with each subsystem and the different cost components reveals that both solar PV and reverse pumping represent the main contributors to the total system expenditure. However, their cost profiles differ significantly: for solar PV, high replacement costs over the system’s lifetime account for a substantial share of the total expenses, whereas in reverse pumping, the initial capital investment dominates the cost structure. In contrast, the mega-batteries subsystem exhibits a relatively low total cost compared to the other subsystems; however, most of its expenditure is attributable to periodic replacements, indicating a limited operational life for its main components. The wind generation subsystem, on the other hand, distributes its costs more evenly among capital, replacements, and O&M, though its overall contribution to the total system cost is less significant in absolute terms. Therefore, investment strategies should carefully consider both the heavy upfront capital required for reverse pumping facilities and the substantial replacement needs inherent to PV and battery systems throughout their operational lifecycle.

4.1.2. Efficiency Scenario with EV Mobility Policies

The Efficiency Scenario represents the lower-bound demand projection for 2040, around 81.5 GWh, considering strong DSM, sustainable mobility, and energy-efficiency measures that reduce electricity use despite full EV penetration.

Table 7. Summary of the installed generation power, storage power and capacity, and energy production of the Efficiency scenario.

Generation Systems
Technology	Power (MW)		Energy		Surpluses
			GWh	%	GWh	%
Solar PV	42.0		72.08	60.95	-	-
Wind	11.5		46.19	39.05	-	-
Total	53.5		118.27	100	31.26	26.43
Storage Systems
Technology	Power In /Out (MW) 1		Stored Energy (MWh)	Energy In		Efficiency (%)
				MWh	% 2
Hydro-Pump	25	12	500	17.08	14.44	81
Batteries	9.79	9.79	39.16	5.97	5.05	93.7
Total	34.79	21.79	539.16	23.51	19.49	-

¹ Power In/Out refers to the Pumping/Turbination and Charge/Discharge processes for the reverse pumping and batteries, respectively; ² percentage of energy recovered by the storage systems as a percentage of the total generated energy.

presents a summary of the key characteristics of the generation and storage systems. Solar PV generation represents the largest installed capacity, comprising approximately 55% of the total with 42 MW of nominal power, while wind energy accounts for the remaining nearly 45%, around 11.5 MW. As in the prior scenario, the system is capable of absorbing virtually all unused energy, except when the storage units reach full capacity, despite having over 500 MWh of storage available. Additionally, approximately 20% of the demand is supplied back to the grid as needed through the storage systems. The substantial storage capacity effectively reduces energy excesses to around 30% of the total electricity generated.

Table 8. Capital, replacement, O&M, and total discounted costs for the sub-systems of the Efficiency scenario.

Sub-Systems	Capital (M€)	Replacement (M€)	O&M (M€) 1	Total (M€) 1
PV	37.8	30.26	12.35	80.4
Wind	15.5	9.20	6.01	26.7
Reverse Pumping	54.75	0	12.21	67.0
Mega-Batteries	4.6	7.35	1.154	13.1
Total	108.7	46.8	31.7	186.6

¹ Total costs over the 50-year life of the project.

summarizes the main financial variables of the proposed system. The total capital expenditure for the entire electrical installation is just under 110 M€. Of this amount, the initial investment represents nearly 60%, replacement costs for the subsystems account for 25%, and operating and maintenance (O&M) costs comprise approximately 17%. Together, these costs total close to 190 million euros, yielding a LCOE of 102.6 €/MWh. The analysis of the associated costs within each subsystem is identical to the description in Table 6, as both systems incur the same costs for each technology. However, there are variations in the relative costs of each subsystem, although in both cases, solar PV generation systems and irreversible pumping dominate.

4.1.3. Comparative Summary for the Deterministic Approach

Figure 9 summarizes the major characteristics of both scenarios. As shown, the BAU scenario comprises solar PV (53 MW) and wind (16 MW). Large storage systems (over 40 MW power, ~800 MWh capacity) absorb most excess energy, keeping surpluses to about 27% of total generation. Storage includes pumped hydro (32 MW, 750 MWh, 81% efficiency) and batteries (nearly 10 MW, 39 MWh, 94% efficiency). This mix ensures efficient use of renewable energy and system flexibility.

In the Efficiency scenario, the generation mix consists of about 42 MW of solar PV and 11.5 MW. Storage systems have a combined power capacity of approximately 35 MW and 539 MWh of energy storage, including pumped hydro (25 MW, 500 MWh) and batteries (around 10 MW, 39 MWh).

