3.3.1. Description of Agents

Agents are autonomous systems which can operate in a fully-decentralized manner, i.e., it does not depend on a central (master) element. An agen<sup>t</sup> can represent a group of appliances or bigger grid elements, such as a cluster of households. The utilities can retrieve data collected by the agents and use this information to control the power supply and decide which power sources to use. MAS-based software architecture for micro grids has been experimented in previous studies [35–49,51–53]. Agent-based software architectures are a potential technique in smart grids, hence plenty of research articles exist in this domain.

Existing distribution systems operate as a Master–Slave concept, responding to the central command with no possibility of dynamic reconfiguration. They do not support interoperability, and utilities often are locked into solutions from a single vendor. Multi agents' application in grid environments enables the local devices to be more self-operable and support decentralized monitoring and controlling operations. Data services, functional logic services, and business logic services are described in smart grid architectures. An intelligent agen<sup>t</sup> consists of four components:


### 3.3.2. Agent Based Architecture in Literature

Multi Agent Systems are "autonomous decision makers that communicate their preferences, negotiate sub-goals, and coordinate their intentions in order to achieve the individual or system goals" [54,55]. They are suitable for smart grid architectures as they are intelligent and autonomous, exhibit distributed control, act flexibly to the changes in environment, and cooperate with each other towards a common goal.

Malik et al. analyzed various aspects of agen<sup>t</sup> usage in a smart grid environment [3]. This paper evaluates the agent's application in smart grid functional areas, namely control, fault management, self-healing properties, energy balance management, and distribution side management. Zhabelova et al. investigated various aspects of distributed architectures for smart grid applications and unique contributions within this domain [29,54,56,57]. These works practically implement agent-based solutions for smart grids and facility migration of smart grid technology to the next level of research.

The limitations of current distribution systems are addressed with agent-based techniques [29]. The proposed integration of the two international standards IEC 61499 and IEC 61850 has enabled development of a prototype of the automation architecture, where the intelligence resides at the device level [56,58,59].

An agent-based architecture is proposed for fault location, isolation, and supply restoration (FLISR) applications [15,57]. A hybrid approach combining both the advantages of reactive and deliberative architecture approaches is implemented. This paper deals with two reactive behaviors, five plans, the intention stack, and the interpreter. The design process is methodological and agents are reproducible. An actual fault condition in a substation environment is considered and the mechanism of agen<sup>t</sup> reactions is tested.

The concept of a JIAC (Java Intelligent Agent Component) framework is integrated with a service-oriented paradigm [2]. The framework provides the agent-platform, comprising agen<sup>t</sup> nodes, is physically distributed, and it enables runtime environments for JIAC agents. These are demonstrated in two projects: The Intelligente L¨osungen zum Schutz vor Kaskadeneffekten in voneinander abh¨angigen kritischen Infrastrukturen (ILIas) project, and Gesteuertes Laden V2.0.

Agent-based design in smart grid applications requires a modular software architecture [60]. This work attempts to implement different aspects of communication agents, namely communication perspective, access to services, and handling subscriptions. Agents can have the following attributes: name, parent, child, and unique ID. Protocol adapters are proposed to bridge between agents and appliances.

The Multi Agent Architecture for smart grids is explored in a previous study [61]. This paper defines different roles in the smart grid as producer, storage, consumer, prosumer, and distribution. It takes JADE as a base framework to implement the proposed architecture. The salient features of this research are time dependent monitoring and computation support for internal and external physical models, flexibility in computation, and it eases the integration of distributed control algorithms.

### **4. Future Work and Conclusions**

#### *4.1. Smart Grid Architecture Proposal Based on Data Exchange*

Smart grid architecture is a complex, data intensive, and safety-critical system. The data collection and exchange among consumers, DERs, distribution, and cloud systems decides the basic structure of smart grid architecture. A typical smart grid system is similar to a datacenter system, in that the devices sit is a remote location and require a means for a supervisor or administrator to access the system remotely. Remote access is meant to happen over a secure connection and will also need reliable data transfer, a light, and fast protocol definition, which is one of the focus areas of this document. The bit-level protocol definition ensures fast, reliable, and secure communication in a smart grid environment [62]. Cloud integration techniques also enable leverage of many technologies (automotive, home automation) into the smart grid domain.
