Distributed Cooperative Control With Applications To Smart Grid Operations

The development of complex systems and network science technologies has promoted the coming of the era of "Internet+", and the network has become a strong link which is connecting all walks of life, such as transportation networks, aviation networks, power grids, energy networks, biological networks. The internal nodes of the network are interconnected through a physical coupling or a communication coupling; the information exchange between the network and the network forms a multi-layer structure, such as the network of networks. In the network, we can treat each individual node as an agent, which has the abilities of decision-making and coordinate optimization calculation. All the network nodes have formed a multi-agent system. In this thesis, we are focused on the distributed cooperative control of multi-agent systems, including distributed consensus control, pinning consensus control, distributed containment control; on the other hand, combined with multi-agent scenarios in smart grids, we explore the application of distributed optimization, distribution estimation and the distributed control technologies in the operation of smart grids. The main contents of this dissertation can be summarized as follows:(1) Consensus seeking problems are investigated for linear multi-agent systems. Two kinds of state observers, the decentralized Luenberger observer and distributed pinning networked observer, are proposed to estimate the group agents’ state. Then, based on the observed state information, a novel distributed hybrid output feedback protocol is proposed. By eigenval-ue analysis method, some necessary and sufficient criteria have been established such that the asymptomatic consensus can be achieved for the linear multi-agent system under a fixed directed communication topology. Furthermore, two multi-step algorithms are provided to de-termine the observer gains and the feedback gains for the proposed observer and distributed hybrid protocol.Next, we are concerned with distributed pinning consensus problem for a class of nonlinear multi-agent system with observer-based protocols. Two types of state observers including local observer and distributed pinning observer are proposed for the single nonlinear agent with the first one designed by the local output information and the second one designed via the relative output information of its neighboring agents. According to the state information observed, a distributed pinning observer-based protocol is proposed for the leader-following consensus of the multi-agent system. Furthermore,two multi-step algorithms are presented to construct the observer gains and the protocol parameters for the proposed protocols respectively. It is shown that under the condition that the pinning joint communication topology contains a directed spanning tree, the sufficient criteria established can not only ensure the observation error to be globally asymptotically stable, but also guarantee the consensus of the multi-agent system to be solved asymptotically.(2) The hierarchical cooperative control problem is concerned for a two-layer networked multi-agent system under switching directed topologies. The group cooperative objective is to achieve finite-time formation control for the upper layer of leaders and containment control for the lower layer of followers. Two kinds of cooperative strategies including centralized-distributed control and distributed-distributed control are proposed for two types of switching laws:random switching law with the dwell time and Markov switching law with stationary distribution. By utilizing the state transition matrix methods and matrix measure techniques, some sufficient conditions are derived for asymptotical containment control and exponential almost sure containment control respectively.Then, the containment control problem is considered for nonlinear multi-agent systems with directed communication topology. Under the guidance of designed distributed communica-tion protocols with/without previous state information, the followers are expected to converge to a dynamic convex hull spanned by multiple leaders. Two multi-step algorithms are proposed to construct the corresponding protocols, the state feedback protocol and the delay-coupled protocol, under which the containment control can be achieved asymptotically. Furthermore, it is found that the delay-coupled protocol is rather sensitive to time delays. That is, real-time tracking will become impossible by only using long-dated previous state information.(3) Demand-side resources have considerable potential, which can provide efficient load shedding operations so as to maintain the supply-demand balance together with traditional generating units. Here, we are concerned with active power dispatch and control problems by considering both generating units and demand-side resources, where some controllable loads in demand-side are aggregated under each bus load agent. We propose a multi-level dispatch and control architecture for power systems with demand-side resources, which can handle the optimization and control problems in different time scales. Specifically, an upper layer performs calculation of system’s look-ahead dispatch for generating units and bus load agents, while a middle level computes control actions for multiple load clusters in each bus load agent via a distributed consensus estimation algorithm, and multiple energy management units in the lower layer respond the control instruction issued from the load cluster based on a distributed pinning control algorithm. For the middle and lower layers, participants can communicate with its neighbors via sparse communication networks for the sake of distributed calculation or distributed control, which is more robust than the centralized strategies in terms of dealing with massive amounts of participation units.Next, a novel distributed optimal dispatch algorithm is proposed for coordinating the op-eration of multiple micro units in a microgrid, which has incorporated the distributed consensus algorithm in multi-agent systems and the λ-iteration optimization algorithm in economic dis-patch of power systems. Specifically, the proposed algorithm considers the global active power constraint by adding a virtual pinner and it can deal with the optimization problem with any initial states. That is, it can realize the global optimization and avoid the defect of the initial conditions’sensitivity in the optimization problem. On the other hand, the proposed optimiza-tion algorithm can either be used for off-line calculation or be utilized for on-line operation and has the ability to survive single-point failures and shows good robustness in the iteration process.(4) In the power demand side, responsive loads can provide fast regulation and ancillary services as reserve capacity in interconnected power systems. We present a distributed pinning demand side control (DSC) strategy for coordinating multiple load aggregators, i.e., aggregated responsive loads, to provide frequency regulation services. Specifically, a leader-following com-munication protocol is considered for the load aggregators, in which there is a centralized pinner (leader) and multiple load aggregators (followers). The regulation objective is generated from the pinner and only shared with a small fraction of load aggregators. Moreover, a multi-step algorithm is proposed to determine the control gains in the DSC, which not only guarantees the stability of the close-loop system, but also restrains the plant disturbance. Furthermore, the distributed pinning DSC algorithm is integrated into the traditional centralized PI AGC framework, which has formed the coupled secondary frequency control structure. It has been shown that the total power mismatch in each control area is shared with both AGC units and load aggregators and the system frequency can be improved by considering the distributed pinning DSC for load aggregators.Furthermore, we propose a distributed load following strategy for multiple aggregated thermostatically controlled loads (TCLs). In demand response programmers, these aggregated TCLs are often affiliated to a load agent, which can denote all the thermostatically controlled loads (such as, air conditioners, water heaters) in an intelligent residential district. Specifically, each TCL aggregator is described by an approximate bilinear model with the temperature setpoint as the control input and aggregated power as the output signal. Based on the spare communication links, each aggregator just needs to communicate with its neighbors and partial aggregators are connected to the load agent, and then the total power tracking of the load agent can be achieved by a distributed pinning control strategy. Such a distributed strategy is fairly robust and has good scalability performances, which can realize the plug and play of TCL aggregators. The reference power trajectory for the load following service can be derived from the dispatch center optimized by the dispatch costs of load agents.