Resilient control of multi-microgrids against false data injection attack

Highlights

• Multi-microgrid systems, including several microgrids and distributed energy resources are investigated.

• The proposed method considers numbers of faults and attacks as a consequence of which malfunctioning can occur on a large scale.

• False data injection and replay attack by considering the multi-microgrid as a multi-agent system is considered and the effects are minimized.

• The malicious agents become isolated with the help of Weighted Mean Subsequence Reduced (W-MSR) algorithms.

• The proposed controller is able to maintain the system’s desired performance when false data is injected to the system or valid data is received with time-delays.

Abstract

A multi-microgrid system, including several microgrids and distributed energy resources, is always threatened by numbers of faults and attacks as a consequence of which malfunctioning can occur on a large scale. Thus, minimizing the effects of such disruptions is of paramount importance. This paper addresses the problem of mitigating a multi-microgrid system that faces false data injection and replay attacks by considering the multi-microgrid as a multi-agent system in which each microgrid as an agent represents a node in a weighted directed graph. The problem of consensus among normal agents is studied when microgrids and their communications are attacked. The malicious agents become isolated with the help of Weighted Mean Subsequence Reduced (W-MSR) algorithms in which all normal agents neglect the extreme values received from their neighbors. The proposed controller is able to maintain the system’s desired performance when false data is injected into the system, or valid data is received with time-delays. Finally, numerical examples and simulation results are provided.

Introduction

As the world is gradually decreasing the usage of conventional energy resources due to their pollution, causing the greenhouse effect, and other innumerable environmental harms, microgrids are substituting these resources. A microgrid system can be defined as a group of distributed power resources that can operate independently from the main grid; and a multi-microgrid system is a group of microgrids with several distributed generation units connected to the grid [1]. In recent years, numerous advantages of employing microgrids rather than traditional energy resources in terms of reliability and cost-efficiency have attracted scholars’ attention. Besides, distributed energy resources (DERs), which are energy units in distributed locations and responsible for producing the network’s whole energy, can be used in microgrid systems. Consequently, the microgrid systems turn to environment-friendly power generating facilities since DERs are capable of using various kinds of energy resources like solar energy, wind power, water turbine, and nuclear energy. In this regard, microgrids can perform without causing any damages to the environment.

Even though the merits of using DERs and microgrids cannot be denied, these systems face many threats, faults, and adversarial attacks. Computational resource scarcity, communication constraints, and sensitive information in data cause many cybersecurity challenges that microgrids face [2]. Moreover, detection of faults and attacks in DERs can as well be quite challenging owing to the fact that they operate dynamically and are based on nature which leads them to be unpredictable [3]. As an example, the penetration of generations that generate energy from resources such as solar or wind, can differ based on their place, the system’s characteristics, and other factors.

A set of agents that interact with one another is known as a multi-agent system (MAS). Using such systems help complex problems become solved by dividing them into several simple tasks [4]. Due to intelligent interactions between agents, MASs are capable of solving problems that are unsolvable by a single agent. Microgrid systems are complex systems and their problems cannot usually be solved with common methods. Moreover, using DERs, increases the system’s complexity [5]. According to previous studies (like [6], [7]), microgrid systems can be modeled as MASs such that each microgrid is considered as an agent. Consequently, to overcome the challenges with microgrid and multi-microgrid systems, employing multi-agent technology is highly recommended. In this paper, a multi-microgrid system is considered as a multi-agent system in which each microgrid represents an agent. The model of MAS is shown by a directed graph representing agents as nodes and the communications with paths between them.

Owing to the fact that components in some systems may be limited in transmission power and memory, designing distributed algorithms that enable performing a global task with only having access to local information and limited computations is crucial. These kinds of problems are usually addressed in the area of consensus problems [8]. In a network consisting of agents communicating with each other, reaching an agreement based on each agent’s value and the ones with which it is communicating, is called reaching consensus [9]. When power energy resources are being used with the aim of balancing the production and demand in the system, reaching such an agreement is essential.

