状态观测机制下多智能体系统的集群控制研究及应用

Group control of Multi-agent systems (MASs) has been a hot research topic in the last decade, while the distributed technique has unique advantages in dealing with scalability, plug-and-play and robustness of coupled systems. This project considers the distributed group control problem of multi-agent system under state-observation mechanism and its application in the demand side control of thermostatically controlled loads in smart grids. Firstly, a decentralized full-order/reduced-order observer and a distributed parallel full-order / reduced-order observer are designed for linear, bilinear and nonlinear multi-agent systems and different types of state observer designing algorithms are provided. Based on the observed state value of multi-agent system, distributed group control protocols, such as distributed consensus protocol、distributed containment control protocol, are designed so as to achieve the aggregate distributed consensus tracking or aggregate interval distributed containment tracking of multi-agent system. Integrated linear system theory, complex network theory, Lyapunov stability theory, we analysis the stability of the coupled close-loop observation and control system, and propose the distributed protocol designing algorithms. As for the application of distributed technique, we consider the active ancillary service control of the massive flexible thermostatically controlled loads on the demand side in smart grid. By load clustering, the aggregate agent model is established according to the type of the auxiliary service and the distributed group control algorithm is applied to realize the distributed active load shifting and distributed active load interval following service, which has laid a theoretical foundation for the plug-and-play function of the practical load aggregators.

多智能体系统的集群控制是近十年来学术界的研究热点问题，其中分布式技术在处理可延展性、即插即用、鲁棒性等方面具有独特的优势。本项目考虑状态观测机制下多智能体系统的分布式集群控制问题及其在智能电网温控负荷控制中的应用。首先，以线性和非线性多智能体系统为研究对象，设计分散式局部全/降阶观测器、分布式并行全/降阶观测器，给出不同类型的状态观测算法；基于观测的多智能体系统状态值，设计分布式的集群控制协议，实现多智能体系统的分布式一致性控制与分布式包含控制。综合线性系统理论、复杂网络理论、Lyapunov稳定性理论，分析耦合观测控制系统的稳定性，给出分布式协议设计算法。从智能电网海量温控负荷的有功辅助服务控制出发，通过负荷聚类，依据辅助服务类型建立聚合智能体模型，采用分布式的集群控制算法，实现多个负荷聚合智能体的分布式负荷削减、分布式有功负荷区间跟随服务，为实际负荷聚合体的即插即用功能奠定理论基础。

本项目围绕多智能体系统的集群建模、分布式优化、分布式控制理论与智能电网中的需求响应柔性负荷集群的优化、控制应用展开系统深入的研究，并在平面多智能体系统的运动稳定性分析，有限时间和固定时间一致性控制算法设计、基于分布式状态观测的分布式优化求解，基于分布式状态观测的线性矩阵方程组求解等方面扩展了研究。与此同时，在应用领域，探究了聚合柔性空调负荷集群、聚合柔性热水器负荷集群、聚合柔性电动汽车集群的集中式、分布式负荷削减与负荷跟随控制算法，实现了柔性负荷参与需求响应的互动运营，为实际的负荷聚合调度控制提供了技术支撑。在项目资助下，研究成果发表在诸如IEEE Trans. Auto.等《SCI》期刊收录论文10篇，中文核心期刊1篇，外文专著1部。项目执行期间，项目负责人获得江苏省科技进步二等奖1项，协助培养博士生2名，硕士生3名。项目投入经费25万元，支出16.33万元，各项支出与预算相符合。剩余经费8.67万元，计划用于本项目的后续研究支出。