In both scenarios, the solar PV subsystem, which represents approximately 75% of installed capacity and about 60% of electrical generation, is justified by its relatively low cost. The design emphasizes self-consumption, utilizing the large available rooftop surface areas on the island while avoiding large-scale solar farms to minimize land use and visual impact, an important consideration for islands with significant tourism sectors. Although high solar PV integration carries risks such as electrical system oversizing and increased energy surpluses due to generation variability, these challenges necessitate the use of complementary systems to maintain balance.

Wind generation plays a significant role due to the high-capacity factors achievable in the Canary Islands, particularly for offshore installations, which currently stand at around 45%. Despite inherent variability, local conditions yield relatively stable generation. Thus, wind contributes roughly 40% of the generation, with an installed capacity of approximately 25% of the system’s total. Specific local constraints and existing infrastructure, currently including five wind turbines on the island, have informed site selection and system design.

The substantial power and energy capacity of the storage subsystems, which can recharge power on the order of 60% of installed generating capacity, and storage capacity approximately 40 times the island’s peak power demand, notably enhance system flexibility. These storage units effectively absorb excess energy during periods of overgeneration, accounting for approximately 20% of the total generated energy, and return it to the grid as needed.

The energy surplus within the proposed system remains relatively low, around 30%, which is efficient given the scale of the scenario. This performance is largely attributable to the substantial storage capacity provided by the combined reverse pumping and megabattery subsystems, which effectively absorb excess generation and thereby minimize overall energy wastage.

As both scenarios were modeled with the same renewable sources, to gain a comprehensive understanding of the behavior of the generation-storage system, a complete annual representation of each generation and storage unit is necessary. As shown in Figure 10, solar PV generation remains fairly stable throughout the year, with the lowest levels observed in the winter months, and values remaining relatively consistent from November to February. On the other hand, wind power generation varies considerably, with peaks in July and August and lows in October, closely followed by November, while the rest of the months have fairly similar generation levels. Although November combines low wind power generation with relatively low solar power generation, making these months the most critical. The solar resource clearly contributes more during the warmer months. As is known, the lowest production occurs in the winter months, while the rest of the time the generation is similar. Although summer has more sunlight hours, generation is reduced due to system performance penalties caused by relatively high temperatures (approximately from day 120 to 240, corresponding to the months of February to October). Therefore, solar remains one of the main renewable resources in the Canary Islands due to its favorable location, and in the case of El Hierro, it is currently significantly underutilized. Regarding the wind resource, the highest contributions occur in the summer months (days 170–240, approximately corresponding to July and August), while the minimums occur between days 260 and 330 of the year (mid-September to late November). However, the rest of the year shows relatively high values, with some exceptions, especially in December, January, and April.

Figure 11 illustrates the monthly performance of the two storage subsystems for the two scenarios, based on hourly data represented through box-and-whisker plots. The reverse pumped hydro has storage capacities of 750 and 500 MWh for the BAU and Efficiency scenarios, respectively. Both scenarios include battery systems of approximately 40 MWh; therefore, the reversible pumped hydro unit largely defines the overall system performance. In general, during summer months, particularly in July and August, both scenarios and storage systems are scarcely utilized, maintaining median States of Charge (SoC) close to full capacity. In contrast, October and November exhibit the lowest charge levels, with median SoC values around 50% for pumped hydro and close to 0% at minimum.

For pumped storage (Figure 11a), October and November remain the most critical months, with SoC values dropping close to zero in both scenarios. December and January show milder deviations in the Efficiency case than in BAU. In the BAU scenario, January’s median SoC is about 70% with minima near 10%, whereas in the Efficiency scenario, the median exceeds 95% and never falls below 90%. December follows a similar but less pronounced pattern: the BAU median slightly above 90% declines to about 50% at minimum, while in the current case, the median remains comparable and the minimum improves to above 80%. During the remaining months, both median and minimum SoC values in the two scenarios are very similar, typically above 90%, with the only exception being September, where minima approach 80%.

