**1. Introduction**

The current state of the energy generation landscape is undergoing a significant change as concerns are raised over climate change, energy cost, and energy security. The aim as stipulated in the Paris Agreement [1] of keeping the average surface temperature increase below 2 ◦C by 2050 is unlikely given global trends [2], and will be impossible without an ambitious sustainable energy development and technological innovation [3]. Recent events on the global stage have also caused nations in Europe and around the world to reconsider their energy security strategies [4,5]. The adoption of renewable energy at scale should include measures to increase the effectiveness whilst providing cost reductions [6].

The introduction of renewable energy systems (RES), including photovoltaic (PV) solar panels and wind turbines, have been the key driving force in removing global dependence on fossil fuels from the energy sector [7]. These types of generation assets are known as non-dispatchable as they are completely dependent on weather conditions [8] and so cannot be precisely controlled. This is a problem for transmission service operators (TSOs) as ensuring the voltage and frequency are balanced at a grid level becomes challenging [9]. The increasing volume of decentralised RES installed at the demand side is also problem for grid operators [10], as they can induce bidirectional grid flows and put additional strain on the network.

**Citation:** Garner, R.; Dehouche, Z. Optimal Design and Analysis of a Hybrid Hydrogen Energy Storage System for an Island-Based Renewable Energy Community. *Energies* **2023**, *16*, 7363. https:// doi.org/10.3390/en16217363

Academic Editors: Luis Hernández-Callejo, Jesús Armando Aguilar Jiménez and Carlos Meza Benavides

Received: 2 August 2023 Revised: 26 October 2023 Accepted: 27 October 2023 Published: 31 October 2023

**Copyright:** © 2023 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

One solution to this problem is the use of energy storage systems (ESS) to store excess energy and increase the share of the total RES production directly through selfconsumption [11,12]. Electro-chemical storage such as batteries have been deployed in many cases for use as grid-level storage [13–15], as they have the advantage of fast response to demand and can be installed in most global climates. A number of battery technologies including high-performance solid state chemistries are a promising solution due to their long-term stability and high capacity retention [16,17]. Most grid storage applications deploy LiFePO4 variants as they are widely available and have a relatively low cost [18]. Crucially for the research methods used in this work, the retrieval of reliable cost and environmental data is vital for an accurate result, which for LiFePO4 is widely available within the literature. Hydrogen has often been considered for long-term seasonal storage [19], due in part to the mentioned capacity retention challenges of battery storage. Hydrogen is also a flexible energy vector for many other uses, such as heating and industrial processes [20]. A hybrid battery and hydrogen ESS has a great potential to increase the share of renewables within the energy mix [21], thus decreasing the reliance on traditional power stations. The advent of widely available ESS has meant that it is now possible to emphasise the self-consumption of energy at a local level to reduce the problems of grid stress and planning. A 'prosumer' (an end user that is able to both consume and produce energy [22]) or group of prosumers could install decentralised RES coupled with storage technology, and self-consume the power generated at a local level.

To address these challenges, this work presents and evaluates the application of decentralised renewable energy communities (RECs). RECs in practice have many advantages and solve the most common issues associated with increased decentralised generation, while also promoting the further self-consumption of electricity. In a REC configuration, consumers and prosumers are no longer restricted to buying and selling energy from their utility company and can virtually 'share' the excess energy between actors within the energy community itself. This is mutually beneficial for both the network operator, as they no longer need to manage unpredictable grid flows, and for the REC participants as they receive direct renumeration and a reduction in carbon emissions.

The REC considered in this work was based on the policy recommendations recently implemented by a number of EU countries outlined in the Renewable Energy Directive (RED-II) (EU) 2018/2001 [23]. The directive defines a REC as "a legal entity that is based on open and voluntary participation, it is autonomous and controlled by shareholders or members located in the proximity of renewable energy plants belonging to the community itself. The members may be physical persons, companies or local authorities. . . ". While the directive has been transposed into several other national laws and decrees, including Austria [24], France [25], Germany [26], and the Netherlands [27], the REC modelled in this work most closely resembles the framework practiced in Italy as discussed by Trevisan et al. [28]. Although the study was in Spanish territory, it was chosen to follow the Italian implementation as there are more example cases available and, as of 2021, further improvements to the 2019 Spanish REC policy are currently in progress [29].

