Data-Driven Predictive Control for Ensuring Grid Security with High Penetration of Wind Energy

为了在英国实现净零碳排放，发电必须完成从以化石燃料为基础的生产向可再生能源(RES)的过渡。由此导致的RE渗透率的增加使电网的运营更具挑战性。对风能和太阳能发电天气的依赖带来了供应能力的不确定性。在传统发电厂中，旋转涡轮机的惯性有助于维持电网稳定。然而，RE使用被称为逆变器的电力电子设备来提供所需的电压和频率。逆变器不会给电网提供惯性，这使它更容易受到不稳定的影响。计划中的研究重点是通过开发新的控制方法和网络架构来解决电网稳定性和供应与消费者需求匹配的挑战。一种名为直接数据驱动控制的新范式已开始在研究界获得吸引力。系统输入是直接从收集的数据中找到的，而不是通过使用数据估计近似模型，然后使用系统行为的模型预测来确定最佳输入。到目前为止，只有一小部分直接数据驱动控制策略应用于电力系统场景中进行了仿真。这种方法的潜在应用范围从控制功率转换器中的信号到管理风力发电场。这项拟议的研究旨在评估直接数据驱动方法在控制电网中的适用性，并确保在高RE渗透率的情况下电网稳定。研究计划包括对由数据驱动控制器管理的电力网络进行模拟。将模拟多种网络场景，以评估该方法和最佳控制器设计的适用性。这项研究的目的是提高对哪种方法可以最好地处理系统中的不确定性和非线性的理解，因为这目前是一个悬而未决的问题。另一个目标是在实验室环境中的小规模微电网上测试直接数据驱动控制策略。净零过渡的一个标志是通过当地小规模生产方法(如屋顶太阳能电池)进行的分布式发电日益普遍。与传统的集中式方法相比，这提供了更大的操作灵活性，但需要在控制和通信技术方面进行重大改进，才能充分利用其好处。出现了一个称为全息方法的概念，它可能会带来这种改进。一个全息网络是由形成一个整体结构的合子组成的；它类似于一个层次结构。Holon既是系统的一部分，也是系统本身的一部分；智能家居管理家庭内部的电力需求，同时作为地区电网的一个元素存在。整体架构能够在运行期间进行调整，以平衡竞争需求和管理网络故障。关于电力网络的这种实现，目前已发表的研究有限，就作者所知，几乎没有关于合弄自动适应网络变化的能力的研究。拟议的研究旨在解决这些差距，并开发一种可实现的管理电力网络的体系结构的方法。另一个目标是开发概述完整体系的基本行为的理论，该理论可用于为特定用例设计完整体系。直接数据驱动控制和全息控制方法可以通过使用运行数据来优化网络结构和确定适当的控制输入。直接数据驱动控制和全息控制方法都显示出支持将今天的电网发展到未来网络的潜力，将提供净零电网；向可持续的全球电网过渡的关键部分

In order to achieve net-zero carbon emissions in the UK, electricity generation must complete the transition from fossil fuel-based production to Renewable Energy Sources (RES). The resulting increase in penetration of RES makes operation of the power grid more challenging. Reliance on the weather for wind and solar generation introduces uncertainties in supply capability. In traditional power plants, the inertia of rotating turbines helps maintain grid stability. However, RES use power electronics known as inverters to provide the required voltage and frequency. Inverters do not provide inertia to the power grid which makes it more susceptible to instability. The planned research is focused on addressing the challenges of grid stability and of matching supply to consumer demand through the development of novel control approaches and network architectures.A new paradigm known as direct data-driven control has begun to gain traction in the research community. System inputs are found directly from gathered data rather than by estimating an approximate model using data and then determining the optimal inputs using model predictions of system behaviour. This results in theoretical improvements in performance.To date, only a small number of direct data-driven control strategies applied to power system scenarios have been simulated to the author's knowledge. Potential applications for this approach range from controlling signals in power converters to managing a wind farm. The proposed research aims to assess the suitability of direct data-driven methods for controlling power networks and to ensuring grid stability with a high penetration of RES. The research plan involves carrying out simulations of power networks managed by a data-driven controller. Multiple network scenarios will be simulated to assess the suitability of the approach and the optimal controller design. The research aims to improve understanding of which methods may best cope with uncertainties and nonlinearities in the system as this is currently an open question. A further goal is to test the direct data-driven control strategy on a small-scale microgrid within a laboratory setting.One marker of the net-zero transition is the increasing prevalence of distributed electricity generation through local, small-scale production methods such as roof-mounted solar cells. This offers greater operational flexibility compared to the traditional centralized approach but requires a significant improvement in control and communications technology to take full advantage of its benefits.A concept known as the holonic approach has emerged that may deliver such an improvement. A holonic network is composed of holons forming a holarchy; it is analogous to a hierarchy. A holon is both part of a system and a system in itself; a smart home manages power demands within a home whilst existing as an element of the district power grid. The holarchy is able to adapt during operation to balance competing demands and manage network faults. Limited research has been published concerning such an implementation for power networks and almost no research has been published to the author's knowledge regarding the ability of holons to automatically adapt to changes in the network.The proposed research aims to address these gaps and develop the approach towards an implementable architecture for managing power networks. A further aim is to develop theory outlining the fundamental behaviour of a holarchy that could be used in designing a holarchy for specific use cases. The direct data-driven control and holonic approaches could be combined by using operational data to optimize the network structure and determine appropriate control inputs.Both direct data-driven control and the holonic approach show potential to support the development of today's power networks to the networks of the future that will deliver a net-zero power grid; a crucial part of the transition to a sustainable global e