ContRol methods for rELiable sensIng informAtion in interConnected Energy systems

物联网（IoT）处于电力和能源系统转型的最前沿，通过实现实时监控和分布式控制等新功能，为全球经济可持续增长提供清洁能源。智能电表、灵活需求、车辆到电网技术以及智能建筑所带来的令人兴奋的机遇都依赖于能够访问大量实时数据。由于传感和通信技术的进步和可负担性，这一点现在已经成为可能。然而，这些系统的重要性和有效性取决于所感测、传达和处理的数据的及时性和准确性。如果使用的信息不可靠，例如由于传感器故障、通信问题或网络攻击，会发生什么？事实上，价格实惠的传感器可能容易出现传感器故障，导致测量数据丢失或不正确;对始终连接的设备的需求可能会受到通信问题的影响，例如延迟或数据包丢失，导致信息过时或丢失。最后，针对工业控制系统的新型复杂网络攻击（称为网络物理攻击）可能会故意修改某些信息，从而对系统造成物理后果。最近在乌克兰发生的袭击导致配电中断，表明了这些袭击的可行性和可怕影响。通过从监控传感器和设备中获取测量值，获取信息以做出决策，随后定义系统的操作，并重复此循环，物联网系统实现了所谓的反馈控制。使用过时或受损的数据可能会导致低效的解决方案，甚至是危险的操作条件。在这样的框架内适当处理控制系统的能力是必不可少的：可靠的传感信息是新兴能源系统的基础，以及可靠的控制系统。拟议的计划提供了一个关键的开放研究问题的答案：如何安全有效地控制基于物联网的新兴能源系统应用，在那里可能具有挑战性，以保证传感信息的可靠性？事实上，现有的方法并不适合这种新的互联和复杂的场景。本项目的目标是设计新的方法来监控传感信息的可靠性，包括传感器异常检测和定位，以及新的控制架构，以适应可能不可靠的传感信息，特别是针对互联物联网场景，如电动汽车充电，微电网和智能建筑的需求和能源管理。为了实现这些目标，直觉是增强基于优化的分布式系统的传统控制方法，并在图上使用创新的机器学习技术。这些方法非常适合所考虑的能源系统，可以表示为一个网络的互联子系统与负载，发电机，存储，设备和传感器。基于图的学习技术将利用系统的已知网络结构来识别网络不同元素之间的关系，并估计和重建丢失或受损数据的价值。这一想法代表了系统控制研究中的一种新颖性。所开发的方法将被系统运营商、中小企业和从事传感和物联网领域的ICT公司采用，以提高其系统的可靠性，保护运营商和用户，从而能够引入高效和绿色能源系统的新技术，从而在安全方面为社会带来巨大利益。韧性和可持续性。

The Internet of Things (IoT) is at the forefront of a transformation in electric power and energy systems to provide clean energy for sustainable global economic growth, by enabling novel capabilities, such as real-time monitoring and distributed control. The exciting opportunities given by smart meters, flexible demand, vehicles to grid technologies, and smart buildings all rely on having access to a large amount of real-time data. This is nowadays possible thanks to the advancements and affordability of sensing and communication technologies. However, the importance and effectiveness of these systems relies in the timeliness and accuracy of the data which is sensed, communicated and processed. What happens if the used information is not reliable, for example due to sensor faults, communication problems or cyber-attacks? In fact, affordable sensors could be prone to sensor faults, leading to missing or incorrect measurements; the need for always connected devices could be compromised by communication issues, such as delays or packet losses, resulting in outdated or missing information. Finally, novel sophisticated cyber-attacks, called cyber-physical attacks, targeting Industrial Control Systems, may intentionally modify some information to cause physical consequences on the systems. Recent attacks in Ukraine resulting in the disruption of power distribution have shown the feasibility and terrible effects of these attacks. By taking measurements from monitoring sensors and devices, deriving information to take decisions and subsequently defining actions for the system, and repeating this cycle, IoT systems implement a so-called feedback control. The use of outdated or compromised data could lead to inefficient solutions or even dangerous operation conditions. The ability to appropriately deal with control systems within such frameworks is an imperative: reliable sensing information is fundamental for emerging energy systems, as well as reliable control systems.The proposed programme provides answers to a key open research question: How to safely and efficiently control emerging energy systems applications based on the IoT, where it might be challenging to guarantee the reliability of the sensing information? In fact, existing methods are not suitable for this novel interconnected and complex scenario.The goal of this project is to design novel methods to monitor the reliability of sensing information, including sensors anomaly detection and localisation, and new control architectures resilient to possibly unreliable sensing information, specifically for interconnected IoT scenarios such as electric vehicles charging, demand and energy management in microgrids and smart buildings. To achieve these objectives, the intuition is to enhance traditional control methods for distributed systems based on optimisation with innovative machine learning techniques on graphs. These methods well suit the considered energy systems that can be represented as a network of interconnected subsystems with loads, generators, storage, devices and sensors. Graph-based learning techniques will exploit the known network structure of the system to identify the relationships between the different elements of the network and to estimate and reconstruct the value of missing or compromised data. This idea represents a novelty in the research for systems control.The developed methodologies will be adopted by systems operators, SMEs and ICT companies working in the sensing and IoT sectors for energy, to enhance the reliability of their systems, to protect operators and users, enabling the introduction of novel technologies for efficient and green energy systems, thus bringing a huge benefit to the society in terms of safety, resilience and sustainability.