As laid out in decree-law 199/2021 [30], a group of self-consuming members within the REC must be located within the same low-voltage (LV) network downstream of the same LV/MV substation. Energy is shared in the existing physical network using a virtual network model. The difference between the energy consumed and energy produced by the REC is resolved over each one-hour period to determine the capacity available to be shared [31]. The model created in this work uses the principles of the relevant regulation to design the virtual REC.

A number of studies including operational renewable energy communities have investigated the use of ESS within a REC to further improve the economic performance of centralised renewables. Trevisan et al. presented an optimised energy model considering PV solar and ESS to provide renewable power to a port REC, showing a decrease in energy bills of 28% compared with the business-as-usual case [32]. Bartolini et al. investigated how to size a mixed RES to fully self-consume all generation at a community level, as

well as meeting the heat energy needs, and showed that using hydrogen generation and storage is an economically viable alternative to battery systems [33]. Although less explored in the literature, there have also been studies focused on the environmental and emissions reductions possible with such a community-based system. Wang et al. proposed a community-based virtual power plant solution in Japan with PV and battery ESS with the ability to reduce carbon emissions by 16.26% [34].

Several different modelling and optimisation software tools have also emerged to assist in model-based design and assessment. An in-depth review by Cuesta et al. presented popular renewable energy modelling tools, including the ability to model different renewable assets and output different technical, economic, environmental, and social key performance indicators [35]. Software such as HOMER (version 3.16), TRNSYS (version 18), and MATLAB/Simulink (2022b) are most often used due to their ease of use and available documentation. However, they can be restrictive for some REC cases due to their proprietary nature. Creating the model in Python will provide the flexibility of an open-source platform and a scalable product suitable for deployment as a lightweight software or web applications.

A number of optimisation procedures have been addressed and utilised in the literature to determine the optimal design of hybrid RES and ESS. Most cases vary the design capacities to achieve one or more competing criteria such as economics, grid independence, and environmental impact. Niveditha and Rajan Singaravel consider a multi-objective design criteria for achieving near zero energy buildings (NZEB), using the functions of cost, loss of load probability (LLP), and total energy transfer (TET) to determine the best sizing arrangement for the PV-wind-battery storage system [36]. Zhang et al. presents a capacity configuration for both an on-grid and off-grid mixed renewable system with hydrogen and batteries [37]. The NSGA-II algorithm was used to determine the trade-off relationship between system cost, renewable curtailment, and loss of load probability (LLP), which can be considered analogous to grid independence for grid-connected configurations. Xu et al. considers the design of an off-grid PV-wind-hydrogen storage system using the multiobjective criteria of LCOE, LLP, and power abandonment rate (PAR). The pareto optimal solution produces an LCOE of 0.226 USD/kWh at acceptable LLP and PAR values [38]. Studying the emissions associated with the grid independence would more accurately determine the positive environmental impact, which was of particular focus in this work. Results from the literature also do not consider the implementation of such an optimization procedure for RECs, and the impact of trading arrangements between members. Other algorithms including multi-objective particle swarm optimisation (MOPSO) [39] and multiobjective evolutionary algorithm with decision-making (MOEA-DM) [40] have also been applied to ESS design; however, NSGA-II remains very popular and has proven robustness in energy flow optimisation problems [41].

#### **2. Contribution**

In this study, a techno-economic and carbon emissions assessment was conducted for a decentralised REC. The case study location was chosen as Formentera; a largely rural Balearic Island located in the Mediterranean Sea as illustrated in Figure 1. Emphasis is put on the isolated nature of the energy grid, which naturally increases the energy cost and embedded carbon of electricity usage, making it an ideal location for the study. A comparison of the base case scenario was used to compare the improvements made with the implementation of the REC.