Battery behavior (Figure 11b) follows similar overall trends, with slight variations. In July and August, battery utilization remains low in both scenarios, although it is lower in BAU scenario, where medians approach 100% and minima lie near 90% and 80%, respectively. In the Efficiency scenario, median SoC values decrease slightly to above 90%, and the minimum range is between 50% and 60%. For the rest of the year, battery usage is generally higher in the Efficiency scenario than in BAU, with minima dropping to 0–20% in both cases, while median SoC values remain higher in BAU. Both systems show increasing storage activity toward winter, following an inverted U-shaped pattern through the year—from about 70% to near 100% and back to 80% in BAU, and from 50% to almost 100% and back to 60% in the current scenario. When accounting for their respective storage capacities, the two systems display closely aligned operational behaviors.

As shown in Figure 12a, cost analysis identifies solar PV energy and pumped hydro storage as the primary contributors to total system expenditure, although with distinct cost profiles: PV systems incur substantial replacement costs over their lifetime, whereas pumped hydro is characterized by high upfront capital investment. Megabatteries exhibit relatively low total costs but require frequent replacements due to the limited lifespan of their components. Wind generation costs are more evenly distributed among capital, Operation and Maintenance (O&M), and replacements, but their overall contribution to total cost is smaller. Hence, investment strategies should balance the high initial capital of pumped storage against ongoing replacement needs for solar PV and battery systems during their operational life. The LCOE for both resulting energy mixes is relatively low, slightly above 100 €/MWh, making it cost-competitive (Figure 12b). The surpluses in both cases are also below 30%, while about 30% of the demanded energy passes through the storage systems.

4.2. Stochastic Approach

4.2.1. Forecasting of Weather and Demand Curves

This first section develops the analysis of the trend-continuation scenario of demand combined with the total penetration of EVs, also following the current trend of terrestrial vehicle usage. To carry out the best estimate approach, considering uncertainties (BEPU analysis), 93 cases are randomly sampled based on the probability distributions of the variables presenting uncertainty, namely, the two variable generation sources and the demand applied in this scenario. In other words, a typical scenario is simulated first, which may have averaged values or even typical performance values of these variables for a particular case. Then, simulations of the 93 scenarios are carried out, where the uncertain variables have been sampled—in the current case, 3 variables and 8784 values (hours that constitute the year 2040). After these simulations, as will be developed throughout this section, both the generation and storage capacities should be increased to ensure no lack of coverage of demand under any of the sampled conditions. In fact, this has been carried out with 95% confidence and coverage levels.

To provide an initial idea of the effect of studying the uncertainties associated with the different input variables, Figure 13 shows the evolution of the hourly demand curves during a typical winter and summer week. Therefore, integrating statistical methodologies and input variable–based analyses provides a reliable estimate, offering insights into the possible variability and the factors that can shape the energy landscape in the coming years. In particular, Figure 13 illustrates the projected electric demands for the 93 cases and the base demand case (this study presents an average evolution for the 11 years analyzed) of 1 January and 8 August, both as examples. The forecasted demands are shown with lines of different colors, forming a demand behavior band, but this region should not be taken as an uncertainty band, since each of the 93 scenarios has its characteristic curve, so generally none of the scenarios will remain throughout the year in the mid or low range. However, it does give an idea of the variability present. Finally, the red line represents the base case demand.

Regarding Figure 14, the average hourly solar irradiance curves can be seen, also for a typical winter and summer day. Likewise, these curves were obtained using the BEPU methodology. Similarly, Figure 15 provides an hourly average of the wind speed curves obtained through the BEPU methodology. These graphs reflect the fluctuating nature of both the irradiance and wind speeds, which is important to understanding the intermittency and variability of both generation sources.

Therefore, in a first approximation, there is a base case, for which the average conditions of the system presented in the three previous figures are used (deterministic modeling). In a second phase, the intrinsic variability of both the two renewable generation sources and demand is taken into account. To do this, it is necessary to model the performance of the system under the 93 cases described (each of the 93 samples in triplets of data for the 8784 h of the year 2040). Since the system that is expected to be capable of meeting demand will not actually be so if these sources of variability in system performance are taken into account.

4.2.2. Business-as-Usual (BAU) Scenario

Upon analyzing the 93 simulated scenarios, several key observations emerge. Of these, only 40 scenarios achieved complete demand coverage at all times (Figure 16). While it is true that most of the 53 scenarios exhibiting unmet demand did so at low levels, it must be acknowledged that any imbalance in a small-scale system can potentially result in a blackout, necessitating careful consideration. The most adverse case exhibits an unmet demand of 1.3%, corresponding to approximately 1.4 GWh per year. At the opposite extreme, the case with total demand coverage also presents the highest energy surplus, accounting for approximately 33% of the total generation.

