**1. Introduction**

A transition from a fossil-fuel-based energy system to a decarbonized one is key to performing a cost-e ffective strategy to mitigate climate change [1] and achieve the 2 ◦C threshold aim of the Paris agreement. Within this context, renewable energy sources (RESs) represent the most promising technology for the transition and the future system. RESs are almost free-emission technologies and, during the last few years, RESs achieved economic competitiveness against conventional energy sources. However, their deployment in traditional power systems is not absent of challenges. The stochastic nature of renewable generation, the non-storable characteristic of electricity in a cost-e ffective way, and the low elasticity in demand associated with its di fficulties to participate in electricity markets [2] make their variability a major issue with a wider impact on smaller systems. Moreover, the final energy consumption will tend to become electric in order to reduce emissions. Thus, future loads will impose new demands and challenges to the power system such as the massive penetration of electric vehicles (EV) to electrify transport.

In order to overcome this problem, the smart grid concept was an accepted solution for some time now. Smarts grids are electricity networks that intelligently integrate their users' actions to e fficiently deliver economic, secure, and sustainable electricity [3]. The implementation of smart grids implies broad and sophisticated functionalities of electric transport and distribution systems, improving their flexibility, allowing bidirectional energy flows, and facilitating RES and demand response (DR) integration. The demand response is based on developing active participation of customers with new requirements that take into account technology and equipment for customer communications, relations, and services. However, just with the participation of demand, the security of supply will still be jeopardized with larger levels of stochastic production associated with renewable generation. Thus, storage systems will also be required to provide flexibility and ensure reliability to the system [4]. Moreover, the batteries' cost reductions make them a key component in the future power systems [5].

Currently, the electricity sector finds itself making three classes of transformations: firstly, the improvement of the current infrastructure; secondly, the addition of the digitalization of power systems, which is the essence of communications and data generation in smart grids; thirdly, business process transformation to perform, in addition to the traditional activities, new ones, or providing infrastructure and data to agents such as aggregators and virtual power plants (VPPs). These agents do new activities related to meeting customer needs and expectations in a more e fficient way than the traditional centralized system. These three transformations were approached in several di fferent ways, which were mainly described on a very abstract level [6] or focused on specific aspects such as just information and communication technology (ICT) [7]. Di fferent standardization bodies developed specific concepts such as the American National Institute of Standards and Technology (NIST) framework and roadmap for smart grid standards [8] and the European Smart Grid Architecture Model (SGAM) [9]. However, the necessary new activities, agents, and interactions among them in the future electricity markets are not clearly defined and authors still characterize them in di fferent ways. Therefore, it is necessary to align specific agents to established practical conceptual architectures as suggested by Neuriter et al., [10].

The functionality of the future power systems and markets may look quite di fferent according to the local social, regulatory, or economic environment. Nevertheless, they have common applications and requirements for digital processing and communications to implement advanced control in all elements of the power system, allowing for bidirectional communication and energy flows [8], understanding the automation of processes and systems as digital processing to retrieve data and perform actions. According to this context, smart grids enable greater information managemen<sup>t</sup> and efficiency compared to conventional power systems, thus allowing the exploitation of the benefits associated with RES, demand response, storage systems, and real-time competition and response in local markets. Local markets are arising as a new mechanism to provide an e fficient allocation and pricing of the growing distributed generation (DG) and flexible demand [11,12].

Thus, smart grids are emerging as a solution for the future of power systems [13]. This broad concept that comprises many di fferent agents, actors, and technology was approached in di fferent ways. Its future faces di fferent problems and sub-problems, which were widely studied. According to Reference [14], some of these are operation and management, energy storage, security, stability, and protection, demand control, or service restoration, among others.

For instance, some authors proposed multi-agent systems that optimize resource scheduling in smart grids [15,16]. These agents enable the system to behave in a more reliable and e fficient way. However, the description of these agents does not follow any standardized premise. The authors of References [17,18] proposed energy managemen<sup>t</sup> systems in smart grids. The agents as in Reference [15] did not include a clear definition of the agen<sup>t</sup> boundaries of action or relationships and presented conflicts between them. A review of agent-based models was presented in Reference [19], where the necessity of harmonization between studies was highlighted.

In order to tackle the previously mentioned standardization problems, di fferent meta-architectures were developed. These conceptual architectures provide a family of ontologies to map smart grids and guidelines on how to use standards [7]. The main two developments were the previously mentioned NIST work and SGAM.

In the United States of America (USA), the NIST created relevant conceptual models for the smart grid. NIST considered the approach that the smart grid can be divided into seven domains [8]. These domains and their sub-domains enclose the conceptual roles and services, including stakeholders, interactions, and types of services. On the other hand, the M/490 working group on reference architectures created the SGAM, which can be seen as a similar effort on the European level. SGAM is based on NIST and proposes a model with five interoperability layers, five domains, and six zones, as can be seen in Figure 1. Thus, every element in the model can be located in a three dimension grid according to its interoperability, domain, and zone characteristics [9]. As in the case of NIST, SGAM requires stronger integration between the design and the use cases and formal semantics [20], as it lacks of precise descriptions.

**Figure 1.** Smart Grid Architecture Model (SGAM) iterations, layers, and planes. Own elaboration based on Reference [8].

Highly correlated with smart grid development, the three novel agents of aggregator, storage, and virtual power plant (VPP) are being developed. In all these cases, several authors published studies on the topic. However, if the case of smart grids is still not clear and no standard definitions are used, VPP, storage, and aggregators offer an even wider range of variation and disagreement. The importance of these three agents is relevant for the conception of smart grids since these agents are crucial for the security and reliability of power systems with increasing levels of renewable penetration [21]. For instance, some authors optimized VPP bidding strategies [22–24], renewable energy integration [25,26], the use of demand response in smart grids [27], or the usage of RESs at the residential level [28,29]. However, there exists a lack of a standardized definition, interactions, and roles performed by a VPP.

Demand response is also stated to have an increasing role in power systems due to its potential capacity to help manage renewable variability [30]. Work was done in analyzing the cost of automated DR systems [31], the suitability of different customers [32], the evaluation of the action performance [33,34], or its optimization in smart grid programs [35]. Moreover, its role among active consumers at the distribution level is gaining importance [36]. Storage is seen as the key technology to enable RES integration in the future power systems [4,37]. Under this paradigm, storage systems are already a key agen<sup>t</sup> in the power system as in the case of the Tesla Battery of South Australia [38]. However, the particularities and services that they provide are far from being homogeneous or clear among scholars and systems. Finally, in a similar line, aggregators were approached in different ways by authors and regulators, but also lack a clear common definition [39]. Moreover, authors do not share a common view on the size that optimal aggregation should have. For instance, while the authors of Reference [40] argued that aggregation is only profitable at large levels, the authors of Reference [41] defended that, even at low levels, aggregation offers benefits. In sum, agents are not clearly defined and the interactions between them vary among authors.

The conceptual architecture here developed is based on the NIST framework [8] and builds on providing the relationships and interaction design between the di fferent agents. These agents can be performed by di fferent entities or one entity, company, or organization that could hold more than one of the agents' responsibilities. Reference levels of power, voltage, and minimum bidding levels were parameterized to be chosen depending on the system, thus providing an easy way to implement the conceptual architecture to any power system. Thus, the proposed conceptual architecture can be applied to any type of power sector, independently of the level of decentralization and its size.

The main contributions of this paper are the following:


The rest of the paper is structured as follows: Section 2 outlines the NIST methodology used for building the proposed design to upgrade the current one. Then, the specific agents proposed for a standardized architecture are developed in Section 3. Finally, in Section 4, some conclusions are drawn.
