Online Distributed Optimal Control Designs for Wide-Area Power System Networks

摘要

In this dissertation, we develop distributed control designs for oscillation damping in wide-area power system networks. The power system, with a state-feedback communication graph for damping control, is treated as a cyber-physical system (CPS). On this CPS, the following critical cyber-physical constraints are considered - the actuation constraints on the generator excitation controllers (physical constraint), and the communication constraints on the state-feedback architecture for control (cyber constraint). Using the fact that the wide-area oscillations in power systems are caused by the slow inter-area frequency modes, we design online disturbance-aware optimal controllers which promote sparsity in the feedback communication network by only including the most influential generators in its topology, based on the post-disturbance state of the system. Hence the proposed communication topology is adaptive with respect to the location and strength of the incoming disturbance.;In Chapter 2, we first propose a three-step strategy for decentralized state estimation of all generators in the power system, using measurements from strategically placed Phasor Measurement Units in the power grid. In Chapter 3, we design a centralized Model Predictive Controller (MPC) for selective modal damping in the power system. The MPC is designed in frequency domain so that the control energy can be focused on only the most excited inter-area modes. Next, in Chapter 4 we identify the sets of generators which are influential in damping of the most excited inter-area oscillation modes. This is done by analyzing the post-disturbance state of the system, which in turn depends on the unknown strength and location of the incoming disturbance. A sparse feedback communication topology is constructed based in these sets of in uential generators.;In Chapter 5, we design a distributed MPC such that both the cyber and physical constraints on the CPS are respected. Using a distributed cloud architecture, details on the implementation of the controllers are provided. In Chapter 6, we relax the actuator constraints on the controllers, and design a sparse Linear Quadratic Regulator (LQR) with guarantees on the closed-loop stability. Sparsity is promoted in the following two ways. First, structural constraints on the feedback matrix are imposed on the LQR problem by involving only the influential generators for feedback control. Secondly, ℓ 1-regularization of the LQR cost is done to promote further sparsity within the communication links. In Chapter 6, we analyze the closed-loop CPS resiliency to Denial-of-Service (DoS) cyber attacks on the communication network. It is shown that such DoS cyber attacks can be mitigated by redesigning our sparse controller online, i.e. right after the attack is detected.