Table 9. Performance of the optimized deterministic system over a stochastic approach for the BAU scenario.

	Base Case	Unfavourable Case	Favourable Case
Renewable sources
Solar PV (GWh)	90.96	90.65	91.86
Wind (GWh)	64.66	68.37	71.62
Total (GWh)	155.62	159.02	163.48
Storage systems
Hydro pump (GWh)	24.33	25.02	24.93
Battery (GWh)	3.471	3.403	4.696
Total (GWh)	27.80	28.42	29.63
System Excesses/Unmet Demand
Generation Surpluses (GWh)	42.47	47.31	50.61
(%)	27.29	29.70	30.94
Unmet Demand (GWh)	0	1.407	0
(%)	0	1.29	0

summarizes the performance of the three most representative cases: the reference case and the two extreme outcomes derived from uncertainty in solar, wind, and demand inputs. It is worth noting that, across all cases, surplus energy remains contained. This indicates that the system is not significantly oversized, primarily due to the considerable capacity of the two storage subsystems.

However, although most of the 53 cases with unmet demand have a very low percentage, this is not acceptable from the point of view of reliability. Therefore, the system must be resized to eliminate this risk. There are essentially two approaches: (i) upsizing the proposed system, or (ii) implementing an additional dispatchable generation subsystem. If the second solution is preferred, using exclusively renewable sources, the only feasible option given the island’s constraints would be biomass. Alternatively, albeit with higher GHG emissions, the retention of diesel generator sets could be considered. In this latter case, given the low unmet demand percentages, diesel generation would be required rarely, with a capacity factor of only around 1.3% at the 95% reliability and coverage levels.

Exploring different resizing options offers a general view of the repercussions of uncertainties in the dimensioning of energy systems (

Table 10. Performance of possible resized systems in the stochastic approach for the two extreme cases in the BAU scenario.

	Resized Solar PV		Resized Wind		Resized Batteries		Resized Solar PV/Batteries
	Unfavour. Case	Favou. Case	Unfavour. Case	Unfavour. Case	Unfavour. Case	Favour. Case	Unfavour. Case	Favour. Case
Demand (GWh/year)	108.43	108.07	108.43	108.07	108.43	108.07	108.43	108.07
Resizing (MW)	14 MW PV Pannels		6 Wind Turbines (13.8 MW)		21 Batt (20.56 MW, 82.24 MWh)		7 MW PV/6 Batt (5.874 MW, 23.50 MWh)
Generation (GWh)	183.35	187.59	217.92	225.82	159.02	163.48	171.3	175.59
Generation Surpluses (GWh)	63.39	68.94	101.87	111.23	45.37	52.71	55.59	61.09
(%)	34.57	36.75	43.50	49.26	28.53	32.20	32.45	34.79
System Costs (M€)	269.50	269.05	271.84	271.40	263.51	263.22	260.15	259.84
LCOE (c€/kWh)	11.42	11.43	11.51	11.53	11.17	11.19	11.02	11.04

). As shown, increasing the battery storage with 21 additional modules (approximately 20 MW of power and nearly 80 MWh of extra storage capacity) would satisfy the reliability requirements. In addition, a substantial increase in solar PV generation, specifically, the installation of 14 MW, would be necessary to ensure demand coverage. Expanding wind generation by adding six extra wind turbines (about 14 MW) was also assessed.

Analyzing the performance of the three options under the most adverse scenario, which motivated the maximum oversizing aimed at ensuring the desired system stability, several key observations can be made. The first option involves the greatest degree of oversizing; however, since it relies solely on storage rather than additional generation sources, it maintains the surplus energy levels of the initial system. The additional installation of solar panels (14 MW) leads to a modest increase of just over 6% in system surpluses. In contrast, increasing the number of wind turbines results in a significant rise in surpluses, approximately 15%, primarily because the most adverse performance period for the system occurred during October and November, when wind availability was at its minimum. Finally, the combined configuration of solar and batteries, consisting of a 7 MW peak solar installation and 6 battery modules, produces a moderate increase in surpluses.

Focusing on costs, the lowest LCOE is achieved by the latter option (solar plus batteries), although the re-dimensioned battery system presents a comparable value. In any case, both options incur cost increases exceeding 10% relative to the baseline system identified through deterministic analysis, from the initial 10 c€/kWh generated to more than 11 c€/kWh. The addition of extra wind turbines is the least favorable option from both a cost and surplus energy perspective. While the rise in solar capacity results in costs similar to those of increased wind generation, its surplus levels are closer to those observed for the two options previously identified as most suitable.