While multi-microgrid systems are used, not only detecting attacks is crucial but also employing safe control strategies with the aim of providing appropriate protection from the system is also of high importance [10]. Some security objectives of the system may be threatened while the system is affected by an adversarial attack. Data confidentiality, integrity, and authenticity are some of these objectives that have the highest priority for the transmitting data in networked systems [11]. Data confidentially is the ability to protect data from unauthorized access. Data integrity refers to the ability of maintaining and assuring the accuracy and consistency of data, and data authenticity is defined as the ability to receive the original data and exactly as it was sent. Various types of attacks threaten these abilities of the system. For instance, some kinds of deception and false data injection attacks damage the system by threatening confidentiality and integrity of data. This is while kinds of replay attacks harm the system by preventing it from receiving the original data at the appropriate time. Besides, the other important parameter is the diverse parts of the system on which attacks have effect. Attack on the nodes’ dynamics and on the out-going communications of a node can be considered as two different scenarios that can impair networked systems [12].

This paper considers two types of attacks, one of which is false data injection and the other is replay attack. In the first type, it is assumed that the agents’ dynamics are threatened. Therefore, no control signal can be applied to the agents that are attacked and behave in an adversarial manner. On the contrary, the replay attack discussed in this paper affects the communications between nodes by not letting the original data be received immediately.

The ability of a system to adapt and survive when being vulnerable to faults and attacks is known as the resiliency of that system. The resilient control concept has been studied in various publications in recent years. Due to the significant increase in cyber–physical systems and thus, an increase in attacks, faults, and vulnerabilities in such systems, different types of resilient controllers, employing diverse methods and strategies, have been designed to protect such systems. For instance, authors in [13], [14] have used adaptive controllers, in [15] game theories have been studied, in [16] a resilient controller is designed with the help of self-healing strategies, and in [17] optimal control methods are discussed as one of these strategies. The authors of this paper utilize graph theories and concepts to design a resilient controller for a multi-microgrid system.

There are different approaches to achieve control performance under the attacks. For instance, in [18], [19], the controller is designed based on integrating the attacks’ information. As another example, in [20], giving different reputations to agents is utilized to achieve proper control performance. In this article, the control performance is achieved by disregarding extreme values compared to each agent’s values in all the time steps.

In multi-agent systems, sometimes, only local information is available for agents. This means that only information about their own and their neighbors’ states and values is available for agents. For such systems, authors in [21] have proposed mean-subsequence-reduces (MSR) algorithms as a resilient control strategy to overcome security and protection challenges. Such algorithms are discussed in [22], [23]. MSR algorithms, which are followed by consensus algorithms lead the agents’ states to reach an agreement asymptotically even though faults and attacks occur. The significant advantage of these kinds of algorithms is that the system can be recovered under undesirable circumstances even though agents may not be aware of the whole system’s model, communications between all agents, and all other agents’ values. The merits regarding these types of algorithms have attracted researchers’ attention. For instance, in [24], applying W-MSR algorithms to switched multi-microgrids have been studied, and in [25], opinion shifting strategies have been designed with the help of W-MSR algorithms which provide resilience in some adversarial environments.

Applying a novel approach of MSR algorithm followed by consensus algorithm to a multi-microgrid system containing microgrids with multiple environment-friendly renewable distributed energy resources with the aim of detecting disruptions in such system and mitigating it is discussed in this paper. Considering the multi-microgrid system as a MAS and modeling it with a weighted directed graph, the necessary and sufficient conditions to maintain the system’s resiliency in the occurrence of false data injection attack and replay attack have been provided.

Regarding the wide-range effects of microgrids, the criticality of energy storage and demand, and the complexity of these systems, balancing the energy level in the presence of adversarial attacks or unintended faults is important, a goal which has been achieved in the paper.

The paper is organized as follows: In Section 2, preliminaries and the algorithm are presented. In Section 3, the focus is on resiliency of multi-microgrids under different circumstances. Section 4, is devoted to the numerical examples. Finally, the conclusions are given in Section 5.