The community has shared usage of PV solar and wind power to produce energy, and a hybrid battery and regenerative hydrogen fuel cell to store excess production. The combination of battery and hydrogen minimises the potential shortcomings of decentralised storage. A virtual trading scheme based on the EU decree-law 199/2021 for REC implementation was used to evaluate the energy shared between community members, without considering incentives or feed in tariffs. Through the implementation of key economic and

environmental parameters, the multi-objective optimisation determines the best design topology within the defined REC boundary conditions.

**Figure 1.** Formentera Island, highlighted in red, is located east of the Spanish mainland in the Mediterranean [42].

The multi-objective results reveal an inherent trade-off relationship between low-cost energy and the ability to decarbonise supply, and that this approaches a critical limit at both extremes of the pareto front. This work shows that across the pareto optimal sets, the hybridisation of energy storage provides a better overall performance than a battery-only or hydrogen-only case. Additional constraints can be applied to the objective domain to assist in design decision-making.

The implementation of the model in Python allows for the creation of a scalable product, which following digitisation trends in model-based design could provide a vital tool for communities and policymakers to determine the best method for assisting communities to reach net-zero emissions. To summarise, the novelty of this work is summarised as follows:


#### **3. Materials and Methods**

For the purposes of simplification, the simulation model was discretised into one hour time steps using kWh as the function unit for all energy flows within the system. The case study input assumptions including building load and meteorological datasets are defined first. The meteorological data at the chosen coordinate location were obtained from the National Aeronautics and Space Administration (NASA) Langley Research Center (LaRC) Prediction of Worldwide Energy Resource Project funded through the NASA Earth Science/Applied Science Program [43]. A combination of hourly and monthly energy consumption collected from the case study location was used to recreate typical annual load profiles for each of the seven buildings within the virtual REC. A selection of 24 industrial, commercial, and residential load profiles produced by Farhad et al. (2020) were used to augment the profiles where required [44].

#### *3.1. Renewable Energy Community Implementation*

It was assumed that the community members will have a shared capital investment in the generation and storage assets. PV solar and wind assets can either be installed in the low-voltage energy grid within the same secondary substation of the REC, or spread out between the members, installing in open areas such as rooftops. The stationary ESS consisting of both a lithium-ion battery and a RHFC was installed with the REC boundary conditions in accordance with the EU decree-law 199/2021, with the capacity to accept but also release energy to the physical energy grid. A simple diagram of the system architecture is shown in Figure 2.

**Figure 2.** Renewable energy community system architecture. The community buildings grouped on the left are connected virtually to the distributed generation and storage assets, which are also able to export to the local power grid.

The control strategy consists of a load-following authority, but with additional considerations for the hybrid ESS. Since batteries have an improved performance as short-term storage, these are allowed to discharge first to cover the load of the REC. Once the battery depth-of-discharge (DOD) limit is reached, the hydrogen system is then activated to cover the remaining demand. During the charging phase, this control scheme occurs in reverse. By evaluating the excess energy available between the total REC consumption and production over each one-hour increment in line with decree-law 162/19 for community implementation in Italy, the total virtual energy flows between community members were derived. This case does not consider incentives to reduce financial strain and instead evaluates through a techno-economic assessment over a 20-year project period whether the hybrid system was able to provide net-positive economic and environmental performance over the business-as-usual case.

The electrical load profiles form the foundation of the assessment of economic and environmental improvements to the REC. The community consists of seven member buildings; a community centre, a small school, a large school, local government offices, and three typical residential units. For the community centre, two schools, and offices, sample daily load profiles, as well as the monthly average energy consumption, were collected directly from the test site. For the residential units, a combination of the annual heating, cooling, and appliances usage of 80.7 kWh/m<sup>2</sup> was used to evaluation the typical characteristics of a residence in Spain [45], where the buildings were assumed to be 50 m2 in area. The monthly and yearly consumption was used to create a spline, over which the daily load profile was interpolated and repeated to create the one-year load profiles for each building. The total yearly consumption for each member is included in Table 1, with the monthly

and daily load profiles shown in Figure 3. The three mixed homes have been combined to represent a mixed family building and to improve visibility within the analysis.

**Figure 3.** Building electrical energy service demand model based on the requirements of the renewable energy community. Figure 3b displays the average daily power demand curves of the different building types within the model.