It is important to note that the system will not always operate under the most adverse conditions. This implies that, during a favorable year, the system oversizing implemented to ensure reliability under challenging conditions will lead to even greater inefficiency. Consistent with observations in the adverse scenario, wind generation exhibits the poorest performance under favorable conditions, with losses increasing to nearly 50% of the energy generated. The other systems exhibit a contained increase in surpluses, with very similar behavior between the battery-only re-dimensioned system and the hybrid configuration that combines batteries and solar PV generation.

Another possibility arises from a detailed analysis of the system’s critical operation intervals revealed that energy deficits occur predominantly during the months of October and November. These shortages are particularly significant, with the system frequently failing to meet the majority of demand during these periods. Specifically, power deficits often exceed 10 MW and can reach instantaneous peaks close to 15 MW. Therefore, if a backup generation system is to be deployed, its capacity should be sized according to these maximum observed shortfalls.

4.2.3. Efficiency Scenario with EV Mobility Policies

As presented for the BAU scenario, a deterministic analysis results in a system capable of achieving the objectives set in terms of demand coverage and excess energy control, considering the assumptions made, i.e., specific hourly generation profiles for both solar and wind resources, as well as unique hourly demand values. Consequently, a stochastic approach is carried out. This methodology permits the capture of a wide range of possible operational conditions and assesses the robustness of the system’s design.

The primary metric used to evaluate system performance in each scenario is the percentage of unmet demand, that is, the proportion of total load that could not be supplied. A secondary variable of interest is the fraction of surplus energy produced. The results of all 93 scenarios are synthesized in Figure 17, which shows, for each case, both the percentage of unmet demand and the proportion of surplus electricity.

Analysis of these 93 simulated scenarios reveals several important insights. Notably, only 33 scenarios resulted in complete satisfaction of demand throughout the year. Although unmet demand in the remaining 60 cases was generally low, it is important to recognize that even small mismatches can be dangerous in a microgrid or small-scale system, as any shortfall may lead to potential zero energy supplies. The most severe case observed showed an unmet demand of nearly 1.3%, equating to approximately 1.07 GWh annually. In contrast, the scenario that achieved full demand coverage also exhibited the highest level of surplus generation, with excess electricity reaching more than 12% of total annual production.

Table 11. Performance of the optimized deterministic system over a stochastic approach for the Efficiency scenario.

	Base Case	Unfavourable Case	Favourable Case
Renewable sources
Solar PV (GWh)	72.08	71.62	72.86
Wind (GWh)	46.19	49.29	51.09
Total (GWh)	118.27	120.91	123.95
Storage systems
Hydro pump (GWh)	17.08	16.40	13.29
Battery (GWh)	5.97	7.27	6.53
Total (GWh)	23.15	23.67	19.82
System Excesses/Unmet Demand
Generation Surpluses (GWh)	8.35	13.11	15.29
(%)	7.06	10.84	12.33
Unmet Demand (GWh)	0	1.067	0
(%)	0	1.29	0

provides a summary of the performance observed in the three most representative scenarios: the reference case, and the two extreme outcomes reflecting the greatest impact of uncertainties in solar, wind, and demand data. Across all scenarios, it is notable that the volume of surplus energy remains limited, which suggests that the system is well-sized—neither significantly oversized nor undersized—thanks largely to the substantial storage capacity provided by the two storage subsystems.

The same considerations that apply to the BAU scenario also hold in the Efficiency scenario (

Table 12. Performance of possible resized systems in the stochastic approach for the two extreme cases in the Efficiency scenario.

	Resized Solar PV		Resized Wind		Resized Batteries		Resized Solar PV/Batteries
	Unfavour. Case	Favour. Case	Unfavour. Case	Unfavour. Case	Unfavour. Case	Favour. Case	Unfavour. Case	Favour. Case
Resizing (MW)	32 MW PV Panels		5 Wind Turbines (11.5 MW)		17 Batt (16.64 MW, 66.57 MWh)		6 MW PV/4 Batt (3.916 MW, 15.66 MWh)
Generation (GWh)	175.83	178.87	169.75	175.82	120.91	123.95	131.21	134.25
Generation Surpluses (GWh)	86.33	90.26	81.65	88.70	31.04	38.66	42.58	46.44
(%)	46.2	47.80	45.70	48.40	28.50	30.90	30.4	32.7
System Costs (M€)	248.91	248.72	210.90	210.65	204.74	204.56	201.51	201.32
LCOE (c€/kWh)	13.38	13.71	11.59	11.61	11.25	11.27	11.08	11.09

), with the only difference being that the solar generation system must be significantly expanded to meet the demand of the more restrictive case (power increase of 32 MW). This situation is driven by the fact that this Efficiency scenario features a reduced storage capacity, which is unable to redistribute solar surpluses during critical periods, resulting in significantly more adverse system performance. Therefore, this option should be discarded in this case, leaving the two options identified as most suitable in the BAU scenario—the increased battery system and the hybrid configuration, which combines a modest increase in solar PV with battery storage—as the preferred choices.

In this Efficiency scenario, similar observations can be drawn as those made for the BAU scenario. Specifically, eliminating the lack of coverage by resizing the system—through significant increases in power capacity and/or storage—requires very substantial upgrades. As in the BAU scenario, these big demands cluster in the system’s most challenging performance intervals, during which almost no energy demand can be met. Therefore, the most appropriate solution is likely the implementation of a backup system, which under the most unfavorable conditions would generate around 1% of the annual energy demand. Its capacity would need to be close to the peak demand power, estimated in this scenario to be approximately 10–12 MW.

4.2.4. Comparative Summary for Stochastic Approach

While the deterministic simulations of both the BAU and Efficiency scenarios confirm technical feasibility under nominal conditions, the stochastic extension provides a more realistic appraisal of system resilience against interannual variability in wind, solar, and demand (Figure 18). Incorporating 93 random realizations derived via Wilks’ formula (95/95 confidence-coverage) exposes the sensitivity of the El Hierro energy system to uncertain input.

Uncertainty analysis considering the inevitable variability of solar and wind re-sources, as well as demand, revealed a significant impact on system self-sufficiency and system reliability. A total of more than half of the simulations do not achieve complete coverage of electricity demand throughout the analyzed period, although this typically occurs at a low level. Consequently, from a reliability perspective, any non-zero unmet demand is considered unacceptable in small-scale or isolated systems due to the potential risk of blackouts. The results then demonstrate the need to oversize generation and storage systems to cope with these fluctuations. Simulations of multiple stochastic scenarios suggest that while a deterministic system may meet coverage and surplus control targets, variability and uncertainty in renewable generation and demand often lead to incomplete demand coverage, typically with small deficits. An increase of at least 10% in installed generation capacity is required, which translates into an additional cost increase of approximately 10% in both scenarios.

A comparison of critical operating periods revealed that energy shortages predominantly occur in October and November. Hence, any complementary backup generation system should be dimensioned to cover at least this shortfall. Overall, the results demonstrate that a moderate degree of oversizing, preferably through the expansion of storage capacity or combined solar PV and storage solutions, provides an optimal balance between system reliability, cost efficiency, and renewable energy utilization.

In the most adverse case of the BAU scenario, oversizing the system by adding storage alone maintains surplus energy levels comparable to those of the original system. Adding 14 MW of solar panels modestly increases surpluses by just over 6%, while expanding wind turbine capacity causes a larger surplus increase of approximately 15%, primarily due to low wind availability during critical system performance months (October and November). The combined option of 7 MW solar PV plus battery modules results in a moderate increase in surplus and achieves the lowest LCOE (Figure 19a); however, all oversizing options raise costs by over 10% compared to the baseline system. Among the three options, adding wind turbines is the least favorable option, considering both cost and surplus energy.

Under more favorable conditions, system oversizing leads to increased inefficiencies, particularly for wind generation, which can lose nearly 50% of its generated energy. Other configurations show smaller increases in surplus energy, with very similar behavior between the battery-only and the solar-plus-battery re-dimensioned systems.

A backup generation system sized near the peak demand (about 15 MW) would be another option to manage system oversizing and ensure reliability under challenging conditions. This backup would supply about 1.4% of the annual energy demand. On El Hierro Island, biomass is essentially the only available renewable option for such backup generation, providing a dispatchable, sustainable complement to wind, solar PV, and pumped hydro storage. Integrating biomass-based backup enhances system stability and supports the island’s ambitious goals for high renewable energy penetration.

In the Efficiency scenario, the trends are similar; however, oversizing the system by simply adding battery storage keeps surplus energy levels comparable to those of the original system. Adding 32 MW of solar panels significantly increases surpluses by almost 20%. This, in turn, increases the system’s LCOE by approximately 30% (Figure 19b). Meanwhile, expanding the capacity of the wind turbines causes an increase in surpluses that is practically the same, mainly due to the low wind availability during the critical months for system performance (October and November). The combined option of 6 MW of solar photovoltaic energy plus 4 battery modules results in a moderate increase in surplus and achieves the lowest LCOE; however, all oversizing options increase costs by around 10% compared to the reference system, with solar oversizing increasing costs by approximately 30%.

The comparative assessment of both the BAU and Efficiency scenarios confirms that the long-term stability and reliability of isolated renewable energy systems depend fundamentally on achieving an appropriate balance between generation and storage capacity. Although increasing solar PV and/or wind power capacity can partially alleviate supply deficits under specific operating conditions, such measures inherently generate substantial energy surpluses, especially during favorable resource periods, and persistent shortages during unfavorable ones. The results demonstrate that merely expanding generation does not guarantee a robust system; rather, it increases inefficiency and curtailment without proportionally improving reliability.

In contrast, the coordinated operation of the two storage subsystems—pumped hydro and batteries—proves essential to maintaining a stable energy balance. Reverse pumping storage ensures the buffering of medium-term mismatches, while batteries provide short-term flexibility to manage rapid fluctuations in supply and demand. Their combined action enables the redistribution of energy surpluses from periods of high renewable production to times of scarcity, thereby minimizing both unmet demand and energy waste.

Therefore, the analysis highlights the unavoidable necessity of a well-balanced system architecture where storage plays a central and strategic role. Only through the integration of both reversible pumping and electrochemical storage can the system achieve the dual objectives of reliability and efficiency while progressing toward a fully renewable energy configuration with reduced dependence on backup generation.

5. Conclusions

The study demonstrates that full decarbonization of El Hierro’s energy system by 2040 is technically feasible and economically viable through 100% renewable generation supported by storage. The two extreme scenarios predicted for 2040 were analyzed: a Business-as-Usual (BAU) case with an annual demand of 108.35 GWh and an Efficiency case with an annual demand of 83.50 GWh/year demand achieved through Demand-Side Management (DSM). Both scenarios rely on a mix of solar PV and wind generation, providing approximately 60% and 40% of total output, respectively, complemented by pumped hydro storage (500–750 MWh) and batteries (39 MWh).

The resulting Levelized Cost Of Energy (LCOE) is between 103 and 111 €/MWh, confirming competitiveness against conventional generation. Efficiency measures reduce total demand by 22.9% and capital costs by 23.4% without significantly affecting LCOE. Energy surpluses remain moderate (≈approximately 27%) due to the effective operation of large-scale storage, which maintains the state-of-charge above 90% for most of the year, with deficits only occurring in October and November.

Stochastic analysis revealed that deterministic designs underestimate system needs by 7–10%. Only 43% of BAU and 35% of Efficiency simulations achieved full coverage, with a maximum unmet demand of nearly 1.3%. To ensure 100% reliability, moderate system resizing is required, combining solar and battery expansion (+7 MW PV and +6 modules for BAU; +6 MW PV and +4 modules for Efficiency), which increases costs by 7–8% but maintains balanced surpluses. Alternatively, limited biomass backup generation of 10–15 MW can secure reliability with minimal operation (<1.5% of annual energy).

The investment structure highlights solar PV and pumped hydro as cost drivers, with wind providing complementary stability and batteries ensuring fast response. Demand-side management and electrified mobility policies remain the most cost-effective pathways for decarbonization, as they reduce demand while maintaining system flexibility.

Future research should focus on improving the flexibility and reliability of fully renewable island systems. Key priorities include developing advanced demand management strategies that align consumption with renewable generation, integrating Vehicle-to-X technologies (such as V2G and V2H) to use electric vehicles as flexible energy assets, and implementing smart predictive control systems to enhance grid stability. Studies on hybrid storage solutions combining hydro, batteries, and hydrogen, along with coordinated demand-side management and distributed renewables, could further optimize system efficiency. Ultimately, future research should investigate effective policy frameworks and consumer engagement mechanisms that foster energy efficiency, self-consumption, and long-term sustainability.